Health Maxhxchxhnagement

Preface
3. Performance improvement tools, techniques, and programs; 4. Applications to c

ontemporary healthcare operations issues; and 5. Putting it all together for ope
rational excellence. Although this structure will be helpful for most readers, e
ach chapter stands alone and can be covered or read in any order that makes sens
e for a particular course or student. The first part of the book, Introduction t
o Healthcare Operations, provides an overview of the challenges and opportunitie
s found in todays healthcare environment (Chapter 1). We follow that with a histo
ry of the field of management science and operations improvement (Chapter 2). Ne
xt, we discuss two of the most influential environmental changes facing healthca
re today, evidence-based medicine and pay-for-performance (Chapter 3). In Part I
I, Setting Goals and Executing Strategy, Chapter 4 highlights the importance of
tying the strategic direction of the organization to operational initiatives. Th
is chapter outlines the use of the balanced scorecard technique to execute and m
onitor these initiatives to achieve organizational objectives. Typically, strate
gic initiatives are large in scope, and the tools of project management (Chapter
5) are needed to successfully manage them. Indeed, the use of project managemen
t tools can help to ensure the success of any size project. Strategic focus and
project management provide the organizational foundation for the remainder of th
is book. The next part of the book, Performance Improvement Tools, Techniques, a
nd Programs, provides an introduction to the basic decision-making and problem-s
olving processes and describes some of the associated tools (Chapter 6). Almost
all process improvement initiatives (Six Sigma, Lean, etc.) follow this same pro
cess and make use of some or all of the tools in Chapter 6. Good decisions and e
ffective solutions are based on facts, not intuition. Chapter 7 provides an over
view of data collection and analysis to enable fact-based decision making. Six S
igma, Lean, simulation, and supply chain management are more specific philosophi
es or techniques that can be used to improve processes and systems. The Six Sigm
a (Chapter 8) methodology is the latest manifestation of the use of quality impr
ovement tools to reduce variation and errors in a process. The Lean (Chapter 9)
methodology is focused on eliminating waste in a system or process. Many healthc
are decisions and processes can be modeled and optimized or improved by using co
mputer simulation (Chapter 10). The fourth section of the book, Applications to
Contemporary Healthcare Operations Issues, begins with an integrated approach to
applying the various tools and techniques for process improvement in the health
care environment (Chapter 11). We then focus on a special but important case of
process improvement, patient scheduling in the ambulatory environ-
Pre f a c e
xix
ment (Chapter 12). Supply chain management extends the boundaries of the system
to include both upstream suppliers and downstream customers, and this is the foc
us of the final chapter in the section (Chapter 13). The books last section, Putt
ing It All Together for Operational Excellence, concludes the book with a discus
sion, in Chapter 14, of strategies for implementing and maintaining the focus on
continuous improvement in healthcare organizations. We have included many featu
res in this book that we believe will enhance student understanding and learning
. Most chapters begin with a vignette, called Operations Management in Action, t
hat offers real-world examples related to the content of the particular chapter.
Throughout the book we use a fictitious but realistic organization, Vincent Val
ley Hospital and Health System, to illustrate the various tools, techniques, and
programs discussed. Each chapter concludes with questions for discussion, and P
arts II through IV include exercises to be solved. We have included many example
s throughout the text of the use of various contemporary software tools essentia
l for effective operations management. Healthcare leaders and managers must be e
xperts in the application of these tools and stay current with their latest vers
ions. Just as we ask healthcare providers to stay up to date with the latest cli
nical advances, so too must healthcare managers stay current with these basic to
ols.
Student Resources
We have developed an extensive companion website with links to a vast amount of
supplementary material. This website, ache.org/books/OpsManagement, provides lin
ks to material we have developed, as well as other supplemental material. In par
ticular, we have developed, and make available, various Excel templates, Arena m
odels, tutorials, exercises, and PowerPoint presentations for each chapter. Addi
tionally, links to many of the cited articles and books can be found on this web
site. Finally, the site provides links to a wide variety of information and mate
rial related to specific topics, including videos, webcasts, web demonstrations,
exercises, and tutorials. Because new and valuable information is constantly be
ing added to the web, we encourage readers to share any relevant sites they find
so that we can include them on the companion website. A password is not necessa
ry for access to the website. Included with this book is the student version of
Arena software. This simulation software provides a powerful tool to model and o
ptimize healthcare processes such as patient flow and scheduling. The animation
features of Arena provide the healthcare professional with an accessible and eas
ily understood tool to engage colleagues in the intense and complex work of oper
ations improvement.
xx
Preface
Instructor Resources
For instructors who choose to use this book in their courses, accompanying resou
rces are available online. For access information, e-mail hap1@ache.org. Contain
ed in these resources are answers or discussion points for the end-ofchapter que
stions and exercises; teaching tips; and recommended teaching cases, with links
to sources as needed. We hope that this text is helpful to you and your organiza
tion on your journey along the path of continuous improvement. We are interested
in your progress whether you are a student of healthcare administration, new me
mber of the health administration team, seasoned department head, or physician l
eader. Please use the e-mail addresses on the companion website to inform us of
your successes and let us know what we could do to make this a better text so th
at we, too, can continuously improve. Dan McLaughlin Julie Hays
PART
I
introduction to healthcare operations
CHAPTER
1
THE CHALLENGE AND THE OPPORTUNITY
CHAPTER OUTLINE
Introduction Purpose of this Book The Challenge The Opportunity Evidence-Based M
edicine Knowledge-Based Management A More Active Role for the Consumer A Systems
Look at Healthcare The Clinical System System Stability and Change An Integrati
ng Framework for Operations Management in Healthcare Vincent Valley Hospital and
Health System Introduction to Healthcare Operations Setting Goals and Executing
Strategy Performance Improvement Tools, Techniques, and Programs Applications t
o Fundamental Healthcare Operations Issues Conclusion Discussion Questions Refer
ences
2
KEY TERMS AND ACRONYMS

Agency for Healthcare Research and Quality (AHRQ) consumer-directed healthcare e
vidence-based medicine (EBM) health savings account Institute of Medicine (IOM)
knowledge-based management (KBM) patient care microsystem Vincent Valley Hospita
l and Health System (VVH)
3
Introduction to Healthcare Operations

Introduction
The challenges and opportunities in todays complex healthcare delivery systems de
mand that leaders take charge of their operations. A strong operations focus can
reduce costs, increase safety, improve clinical outcomes, and allow an organiza
tion to compete effectively in an aggressive marketplace. In the recent past, th
e success of many organizations in the American healthcare system has been achie
ved through executing a few key strategies: First, attract and retain talented c
linicians; next, add new technology and specialty care; and finally, find new me
thods to maximize the organizations reimbursement for these services. In most org
anizations, new servicesnot ongoing operationsrepresented the key to success. Howe
ver, that era is ending. Payer resistance to cost increases and a surge in publi
c reporting on the quality of healthcare are strong forces driving a major chang
e in strategy. To succeed in this new environment, a healthcare enterprise must
focus on making significant improvements in its core operations. This book is ab
out how to get things done. It provides an integrated system and set of contempo
rary operations improvement tools that can be used to make significant gains in
any organization. These tools have been successfully deployed in much of the glo
bal business community for more than 30 years (Hammer 2005) and now are being us
ed by leading healthcare delivery organizations. This chapter outlines the purpo
se of the book, identifies challenges that current healthcare systems are facing
, presents a systems view of healthcare, and provides a comprehensive framework
for the use of operations tools and methods in healthcare. Finally, Vincent Vall
ey Hospital and Health System (VVH), which is used in examples throughout the bo
ok, is described.
Purpose of this Book
Excellence in healthcare derives from three major areas of expertise: clinical c
are, leadership, and operations. Although clinical expertise and leadership are
critical to an organizations success, this book focuses on operations how to deliv
er high-quality care in a consistent, efficient manner. Many books cover operati
onal improvement tools, and some focus on using these tools in healthcare enviro
nments. So, why a book devoted to the broad topic of healthcare operations? Beca
use there is a real need for an integrated approach to operations improvement th
at puts all the tools in a logical context and provides a road map for their use
. An integrated approach 4
Chapter 1: The Challenge and the Opportunity

5
uses a clinical analogyfirst find and diagnose an operations issue, then apply th
e appropriate treatment tool to solve the problem. The field of operations resea
rch and management science is too deep to cover in one book. In Healthcare Opera
tions Management, only tools and techniques that are currently being deployed in
leading healthcare organizations are covered in enough detail to enable student
s and practitioners to get things done in their work. Each chapter provides many r
eferences for deeper study. The authors have also included additional resources,
exercises, and tools on the website that accompanies this book. This book is or
ganized so that each chapter builds on the next and is cross-referenced. However
, each chapter also stands alone, so a reader interested in Six Sigma could star
t in Chapter 8 and then move back and forth into the other chapters. This book d
oes not specifically explore quality in healthcare as defined by the many agencies
that have a mission to ensure healthcare quality, such as the Joint Commission,
National Committee for Quality Assurance, National Quality Forum, or federally
funded Quality Improvement Organizations. The Healthcare Quality Book: Vision, S
trategy and Tools (Ransom, Maulik, and Nash 2005) explores this perspective in d
epth and provides a useful companion to this book. However, the systems, tools,
and techniques discussed here are essential to make the operational improvements
needed to meet the expectations of these quality-assurance organizations.
The Challenge
The United States spent more than $2 trillion on healthcare in 2007the most per c
apita in the world. With health insurance premiums doubling every five years, th
e annual cost for a family for health insurance is expected to be $22,000 by 201
0all of a workers paycheck at ten dollars an hour. The Centers for Medicare & Medi
caid Services predict that within the next decade, one of every five dollars of
the U.S. economy will be devoted to healthcare (DoBias and Evans 2006). Despite
its high cost, the value delivered by the system has been questioned by many pol
icymakers. Unexplained variations in healthcare have been estimated to result in
44,000 to 98,000 preventable deaths every year. Preventable healthcare-related
injuries cost the economy between $17 billion and $29 billion annually, half of
which represents direct healthcare costs (IOM 1999). In 2004, more than half (55
percent) of the American public said that they were dissatisfied with the quali
ty of healthcare in this country, compared to 44 percent in 2000 (Henry J. Kaise
r Foundation, Agency for Healthcare Research and Quality, and Harvard School of
Public Health 2004).
6
These problems were studied in the landmark work of the Institute of Medicine (I
OM 2001), Crossing the Quality ChasmA New Health System for the 21st Century. The
IOM panel concluded that the knowledge to improve patient care is available, bu
t a gapa chasmseparates that knowledge from everyday practice. The panel summarize
s the goals of a new health system in six aims. (Box 1.1)
BOX 1.1 Six Aims of a New Health System
Patient care should be 1. Safe, avoiding injuries to patients from the care that
is intended to help them; 2. Effective, providing services based on scientific
knowledge to all who could benefit, and refraining from providing services to th
ose not likely to benefit (avoiding underuse and overuse, respectively); 3. Pati
ent-centered, providing care that is respectful of and responsive to individual
patient preferences, needs, and values, and ensuring that patient values guide a
ll clinical decisions; 4. Timely, reducing wait times and harmful delays for bot
h those who receive and those who give care; 5. Efficient, avoiding waste of equ
ipment, supplies, ideas, and energy; and 6. Equitable, providing care that does
not vary in quality because of personal characteristics such as gender, ethnicit
y, geographic location, and socioeconomic status.
SOURCE: Reprinted with permission from Crossing the Quality ChasmA New Health Sys
tem for the 21st Century 2001 by the National Academy of Sciences, Courtesy of t
he National Academies Press, Washington, D.C.
The IOM panel recommended ten steps to close the gap between care with the above
characteristics and current practice (Box 1.2).
BOX 1.2 Ten Steps to Close the Gap
The ten steps to close the gap are: 1. Care based on continuous healing relation
ships. Patients should receive care whenever they need it and in many forms, not
just face-to-face visits. This rule implies that the healthcare system should b
e responsive at all times (24 hours a day, every day), and that access to care s
hould be provided over the Internet, by telephone, and by other means in additio
n to face-to-face visits. 2. Customization based on patient needs and values. Th
e system of care should be designed to meet the most common types of needs, but
have the capability to respond to individual patient choices and preferences.

7
3. The patient as the source of control. Patients should be given all relevant i
nformation and the opportunity to exercise whatever degree of control they choos
e over healthcare decisions that affect them. The health system should be able t
o accommodate differences in patient preferences and encourage shared decision m
aking. 4. Shared knowledge and the free flow of information. Patients should hav
e unfettered access to their own medical information and to clinical knowledge.
Clinicians and patients should communicate effectively and share information. 5.
Evidence-based decision making. Patients should receive care based on the best
available scientific knowledge. Care should not vary illogically from clinician
to clinician or from place to place. 6. Safety as a system property. Patients sh
ould be safe from injury caused by the care system. Reducing risk and ensuring s
afety require greater attention to systems that help prevent and mitigate errors
. 7. The need for transparency. The healthcare system should make available to p
atients and their families information that allows them to make informed decisio
ns when selecting a health plan, hospital, or clinical practice, or when choosin
g among alternative treatments. This should include information describing the s
ystems performance on safety, evidencebased practice, and patient satisfaction. 8
. Anticipation of needs. The health system should anticipate patient needs rathe
r than simply react to events. 9. Continuous decrease in waste. The health syste
m should not waste resources or patient time. 10. Cooperation among clinicians.
Clinicians and institutions should actively collaborate and communicate to ensur
e an appropriate exchange of information and coordination of care.
SOURCE: Reprinted with permission from Crossing the Quality ChasmA New Health Sys
tem for the 21st Century 2001 by the National Academy of Sciences, Courtesy of t
he National Academies Press, Washington, D.C.
BOX 1.2 Ten Steps to Close the Gap (continued)
Many healthcare leaders have begun to address these issues and are capitalizing
on proven tools employed by other industries to ensure high performance and qual
ity outcomes. For major change to occur in the U.S. health system, however, thes
e strategies must be adopted by a broad spectrum of healthcare providers and imp
lemented consistently throughout the continuum of careambulatory, inpatient/acute
settings, and long-term care. The payers for healthcare must engage with the de
livery system to find new ways to partner for improvement. In addition, patients
have to assume a stronger financial and self-care role in this new system. Alth
ough not all of the IOM goals can be accomplished through operational improvemen
ts, this book provides methods and tools to actively change the system to accomp
lish many aspects of them.
8
The Opportunity
Although the current American health system presents numerous challenges, opport
unities for improvement are emerging as well. Three major trends provide hope th
at significant change is possible.
Evidence-Based Medicine
The use of evidence-based medicine (EBM) for the delivery of healthcare is the r
esult of 30 years of work by some of the most progressive and thoughtful practit
ioners in the nation. The movement has produced an array of care guidelines, car
e patterns, and new shared decision-making tools for both caregivers and patient
s. The cost of healthcare could be reduced by nearly 29 percent and clinical out
comes improved significantly if EBM guidelines and the most efficient care proce
dures were used by all practitioners in the United States (Wennberg, Fisher, and
Skinner 2004). Comprehensive resources are available to the healthcare organiza
tion that wishes to emphasize EBM. For example, the National Guideline Clearingh
ouse (NGC 2006) is a comprehensive database of evidence-based clinical practice
guidelines and related documents and contains more than 4,000 guidelines. NGC is
an initiative of the Agency for Healthcare Research and Quality (AHRQ) of the U
.S. Department of Health and Human Services. NGC was originally created by AHRQ
in partnership with the American Medical Association and American Association of
Health Plans, now Americas Health Insurance Plans (AHIP).
Knowledge-Based Management
Knowledge-based management (KBM) employs data and information, rather than feeli
ngs or intuition, to support management decisions. Practitioners of KBM use the
tools contained in this book for cost reduction, increased safety, and improved
clinical outcomes. The evidence for the efficacy of these techniques is containe
d in the operations research and management science literature. Although these t
ools have been taught in healthcare graduate programs for many years, they have
not migrated widely into practice. Recently, the IOM (Proctor et al. 2005) has r
ecognized the opportunities that the use of KBM presents with its publication Bu
ilding a Better Delivery System: A New Engineering/Healthcare Partnership. In ad
dition, AHRQ and Denver Health provide practical operations improvement tools in
A Toolkit for Redesign in Healthcare (Gabow et al. 2003). Healthcare delivery h
as been slow to adopt information technologies, but many organizations are now b
eginning to aggressively implement electronic medical record systems and other a
utomated tools. Hillestad et al. (2005) have suggested that broad deployment of
these systems could save up to $371 billion annually in the United States.

9
A More Active Role for the Consumer
Consumers are beginning to assume new roles in their own care through the use of
health education and information and more effective partnering with their healt
hcare providers. Personal maintenance of wellness though a healthy lifestyle is
one essential component. Understanding ones disease and treatment options and hav
ing an awareness of the cost of care are also important responsibilities of the
consumer. Patients will become good consumers of healthcare by finding and using
price information in selecting providers and treatments. Many employers are now
offering high-deductible health plans with accompanying health savings accounts
(HSAs.) This type of consumer-directed healthcare is likely to grow and increas
e pressure on providers to deliver cost-effective, customersensitive, high-quali
ty care. The healthcare delivery system of the future will support and empower a
ctive, informed consumers.
A Systems Look at Healthcare
The Clinical System
To improve healthcare operations, it is important to understand the systems that
influence the delivery of care. Clinical care delivery is embedded in a series
of interconnected systems (Figure 1.1). The patient care microsystem is where th
e healthcare professional provides hands-on care. Elements of the clinical micro
system include: FIGURE 1.1 A Systems View of Healthcare
Environment Level D Organization Level C Microsystem Level B Patient Level A
SOURCE: Ransom, Maulik, and Nash (2005). Based on Ferlie, E., and S. M. Shortell
. 2001. Improving the Quality of Healthcare in the United Kingdom and the United
States: A Framework for Change. The Milbank Quarterly 79(2): 281316.
10
The team of health professionals who provide clinical care to the patient; The t
ools the team has to diagnose and treat the patient (e.g., imaging capabilities,
lab tests, drugs); and The logic for determining the appropriate treatments and
the processes to deliver this care. Because common conditions (e.g., hypertensi
on) affect a large number of patients, clinical research has determined the most
effective way to treat these patients. Therefore, in many cases, the organizati
on and functioning of the microsystem can be optimized. Process improvements can
be made at this level to ensure that the most effective, least costly care is d
elivered. In addition, the use of EBM guidelines can also help ensure that the p
atient receives the correct treatment at the correct time. The organizational in
frastructure also influences the effective delivery of care to the patient. Ensu
ring that providers have the correct tools and skills is an important element of
infrastructure. The use of KBM provides a mechanism to optimize the use of clin
ical tools. The electronic health record is one of the most important advances i
n the clinical microsystem for both process improvement and the wider use of EBM
. Another key component of infrastructure is the leadership displayed by senior
staff. Without leadership, effective progress or change will not occur. Finally,
the environment strongly influences the delivery of care. Key environmental fac
tors include competition, government regulation, demographics, and payer policie
s. An organizations strategy is frequently influenced by such factors (e.g., a ne
w regulation from Medicare, a new competitor). Many of the systems concepts rega
rding healthcare delivery were initially developed by Avedis Donabedian. These f
undamental contributions are discussed in depth in Chapter 2.
System Stability and Change
Elements in each layer of this system interact. Peter Senge (1990) provides a us
eful theory to understand the interaction of elements in a complex system such a
s healthcare. In his model, the structure of a system is the primary mechanism f
or producing an outcome. For example, an organized structure of facilities, trai
ned professionals, supplies, equipment, and EBM care guidelines has a high proba
bility of producing an expected clinical outcome. No system is ever completely s
table. Each systems performance is modified and controlled by feedback (Figure 1.
2). Senge (1990, 75) defines feedback as any reciprocal flow of influence. In sys
tems thinking it is an axiom that every influence is both cause and effect. As sh
own in Figure 1.2,

11
Employee motivation
+
+
Financial performance, profit
FIGURE 1.2 Systems with Reinforcing and Balancing Feedback
Salaries
+
Add or reduce staff
Actual staffing level
Compare actual to needed staff based on patient demand

higher salaries provide an incentive for higher performance levels by employees.
This, in turn, leads to better financial performance and profitability; increas
ed profits provide additional funds for higher salaries, and the cycle continues
. Another frequent example in healthcare delivery is patient lab results that di
rectly influence the medication ordered by a physician. A third example is a fin
ancial report that shows an overexpenditure in one category that will prompt a m
anager to reduce spending to meet budget goals. A more formal systems definition
with feedback includes a process, a sensor that monitors process output, a feed
back loop, and a control that modifies how the process operates. Feedback can be
either reinforcing or balancing. Reinforcing feedback prompts change that build
s on itself and amplifies the outcome of a process, taking the process further a
nd further from its starting point. The effect of reinforcing feedback can be ei
ther positive or negative. For example, a reinforcing change of positive financi
al results for an organization could lead to higher salaries, which would then l
ead to even better financial performance because the employees were highly motiv
ated. In contrast, a poor supervisor could lead to employee turnover, short staf
fing, and even more turnover.
12
Balancing feedback prompts change that seeks stability. A balancing feedback loo
p attempts to return the system to its starting point. The human body provides a
good example of a complex system that has many balancing feedback mechanisms. F
or example, an overheated body prompts perspiration until the body is cooled thr
ough evaporation. The clinical term for this type of balance is homeostasis. A c
linical treatment process that controls drug dosing via real-time monitoring of
the patients physiological responses is an example of balancing feedback. Inpatie
nt unit staffing levels that drive where in a hospital patients are admitted is
another. All of these feedback mechanisms are designed to maintain balance in th
e system. A confounding problem with feedback is delay. Delays occur when there
are interruptions between actions and consequences. When this happens, systems t
end to overshoot and perform poorly. For example, an emergency department might
experience a surge in patients and call in additional staff. If the surge subsid
es, the added staff may not be needed and unnecessary expense will have been inc
urred. As healthcare leaders focus on improving their operations, it is importan
t to understand the systems in which change resides. Every change will be resist
ed and reinforced by feedback mechanisms, many of which are not clearly visible.
Taking a broad systems view can improve the effectiveness of change. Many subsy
stems in the total healthcare system are interconnected. These connections have
feedback mechanisms that either reinforce or balance the subsystems performance.
Figure 1.3 shows a simple connection that originates in the environmental segmen
t of the total health system. Each process has both reinforcing and balancing fe
edback.
An Integrating Framework for Operations Management in Healthcare
This book is divided into five major sections: Introduction to healthcare operat
ions; Setting goals and executing strategy; FIGURE 1.3 Linkages Within the Healt
hcare System: Chemotherapy
Payers want to reduce costs for chemotherapy Environment
New payment method for chemotherapy is created Organization
Chemotherapy treatment needs to be more efficient to meet payment levels Clinica
l microsystem
Changes are made in care processes and support systems to maintain quality while
reducing costs Patient

13
Performance improvement tools, techniques, and programs; Applications to contemp
orary healthcare operations issues; and Putting it all together for operational
excellence. This schema reflects the authors view that effective operations manag
ement in healthcare consists of highly focused strategy execution and organizati
onal change accompanied by the disciplined use of analytical tools, techniques,
and programs. The book includes examples of applications of this approach to com
mon healthcare challenges. Figure 1.4 illustrates this framework. An organizatio
n needs to understand the environment, develop a strategy, and implement a syste
m to effectively deploy this strategy. At the same time, the organization must b
ecome adept at using all the tools of operations improvement contained in this b
ook. These improvement tools can then be combined to attack the fundamental chal
lenges of operating a complex healthcare delivery organization.
The introductory chapters provide an overview of the significant environmental t
rends healthcare delivery organizations face. Annual updates to industry-wide tr
ends can be found in Futurescan: Healthcare Trends and Implications 20082013 (Soc
iety for Healthcare Strategy and Market Development and American College of Heal
thcare Executives 2008). Progressive organizations will review these publication
s carefully. Then, using this information, they can respond to external forces b
y identifying either new strategies or current operating problems that must be a
ddressed. Business has been aggressively using operations improvement tools for
the past 30 years, but the field of operations science actually began many centu
ries in the past. Chapter 2 provides a brief history. Healthcare operations are
being strongly driven by the effects of EBM and pay-for-performance. Chapter 3 p
rovides an overview of these trends and how organizations can effect change to m
eet current challenges and opportunities. FIGURE 1.4 Framework for Effective Ope
rations Management in Healthcare
Setting goals and executing strategy Performance improvement tools, techniques,
and programs
Fundamental healthcare operations issues
High performance
14
Setting Goals and Executing Strategy
A key component of effective operations is the ability to move strategy to actio
n. Chapter 4 shows how the use of the balanced scorecard can accomplish this aim
. Change in all organizations is challenging, and formal methods of project mana
gement (Chapter 5) can be used to make effective, lasting improvements in an org
anizations operations.
Performance Improvement Tools, Techniques, and Programs
Once an organization has in place strategy implementation and change management
processes, it needs to select the correct tools, techniques, and programs to ana
lyze current operations and implement effective changes. Chapter 6Tools for Probl
em Solving and Decision Makingoutlines the basic steps of problem solving, beginn
ing with framing the question or problem and continuing through data collection
and analyses to enable effective decision making. Chapter 7Using Data and Statist
ical Tools for Operations Improvementprovides a review of the building blocks for
many of the more advanced tools used later in the book. (This chapter may serve
as a review or reference for readers who already have good statistical skills.)
Some projects will require a focus on process improvement. Six Sigma tools (Cha
pter 8) can be used to reduce the variability in the outcome of a process. Lean
tools (Chapter 9) can be used to eliminate waste and increase speed. Many health
care processes, such as patient flow, can be modeled and improved by using compu
ter simulation (Chapter 10), which may also be used to evaluate project risks.
Applications to Contemporary Healthcare Operations Issues
This part of the book demonstrates how these concepts can be applied to some of
todays fundamental healthcare challenges. Process improvement techniques are wide
ly deployed in many organizations to significantly improve performance; Chapter
11 reviews the tools of process improvement and demonstrates their use in improv
ing patient flow. Scheduling and capacity management continue to be major concer
ns for many healthcare delivery organizations, particularly with the advent of a
dvanced access. Chapter 12 demonstrates how simulation can be used to optimize s
cheduling. Chapter 13Supply Chain Managementexplores the optimal methods of acquir
ing supplies and maintaining appropriate inventory levels. In the end, any opera
tions improvement will fail unless steps are taken to maintain the gains; Chapte
r 14Putting it All Together for Operational Excellencecontains the necessary tools
. The chapter also provides a more detailed algorithm that can help practitioner
s select the appropriate tools,

15
methods, and techniques to make significant operational improvements. It include
s an example of how Vincent Valley Hospital and Health System (VVH) uses all the
tools in the book to achieve operational excellence.
Vincent Valley Hospital and Health System
Woven throughout the sections described below are examples designed to consisten
tly illustrate the tools discussed. A fictitious but realistic health system, VV
H, is featured in these examples. (The companion website, ache.org/books/OpsMana
gement, contains a more expansive description of VVH.) VVH is located in a Midwe
stern city of 1.5 million. It has 3,000 employees, operates 350 inpatient beds,
and has a medical staff of 450 physicians. In addition, VVH operates nine clinic
s staffed by physicians who are employees of the system. VVH has two major compe
titor hospitals, and a number of surgeons from all three hospitals recently join
ed together to set up an independent ambulatory surgery center. Three major heal
th plans provide most of the private payment to VVH and, along with the state Me
dicaid system, have recently begun a pay-forperformance initiative. VVH has a st
rong balance sheet and a profit margin of approximately 2 percent, but feels fin
ancially challenged. The board of VVH includes many local industry leaders, who
have asked the chief executive officer to focus on using the operational techniq
ues that have led them to succeed in their businesses.
Conclusion
This book is an overview of operations management approaches and tools. It is ex
pected that the successful reader will understand all the concepts in the book (
and in current use in the field) and should be able to apply at the basic level
some of the tools, techniques, and programs presented. It is not expected that t
he reader will be able to execute at the more advanced level (e.g., Six Sigma bl
ack belt, Project Management Professional). However, this book will prepare read
ers to work effectively with knowledgeable professionals and, most important, en
able them to direct their work.
Discussion Questions
1. Review the ten action steps recommended by IOM to close the quality chasm. Ra
nk them from easiest to most difficult to achieve, and give a rationale for your
rankings.
16
2. Give three examples of possibilities for system improvement at the boundaries
of the healthcare subsystems (patient, microsystem, organization, and environme
nt). 3. Identify three systems in a healthcare organization (at any level) that
have reinforcing feedback. 4. Identify three systems in a healthcare organizatio
n (at any level) that have balancing feedback. 5. Identify three systems in a he
althcare organization (at any level) where feedback delays affect the performanc
e of the system.
References
DoBias, M., and M. Evans. 2006. Mixed SignalsThe CMS 10-Year Spending Projections
Inspire Both Hope and Skepticism, and Leave Plenty of Room for Lobbyists. Modern
Healthcare 36 (9): 68. Gabow, P., S. Eisert, A. Karkhanis, A. Knight, and P. Dick
son. 2003. A Toolkit for Redesign in Healthcare. Washington, D.C.: Agency for He
althcare Research and Quality. Hammer, M. 2005. Making Operational Innovation Wor
k. Harvard Management Update 10 (4): 34. Henry J. Kaiser Foundation, Agency for He
althcare Research and Quality, and Harvard School of Public Health. 2004. Nation
al Survey on Consumers Experiences with Patient Safety and Quality Information. M
enlo Park, CA: Kaiser Family Foundation. [Online information; retrieved 8/28/06.
] www.kff.org/kaiserpolls/ upload/National-Survey-on-Consumers-Experiences-WithPatient-Safety-andQuality-Information-Survey-Summary-and-Chartpack.pdf. Hillesta
d, R., J. Bigelow, A. Bower, F. Girosi, R. Meili, R. Scoville, and R. Taylor. 20
05. Can Electronic Medical Record Systems Transform Health Care? Potential Health
Benefits, Savings, and Costs. Health Affairs 24 (5): 110317. Institute of Medicin
e. 2001. Crossing the Quality ChasmA New Health System for the 21st Century. Wash
ington, D.C.: National Academies Press. . 1999. To Err Is Human: Building a Safer H
ealth System. Washington, D.C.: National Academies Press. National Guideline Cle
aringhouse (NGC). 2006. [Online information; retrieved 8/28/06.] www.guideline.g
ov/. Proctor, P., W. Reid, D. Compton, J. H. Grossman, and G. Fanjiang. 2005. Bu
ilding a Better Delivery System: A New Engineering/Health Care Partnership. Wash
ington, D.C.: Institute of Medicine. Ransom, S. B., J. S. Maulik, and D. B. Nash
, (eds.), 2005. The Healthcare Quality Book: Vision, Strategy, and Tools. Chicag
o: Health Administration Press. Senge, P. M. 1990. The Fifth DisciplineThe Art an
d Practice of the Learning Organization. New York: Doubleday.

17
Society for Healthcare Strategy and Market Development and American College of H
ealthcare Executives. 2008. Futurescan: Healthcare Trends and Implications 200820
13. Chicago: Health Administration Press. Wennberg, J. E., E. S. Fisher, and J.
S. Skinner. 2004. Geography and the Debate over Medicare Reform. Health Affairs 23
(Sept. 2004 Variations Supplement): W96W114.
CHAPTER
2
HISTORY OF PERFORMANCE IMPROVEMENT
CHAPTER OUTLINE
Operations Management in Action Overview Background Knowledge-Based Management H
istory of Scientific Management Mass Production Frederick Taylor Frank and Lilli
an Gilbreth Scientific Management Today Project Management Quality Walter Shewha
rt W. Edwards Deming Joseph M. Juran Avedis Donabedian TQM and CQI, Leading to S
ix Sigma ISO 9000 Baldrige Award JIT, Leading to Lean and Agile Baldrige, Six Si
gma, Lean, and ISO 9000 Service Typologies Supply Chain Management Conclusion Di
scussion Questions References
18

agile Agency for Healthcare Research and Quality (AHRQ) Centers for Medicare & M
edicaid Services (CMS) continuous quality improvement (CQI) critical path method
(CPM) Demings 14 points for healthcare enterprise resource planning (ERP) Instit
ute for Healthcare Improvement (IHI) ISO 9000 Jurans quality trilogy just-in-time
(JIT) knowledge-based management (KBM) knowledge hierarchy Lean Malcolm Baldrig
e National Quality Award materials requirements planning (MRP) plan-do-check-act
(PDCA) plan-do-study-act, a variation of plan-do-check-act program evaluation a
nd review technique (PERT) service process matrix service typologies single-minu
te exchange of die (SMED) Six Sigma statistical process control (SPC) supply cha
in management (SCM) systems thinking total quality management (TQM) Toyota Produ
ction System (TPS)
19

Operations Management in Action
During the Crimean War, reports of terrible conditions in military hospitals ala
rmed British citizens. In response to the outcry, Florence Nightingale was commi
ssioned to oversee the introduction of nurses to military hospitals and to impro
ve conditions in the hospitals. When Nightingale arrived in Scutari, Turkey, she
found the hospital overcrowded and filthy. Nightingale instituted many changes
to improve the sanitary conditions in the hospital, and many lives were saved as
a result of these reforms. Nightingale was one of the first people in healthcar
e to collect, tabulate, interpret, and graphically display data related to the e
ffect of process changes on care outcomes. Today, this is called evidence-based m
edicine. To quantify the overcrowding problem, she compared the amount of space p
er patient in London hospitals, 1,600 square feet, to the space in Scutari, abou
t 400 square feet. She developed a standard Model Hospital Statistical Form to e
nable the collection of consistent data for analysis and comparison. In February
1855, the mortality rate of patients in Scutari was 42 percent. As a result of
Nightingales changes, the mortality rate decreased to 2.2 percent by June 1855. T
o present these data in a persuasive manner, she developed a new type of graphic
display, the polar area diagram. After the war, Nightingale used the data she h
ad collected to demonstrate that the mortality rate in Scutari, after her reform
s, was significantly lower than in other British military hospitals. Although th
e British military hierarchy was resistant to her changes, the data were convinc
ing and resulted in reforms to military hospitals and the establishment of the R
oyal Commission on the Health of the Army. Florence Nightingale would recognize
many of the philosophies, tools, and techniques outlined in this text as being e
ssentially the same as those she effectively employed to achieve lasting reform
in hospitals throughout the world.
SOURCES: Cohen 1984; Neuhauser 2003; and Nightingale 1858, 1999.
Overview
This chapter provides the background and historical context of performance impro
vement, which is not a new conceptmany of the tools, techniques, and philosophies
outlined in this text are based in the past. Although the terminology has chang
ed, many of the core concepts remain the same. The major topics in this chapter
include: Systems thinking and knowledge-based management; Scientific management
and project management; 20
C h a p t e r 2 : H i s t o r y o f Pe r f o r m a n c e I m p rove m e n t
21
Quality experts; Service typologies; and Philosophies of performance improvement
, including Six Sigma, Lean, and supply chain management.
Background
Operations management is the design, operation, and improvement of the processes
and systems that create and deliver the organizations products and services. Ope
rations managers plan and control delivery processes and systems within the orga
nization. The goal of operations management is to more effectively and efficient
ly produce and deliver the organizations products and services. Healthcare profes
sionals have realized that the theories, tools, and techniques of operations man
agement, if properly applied, can enable their own organizations to become more
efficient and effective. The operations management information presented in this
book should enable healthcare professionals to design systems, processes, produ
cts, and services that meet the needs of their organizations stakeholders. It sho
uld also enable the continuous improvement of these systems and services to meet
the needs of a quickly changing environment.The healthcare industry is facing m
any challenges. The costs of care and level of services delivered are increasing
; more and more we are able to prolong lives as technology advances and the popu
lation ages. In 2004, 15 percent of the U.S. economy was devoted to medical care
, up from 11 percent in 1987 (CMS 2005). The need for quality care with zero def
ects, or failures in care, is being driven by government and other stakeholders.
The need to produce more of a higher quality product or service at a reduced co
st can only be met through better utilization of resources. The healthcare envir
onment has recognized the need to control costs while increasing both the level
and quality of service. These seemingly contradictory goals can only be reached
if healthcare providers can offer their services more effectively and efficientl
y, better utilizing limited resources that include financial assets, employees a
nd staff, machines and facilities, and time. Healthcare providers have the need
and opportunity to adopt many of the tools and techniques that have enabled othe
r service industries and manufacturing to become both more efficient and effecti
ve. Six Sigma and Lean are two of the philosophies that have been successfully i
mplemented in the manufacturing sector to decrease costs, increase product quali
ty, and improve timeliness of delivery. As Donald Berwick, M.D., president and c
hief executive officer (CEO) of the Institute for Healthcare Improvement (IHI),
says, We have to bring the science of management back into healthcare in a way th
at we havent in a very long time (Allen 2004).
22
To improve systems and processes, one must first know the system or process and
its desired inputs and outputs. This book takes a systems view of service provis
ion and delivery, as illustrated in Figure 2.1.
Knowledge-Based Management
To design effective and efficient systems and processes or improve existing proc
esses, knowledge of the systems and processes is needed. This book focuses on kn
owledge-based management (KBM)using data and information to base management decis
ions on facts rather than feelings or intuition. The knowledge hierarchy, as it is
sometimes called in the literature, is generally attributed to Harlan Cleveland
(1982) or Russell Ackoff (1989), an operations researcher and leading systems t
heorist. The knowledge hierarchy relates to the learning that ultimately underpi
ns KBM and consists of five categories, summarized below (Zeleny 1987) and illus
trated in Figure 2.2. Data: symbols or raw numbers that simply exist; they have
no structure or organization. Organizations collect data with their computer sys
tems; individuals collect data through their experiences. In short, know nothing.
Information: data that are organized or processed to have meaning. Information c
an be useful, but it is not necessarily useful. It can answer such questions as
who, what, where, and when. Know what. Knowledge: information that is deliberate
ly useful. Knowledge enables decision making. Know how. Understanding: allows us
e of what is known and enables the development of new knowledge. Understanding r
epresents the difference between learning and memorizing. Know why. Wisdom: adds
moral and ethical views to understanding. Wisdom answers questions to which the
re is no known correct answer and, in some cases, where there will never be a kn
own correct answer. Know right. FIGURE 2.1 Systems View
INPUT
Transformation process Feedback
OUTPUT
Labor Material Machines Management Capital
Goods or services
23
Wisdom morals Understanding Importance principles Knowledge patterns Information
Data relationships Learning
FIGURE 2.2 Knowledge Hierarchy
A simple example may help to explain this hierarchy. Your height is 67 inches, a
nd your weight is 175 pounds (data). You have a body mass index (BMI) of 26.7 (i
nformation). A healthy BMI is 18.5 to 25.5 (knowledge). Your BMI is high, and to
be healthy you should lower it (understanding). You begin a diet and exercise p
rogram and lower your BMI (wisdom). Finnie (1997, 24) summarizes the relationshi
ps within the hierarchy, and our focus on its less important levels:
There is another aspect to learning that relates to the five types of the conten
t in the mind. We talk about the accumulation of information, but we fail to dis
tinguish between data, information, knowledge, understanding, and wisdom. An oun
ce of information is worth a pound of data, an ounce of knowledge is worth a pou
nd of information, an ounce of understanding is worth a pound of knowledge, an o
unce of wisdom is worth a pound of understanding. In the past, our focus has bee
n inversely related to importance. We have focused mainly on data and informatio
n, a little bit on knowledge, nothing on understanding, and virtually less than
nothing on wisdom.
The roots of the knowledge hierarchy can be traced even further back to eighteen
th century philosopher Immanuel Kant, much of whose work attempted to address th
e questions of what and how we can know. The two major philosophical movements t
hat significantly influenced Kant were empiricism and rationalism (McCormick 200
6). The empiricists, most notably John Locke, argued that human knowledge origin
ates in ones experiences. According to Locke, the mind is a blank slate that fill
s with ideas through its interaction with the world; experience is where all kno
wledge
24
originates. The rationalists, including Descartes and Galileo, on the other hand
, argued that the world is knowable through an analysis of ideas and logical rea
soning. Both the empiricists and rationalists viewed the mind as passive, either
because it received ideas onto a blank slate or because it possessed innate ide
as that could be logically analyzed. Kant joined these philosophical ideologies
and argued that experience leads to knowing only if the mind provides a structur
e for those experiences. Although the idea that the rational mind plays a role i
n defining reality is now common, in Kants time this was a major insight into wha
t and how we know. Knowledge does not flow from our experiences alone, nor from
only our ability to reason; rather, knowledge flows from our ability to apply re
asoning to our experiences. Relating Kant to the knowledge hierarchy, data are o
ur experiences, information is obtained through logical reasoning, and knowledge
is obtained when we take data and apply structured reasoning to that data to ac
quire knowledge (Ressler and Ahrens 2006). The intent of this text is to enable
readers to gain knowledge. We discuss tools and techniques that enable the appli
cation of logical reasoning to data in order to obtain knowledge and use it to m
ake better decisions. This knowledge and understanding should enable the reader
to provide healthcare in a more efficient and effective manner.
History of Scientific Management
Frederick Taylor originated the term scientific management in The Principles of Sc
ientific Management (Taylor 1911). Scientific management methods called for elim
inating the old rule-of-thumb, individual way of performing work and, through st
udy and optimization of the work, replacing the varied methods with the one best w
ay of performing the work to improve productivity and efficiency. Today, the ter
m scientific management has been replaced with operations management, but the in
tent is similar: study the process or system and determine ways to optimize it i
n order to make it more efficient and effective.
Mass Production
The Industrial Revolution and mass production set the stage for much of Taylors w
ork. Prior to the Industrial Revolution, individual craftsmen performed all task
s necessary to produce a good using their own tools and procedures. In the eight
eenth century, Adam Smith advocated the division of labormaking work more efficie
nt through specialization. To support a division of labor, a large number of wor
kers are brought together, and each performs a specific task related to the prod
uction of a good. Thus, the factory system of mass production was born, and Henr
y Fords assem-
25
bly line eventually came into being, setting the stage for Taylors scientific man
agement. Mass production allows for significant economies of scale, as predicted
by Smith. Before Ford set up his moving assembly line, each car chassis was ass
embled by a single worker and took about 1212 hours to produce. After the intro d
uction of the assembly line, this time was reduced to 93 minutes (Bellis 2006).
The standardization of products and work allowed for a reduction in the time nee
ded to produce cars and significantly reduced the costs of production. The selli
ng price of the Model T fell from $1,000 to $360 between 1908 and 1916 (Simkin 2
005), allowing Ford to capture a large portion of the market. Although Ford is c
ommonly credited with introducing the moving assembly line and mass production i
n modern times, they existed several hundred years earlier. The Venetian Arsenal
of the 1500s produced nearly one ship every day and employed 16,000 people (Nat
ionMaster.com 2004). Ships were mass produced using premanufactured, standardize
d parts on a floating assembly line (Schmenner 2001). One of the first examples
of mass production in the healthcare industry is Shouldice Hospital (Heskett 200
3). Much like Ford, who said people could have the Model T in any color, so long
as its black, Shouldice performs just one type of surgery. The hospital performs o
nly routine hernia operations, not more complicated hernia surgery or any other
types of surgery. There exists in healthcare growing evidence that experience in
treating specific illnesses and conditions affects the outcome of that care. Hi
gher volumes of cases often result in better outcomes (Halm, Lee, and Chassin 20
02). Although higher volume alone does not produce better outcomes, the addition
al practice associated with higher volume results in better outcomes. The idea o
f practice makes perfect, or learning curve effects, has led organizations such as
the Leapfrog Group (made up of organizations that provide healthcare benefits)
to make patient volume one of its criteria for quality. The Agency for Healthcar
e Research and Quality (AHRQ) report Making Health Care Safer: A Critical Analysi
s of Patient Safety Practices (Auerbach 2001) devotes an entire chapter to this i
ssue and its effect on practice.
Frederick Taylor
Taylor began his work when mass production and the factory system were in their
infancy. He believed that U.S. industry was wasting human effort and that, as a re
sult, national efficiency (now called productivity) was significantly lower than
it could be. The introduction to The Principles of Scientific Management (Taylo
r 1911) illustrates his intent:
But our larger wastes of human effort, which go on every day through such of our
acts as are blundering, ill-directed, or inefficient, and which Mr. Roosevelt
26
refers to as a lack of national efficiency, are less visible, less tangible, and a
re but vaguely appreciated. . . . This paper has been written: First. To point o
ut, through a series of simple illustrations, the great loss which the whole cou
ntry is suffering through inefficiency in almost all of our daily acts. Second.
To try to convince the reader that the remedy for this inefficiency lies in syst
ematic management, rather than in searching for some unusual or extraordinary ma
n. Third. To prove that the best management is a true science, resting upon clea
rly defined laws, rules, and principles, as a foundation. And further to show th
at the fundamental principles of scientific management are applicable to all kin
ds of human activities, from our simplest individual acts to the work of our gre
at corporations, which call for the most elaborate cooperation. And, briefly, th
rough a series of illustrations, to convince the reader that whenever these prin
ciples are correctly applied, results must follow which are truly astounding.
Note that Taylor specifically mentions systems management as opposed to the indi
vidual; this is a common theme that we revisit throughout this book. Rather than
focusing on individuals as the cause of problems and the source of solutions, t
he focus is on systems and their optimization. Taylor believed that much waste w
as the result of what he called soldiering, which today might be called slacking. He
believed that the underlying causes of soldiering were as follows (Taylor 1911)
:
First. The fallacy, which has from time immemorial been almost universal among w
orkmen, that a material increase in the output of each man or each machine in th
e trade would result in the end in throwing a large number of men out of work. S
econd. The defective systems of management which are in common use, and which ma
ke it necessary for each workman to soldier, or work slowly, in order that he ma
y protect his own best interests. Third. The inefficient rule-of-thumb methods,
which are still almost universal in all trades, and in practicing which our work
men waste a large part of their effort.
To eliminate soldiering, Taylor proposed instituting incentive schemes. While at
Midvale Steel Company, he used time studies to set daily production quotas. Inc
entives were paid to those workers reaching their daily goals, and those not rea
ching their goals were paid significantly less. Productivity at Midvale doubled.
Not surprisingly, Taylors ideas produced considerable backlash. The backlash aga
inst increasingly popular pay-for-performance programs in healthcare today is an
alogous to that experienced by Taylor.
27
Taylor believed there was one best way to perform a task and that careful study an
d analysis would lead to the discovery of that way. While at Bethlehem Steel Cor
poration, he studied the shoveling of coal. Using time studies and a careful ana
lysis of how the work was done, he determined that the optimal amount per load w
as 21 pounds. Taylor then developed shovels that would hold exactly 21 pounds fo
r each type of coal (workers had previously supplied their own shovels; NetMBA.c
om 2005). He also determined the ideal work rate and rest periods to ensure that w
orkers could shovel all day without fatigue. As a result of Taylors improved meth
ods, Bethlehem Steel was able to reduce the number of workers shoveling coal fro
m 500 to 140 (Nelson 1980). Taylors four principles of scientific management are
to: 1. Develop and standardize work methods based on scientific study and use th
ese to replace individual rule-of-thumb methods; 2. Select, train, and develop w
orkers rather than allowing them to choose their own tasks and train themselves;
3. Develop a spirit of cooperation between management and workers to ensure tha
t the scientifically developed work methods are both sustainable and implemented
on a continuing basis; and 4. Divide work between management and workers so tha
t each does an equal share, where management plans the work and workers actually
do the work. Although some of Taylors ideas would be problematic todayparticularl
y the notion that workers are machinelike and motivated solely by money many of his
ideas can be seen in the foundations of new initiatives such as Six Sigma and Lea
n.
Frank and Lillian Gilbreth
The Gilbreths were contemporaries of Frederick Taylor. Frank, who worked in the
construction industry, noticed that no two bricklayers performed their tasks the
same way. He believed that bricklaying could be standardized and the one best w
ay determined. He studied the work of bricklaying and analyzed the workers motion
s, finding much unnecessary stooping, walking, and reaching. He eliminated these
motions by developing an adjustable scaffold designed to hold both bricks and m
ortar (Taylor 1911). As a result of this and other improvements, Frank Gilbreth
reduced the number of motions in bricklaying from 18 to 5 (International Work Si
mplification Institute 1968) and raised output from 1,000 to 2,700 bricks a day
(Perkins 1997). He applied what he had learned from his bricklaying experiments
to other industries and work. In his study of surgical operations, Frank Gilbret
h found that doctors spent more time searching for instruments than performing t
he surgery. In response, he developed a technique still seen in operating rooms
today: When
28
the doctor needs an instrument, he extends his hand, palm up, and asks for the i
nstrument, which is then placed in his hand. Not only does this eliminate search
ing for the instrument, but it allows the doctor to stay focused on the surgical
area and therefore reduces surgical time (Perkins 1997). Frank and Lillian Gilb
reth may be more familiarly known as the parents in the book (Gilbreth and Carey
1948), movie (Lang 1950), and remake of the movie (Levy 2003) Cheaper by the Do
zen. The Gilbreths incorporated many of their time-saving ideas in the family ar
ena. For example, they only bought one type of sock for all 12 of their children
, thus eliminating timeconsuming sorting.
Scientific Management Today
Scientific management fell out of favor during the Depression, partly because of
the belief that it dehumanized employees, but mainly because it was believed th
at productivity improvements resulted in downsizing and greater unemployment. No
t until World War II was there a resurgence of interest in scientific management
, or operations research, as it came to be called. Despite this period of disfavor
, modern operations management has its roots in the theories of scientific manag
ement. In healthcare today, standardized methods and procedures are used to redu
ce costs and increase the quality of outcomes. Specialized equipment has been de
veloped to speed procedures and reduce labor costs. In some sense, we are still
searching for the one best way. However, care must be taken to heed the lessons of
the past. If the new tools of operations management are perceived to be dehumaniz
ing or to result in downsizing by healthcare organizations, their implementation
will meet significant resistance.
Project Management
The discipline of project management began with the development of the Gantt cha
rt in the early twentieth century. Henry Gantt worked closely with Frederick Tay
lor at Midvale Steel and in Navy ship construction during World War I. From this
work, he developed Gantt chartsbar graphs that illustrate the duration of projec
t tasks and visually display scheduled and actual progress. Gantt charts were us
ed to help manage such large projects as the construction of the Hoover Dam and
proved to be such a powerful tool that they are still commonly used today. Altho
ugh Gantt charts were used in large projects, they are not ideal for large, comp
licated projects because they do not explicitly show precedence relationships, t
hat is, what tasks need to be completed before other tasks can start. In the 195
0s, two mathematic project scheduling techniques were developed: the program eva
luation and review technique (PERT) and the critical path method (CPM). Both tec
hniques begin by developing a project network showing the precedence relationshi
ps among tasks and task duration.
29
PERT was developed by the U.S. Navy in response to the desire to accelerate the
Polaris missile program. This need for speed was precipitated by the Soviet launch
of Sputnik, the first space satellite. PERT uses a probability distribution (th
e Beta distribution), rather than a point estimate, for the duration of each pro
ject task. The probability of completing the entire project in a given amount of
time can then be determined. This technique is most useful for estimating proje
ct completion time when task times are uncertain and for evaluating risks to pro
ject completion prior to the start of a project. The CPM technique was developed
at the same time as PERT by the DuPont and Remington Rand corporations to manag
e plant maintenance projects. CPM uses the project network and point estimates o
f task duration times to determine the critical path through the network, or the
sequence of activities that will take the longest to complete. If any one of th
e activities on the critical path is delayed, the entire project will be delayed
. This technique is most useful when task times can be estimated with certainty
and is typically used in project management and control. Although both of these
techniques are powerful analytical tools for planning, implementing, controlling
, and evaluating a project plan, performing the required calculations by hand is
quite tedious, and their use was not widespread. With the advent of commerciall
y available project management software for personal computers in the late 1960s
, use of these techniques increased considerably. Today, numerous project manage
ment software packages are commercially available, and these techniques are used
extensively in industry. Microsoft Project, for instance, can perform network a
nalysis based on either PERT or CPM; however, the default is CPM, making this th
e more commonly used technique. Projects are an integral part of many of the pro
cess improvement initiatives found in the healthcare industry. Project managemen
t and its tools are needed to ensure that projects related to quality (Six Sigma
), Lean, and supply chain management are completed in the most effective and tim
ely manner possible.
Quality
Walter Shewhart
Although W. Edwards Deming and Joseph Juran are sometimes referred to as the fat
hers of the quality movement, Walter Shewhart is its grandfather. Both Deming an
d Juran studied under Shewhart, and much of their work was influenced by his ide
as. Shewhart believed that managers needed certain information to enable them to
make scientific, efficient, and economical decisions. He developed statistical
process control (SPC) charts to supply that information (Shewhart 1931). He also
believed that management and production practices need to
30
be continuously evaluated and adopted or rejected based on this evaluation if an
organization hopes to evolve and survive. The Deming cycle of improvement, or D
eming wheel (plan-do-check-act [PDCA] or plan-do-study-act), was adapted from Sh
ewharts work (Shewhart and Deming 1939).
W. Edwards Deming
Deming was an employee of the U.S. Government Services in the 1930s and 1940s, w
orking with statistical sampling techniques. He became a supporter and student o
f Shewhart, believing that his techniques could be useful in nonmanufacturing en
vironments. Deming applied SPC methods to his work at the National Bureau of the
Census to improve clerical operations in preparation for the 1940 population ce
nsus. In some cases, productivity improved by a factor of six. Deming taught sem
inars to bring his and Shewharts work to U.S. and Canadian organizations, where m
ajor reductions in scrap and rework resulted. However, after the war Demings idea
s lost popularity in the United States, mainly because demand for all products w
as so great that quality became unimportant; any product was snapped up by hungr
y consumers. After the war Deming went to Japan as an adviser for that countrys c
ensus. While he was there, the Union of Japanese Scientists and Engineers invite
d him to lecture on quality control techniques, and Deming brought his message t
o Japanese executives: Improving quality will reduce expenses while increasing p
roductivity and market share. During the 1950s and 1960s, Demings ideas were wide
ly known and implemented in Japan, but not in the United States. The energy cris
is of the 1970s and resulting increase in popularity of Japanese automobiles and
decline of the U.S. auto industry set the stage for the return of Demings ideas.
The lower prices and higher quality of Japanese automobiles and electronic good
s threatened U.S. industries and the economy. The 1980 television documentary If
Japan Can, Why Cant We (Mason 1980), investigating the increasing competition U.
S. industry was facing from Japan, made Deming and his quality ideas known to an
even broader audience. Much like the Institute of Medicine report To Err Is Hum
an (1999) increased awareness of the need for quality in healthcare, this docume
ntary increased awareness of the need for quality in manufacturing. Demings quali
ty ideas reflected his statistical background, but experience in their implement
ation caused him to broaden his approach. He believed that managers must underst
and the two types of variation. The first type, variation from special causes, i
s a result of a change in the system and can be identified or assigned and the p
roblem fixed. The second type, variation from common causes, is a result of the
natural differences in the system and cannot be eliminated without changing the
system. Although it might be possible to identify the common causes of variation
, they cannot be fixed without the authority and ability to improve the system for
which management is typically responsible.
31
Demings quality ideas went far beyond SPC to include a systematic approach to pro
blem solving and continuous process improvement with his PDCA cycle. He also bel
ieved that management is ultimately responsible for quality and must actively su
pport and encourage quality transformations within organizations. In the preface t
o Out of the Crisis, Deming (1986) writes:
Drastic changes are required. The first step in the transformation is to learn h
ow to change. . . . Long term commitment to new learning and new philosophy is r
equired of any management that seeks transformation. The timid and the faint-hea
rted, and people that expect quick results are doomed to disappointment. Whilst
the introduction of statistical problem solving and quality techniques and compu
terization and robotization have a part to play, this is not the solution: Solvi
ng problems, big problems and little problems, will not halt the decline of Amer
ican industry, nor will expansion in use of computers, gadgets, and robotic mach
inery. Benefits from massive expansion of new machinery also constitute a vain h
ope. Massive immediate expansion in the teaching of statistical methods to produ
ction workers is not the answer either, nor wholesale flashes of quality control
circles. All these activities make their contribution, but they only prolong th
e life of the patient, they cannot halt the decline. Only transformation of mana
gement and of Governments relations with industry can halt the decline.
Out of the Crisis contains Demings famous 14 points for management (Deming 1986).
Although not as well known, he also published an adaptation of the 14 points fo
r medical service (Box 2.1), which he attributed to Drs. Paul B. Batalden and Lo
ren Vorlicky of the Health Services Research Center, Minneapolis. BOX 2.1 Demings
Adaptation of the 14 Points for Medical Service
1. Establish constancy of purpose toward service. a. Define in operational terms
what you mean by service to patients. b. Specify standards of service for a year
hence and for five years hence. c. Define the patients whom you are seeking to s
erve. d. Constancy of purpose brings innovation. e. Innovate for better service.
f. Put resources into maintenance and new aids to production. g. Decide whom th
e administrators are responsible to and the means by which they will be held res
ponsible. h. Translate this constancy of purpose to service to patients and the
community. i. The board of directors must hold onto the purpose. 2. Adopt the ne
w philosophy. We are in a new economic age. We can no longer live with commonly
accepted levels of mistakes, materials not
32
BOX 2.1 Demings Adaptation of the 14 Points for Medical Service (continued)
suited to the job, people on the job who do not know what the job is and are afr
aid to ask, failure of management to understand their job, antiquated methods of
training on the job, and inadequate and ineffective supervision. The board must
put resources into this new philosophy, with commitment to in-service training.
3. a. Require statistical evidence of quality of incoming materials, such as ph
armaceuticals. Inspection is not the answer. Inspection is too late and is unrel
iable. Inspection does not produce quality. The quality is already built in and
paid for. Require corrective action, where needed, for all tasks that are perfor
med in the hospital. b. Institute a rigid program of feedback from patients in r
egard to their satisfaction with services. c. Look for evidence of rework or def
ects and the cost that may accrue. 4. Deal with vendors that can furnish statist
ical evidence of control. We must take a clear stand that price of services has
no meaning without adequate measure of quality. Without such a stand for rigorou
s measures of quality, business drifts to the lowest bidder, low quality and hig
h cost being the inevitable result. Requirement of suitable measures of quality
will, in all likelihood, require us to reduce the number of vendors. We must wor
k with vendors so that we understand the procedures that they use to achieve red
uced numbers of defects. 5. Improve constantly and forever the system of product
ion and service. 6. Restructure training. a. Develop the concept of tutors. b. D
evelop increased in-service education. c. Teach employees methods of statistical
control on the job. d. Provide operational definitions of all jobs. e. Provide
training until the learners work reaches the state of statistical control. 7. Imp
rove supervision. Supervision is the responsibility of the management. a. Superv
isors need time to help people on the job. b. Supervisors need to find ways to t
ranslate the constancy of purpose to the individual employee. c. Supervisors mus
t be trained in simple statistical methods with the aim to detect and eliminate
special causes of mistakes and rework. d. Focus supervisory time on people who a
re out of statistical control and not those who are low performers. If the membe
rs of a group are in fact in statistical control, there will be some low perform
ers and some high performers.
33
e. Teach supervisors how to use the results of surveys of patients. 8. Drive out
fear. We must break down the class distinctions between types of workers within
the organizationphysicians, nonphysicians, clinical providers versus nonclinical
providers, physician to physician. Discontinue gossip. Cease to blame employees
for problems of the system. Management should be held responsible for faults of
the system. People need to feel secure to make suggestions. Management must fol
low through on suggestions. People on the job cannot work effectively if they da
re not offer suggestions for simplification and improvement of the system. 9. Br
eak down barriers between departments. One way would be to encourage switches of
personnel in related departments. 10. Eliminate numerical goals, slogans, and p
osters imploring people to do better. Instead, display accomplishments of the ma
nagement in respect to helping employees improve their performance. 11. Eliminat
e work standards that set quotas. Work standards must produce quality, not mere
quantity. It is better to take aim at rework, error, and defects. 12. Institute
a massive training program in statistical techniques. Bring statistical techniqu
es down to the level of the individual employees job, and help him to gather info
rmation about the nature of his job in a systematic way. 13. Institute a vigorou
s program for retraining people in new skills. People must be secure about their
jobs in the future and must know that acquiring new skills will facilitate secu
rity. 14. Create a structure in top management that will push every day on the p
revious 13 points. Top management may organize a task force with the authority a
nd obligation to act. This task force will require guidance from an experienced
consultant, but the consultant cannot take on obligations that only the manageme
nt can carry out.
SOURCE: Deming, W. Edwards, Out of the Crisis, pp. 199203, 2000 W. Edwards Deming
Institute, by permission of The MIT Press.
BOX 2.1 Demings Adaptation of the 14 Points for Medical Service (continued)
The New Economics for Industry, Government, Education (Deming 1994) outlines the
Deming System of Profound Knowledge. Deming believed that to transform organiza
tions, the individuals in those organizations need to understand the four parts
of his system of profound knowledge: 1. Appreciation for a system: everything is
related to everything else, and those inside the system need to understand the
relationships within it. 2. Knowledge about variation: this refers to what can a
nd cannot be done to decrease either of the two types of variation.
34
3. Theory of knowledge: this refers to the need for understanding and knowledge
rather than information. 4. Knowledge of psychology: people are intrinsically mo
tivated and different from one another, and attempts to use extrinsic motivators
can result in unwanted outcomes. Demings 14 points and system of profound knowle
dge still provide a roadmap for organizational transformation.
Joseph M. Juran
Joseph Juran was a contemporary of Deming and a student of Shewhart. He began hi
s career at the famous Western Electric Hawthorne Plant, site of the Hawthorne s
tudies (Mayo 1933) related to worker motivation. Western Electric had close ties
to Bell Telephone, Shewharts employer, because the company was the sole supplier
of telephone equipment to Bell. During World War II, Juran served as assistant
administrator for the Lend-Lease Administration. Jurans quality improvement techn
iques made him instrumental in improving the efficiency of processes by eliminat
ing unnecessary paperwork and ensuring the timely arrival of supplies to U.S. al
lies. Jurans Quality Handbook (Juran and Godfrey 1998) was first published in 195
1 and remains a standard reference for quality. Juran was one of the first to de
fine quality from the customer perspective as fitness for use. He compared this de
finition with the alternative, and somewhat confusing, definition of quality as
the number or type of features (Michael Porters differentiation). Jurans contribut
ions to quality include the adaptation of the Pareto principle to the quality ar
ena. According to this principle, 80 percent of defects are caused by 20 percent
of problems, and quality improvement should therefore focus on the vital few to g
ain the most benefit. The roots of Six Sigma programs can be seen in Jurans (1986
) quality trilogy, shown in Table 2.1.
Avedis Donabedian
Avedis Donabedian was born in 1919 in Beirut, Lebanon, and received a medical de
gree from the American University of Beirut. In 1955, he earned a masters degree
in public health from Harvard University. While a student at Harvard, Donabedian
wrote a paper on quality assessment that brought his work to the attention of v
arious experts in the field of public health. He taught for a short period at Ne
w York Medical College before becoming a faculty member at the School of Public
Health of the University of Michigan, where he stayed for the remainder of his c
areer. Shortly after Donabedian joined the University of Michigan faculty, the U
.S. Public Health Service began a project looking at the entire field of health
services research. Donabedian was asked to review and evaluate the literature on
quality assessment for this project. This work culminated in his
35
Basic Quality Processes Quality Planning Identify the customers, both external a
nd internal. Determine customer needs. Develop product features that respond to
customer needs. (Products include both goods and services.) Establish quality go
als that meet the needs of customers and suppliers alike, and do so at a minimum
combined cost. Develop a process that can produce the needed product features.
Prove the process capabilityprove that the process can meet quality goals under o
perating conditions. Choose control subjectswhat to control. Choose units of meas
urement. Establish measurement. Establish standards of performance. Measure actu
al performance. Interpret the difference (actual versus standard). Take action o
n the difference. Prove the need for improvement. Identify specific projects for
improvement. Organize to guide the projects. Organize for diagnosisfor discovery
of causes. Diagnose to find the causes. Provide remedies. Prove that the remedi
es are effective under operating conditions. Provide for control to hold the gai
ns.
Table 2.1 Jurans Quality Trilogy
Control
Improvement
SOURCE: Juran, J. M. 1986. The Quality Trilogy. Quality Progress 19 (8): 1924. Repr
inted with permission from Juran Institute, Inc.
famous article, Evaluating the Quality of Medical Care (Donabedian 1966). This was
followed by a three-volume book set, Exploration in Quality Assessment and Moni
toring (Donabedian 1980, 1982, 1985). Over the course of his career, Donabedian
authored 16 books and more than 100 articles focused on quality assessment and i
mprovement in the healthcare sector on such topics as the definition of quality
in healthcare, relationship between outcomes and process, effect of clinical dec
isions on quality, effectiveness of quality programs, and relationship between q
uality and cost (Sunol 2000). Donabedian (1980) defined healthcare quality in te
rms of efficacy, efficiency, optimality, adaptability, legitimacy, equality, and
cost. Donabedian (1966) was one of the first to view healthcare as a system com
posed of structure, process, and outcome, providing a framework for health servi
ces
36
research still used today. He also highlighted many of the issues still faced in
attempting to measure structures, processes, and outcomes. Donabedian defined o
utcomes as recovery, restoration of function, and survival, but he also included
less easily measured outcomes such as patient satisfaction. Process of care con
sists of the methods by which care is delivered, including gathering appropriate
and necessary information, developing competence in diagnosis and therapy, and
providing preventive care. Finally, structure is related to the environment in w
hich care takes place, including facilities and equipment, medical staff qualifi
cations, administrative structure, and programs. Donabedian (1966, 188) believed
that quality of care is not only related to each of these elements individually
, but also to the relationships among them:
Clearly, the relationships between process and outcome, and between structure an
d both process and outcome, are not fully understood. With regard to this, the r
equirements of validation are best expressed by the concept, already referred to
, of a chain of events in which each event is an end to the one that comes befor
e it and a necessary condition to the one that follows.
Similar to Deming and Juran, Donabedian advocated the continuous improvement of
healthcare quality through structure and process changes supported by outcome as
sessment. The influence of Donabedians seminal work in healthcare can still be se
en. Pay-for-performance programs (structure) reward providers for delivering car
e that meets evidence-based goals assessed in terms of process or outcomes. The
5 Million Lives Campaign (and its predecessor, the 100,000 Lives Campaign; IHI 2
006) is a program (structure) designed to decrease mortality (outcome) through t
he use of evidence-based practices and procedures (process). Not only are assess
ments of process, structure, and outcome being developed, implemented, and repor
ted, but the focus is shifting toward the more systematic view of healthcare adv
ocated by Donabedian.
TQM and CQI, Leading to Six Sigma
The U.S. Navy is credited with coining the term total quality management (TQM) in
the 1980s to describe its approach, informed by Japanese models, to quality mana
gement and improvement (Hefkin 1993). TQM has come to mean a management philosop
hy or program aimed at ensuring quality (defined as customer satisfaction) by fo
cusing on quality throughout the organization and product/service life cycle. Al
l stakeholders in the organization participate in a continuous improvement cycle
. TQM (or continuous quality improvement [CQI], as it is referred to in healthca
re) is not defined by any one organization or individual and has come to encompa
ss the theory and ideas of such quality experts as W. Edwards Deming, Joseph M.
Juran, Philip B. Crosby, Armand V. Feigen-
37
baum, Kaoru Ishikawa, and Avedis Donabedian. TQM may therefore mean different th
ings to different people, and implementation and vocabulary vary from one organi
zation to the next. Possibly as a result, TQM programs have become less popular
in the United States and have been replaced with more codified programs such as
Six Sigma, Lean, and the Baldrige criteria. Six Sigma and TQM are both based on
the teachings of Shewhart, Deming, Juran, and other quality experts. Both TQM an
d Six Sigma emphasize the importance of top management support and leadership, a
nd both focus on continuous improvement as a means to ensure the long-term viabi
lity of an organization. The define-measure-analyze-improve-control cycle of Six
Sigma has its roots in the PDCA cycle (see Chapter 8) of TQM. Six Sigma and TQM
have been described as both philosophies and methodologies. Six Sigma can also
be defined as a metric, or goal, of 3.4 defects per million opportunities, where
as TQM never had that specific goal. TQM was never as clearly defined as Six Sig
ma, nor are certification programs specifically associated with TQM. TQM was def
ined mainly by academics and is more abstract and general, whereas Six Sigma has
its base in industryMotorola and GE were early developersand is more specific, pr
oviding a clear framework for organizations to follow. Early TQM efforts focused
on quality as the primary goal; improved business performance was thought to be
a natural outcome of this goal. Quality departments were mainly responsible for
TQM throughout the organization. While Six Sigma makes quality as defined by th
e customer a primary goal and focuses on tangible results, it also takes into ac
count the effects on business performance. No longer is the focus on quality for
qualitys sake, but rather a quality focus is seen as a means to improve organiza
tional performance. Six Sigma training in the use of specific tools and techniqu
es provides common understanding and common vocabulary both throughout and acros
s organizations; this method makes quality the goal of the entire organization,
not just the quality department. Basically, Six Sigma took the theory and tools
of TQM and codified their implementation, providing a well-defined approach to q
uality so that organizations could more quickly and easily adopt Six Sigma.
ISO 9000
The ISO 9000 series of standards are primarily concerned with quality management
, or how the organization ensures that its products and services satisfy the cus
tomers quality requirements and comply with applicable regulations. The five inte
rnational ISO 9000 standards were first published in 1987 by the International O
rganization for Standardization (ISO). In 2002, the ISO 9000 standard was rename
d ISO 9000:2000; the ISO 9001, 9002, and 9003 standards were consolidated into I
SO 9001:2000; and the ISO 9004 standard was renamed ISO 9004:2000. The standards
are concerned with the processes of
38
ensuring quality rather than the products or services themselves. These standard
s give organizations guidelines to develop and maintain effective quality system
s. Many organizations require that their vendors be ISO certified. For an organi
zation to be registered as an ISO 9000 supplier, it must demonstrate to an accre
dited registrar (third-party organizations who are themselves certified as regis
trars) compliance with the requirements specified in the standard(s). Organizati
ons that are not required by their vendors to be certified can use the standards
to ensure quality systems without attempting to be certified. In the interest o
f improving the quality of healthcare and reducing or maintaining costs, many au
tomotive manufacturers have begun to require that their healthcare providers be
ISO 9000 certified. Driven by this, ISO developed guidelines for implementing IS
O 9000 quality management systems in the healthcare sector, IWA 1, Quality Manage
ment SystemsGuidelines for Process Improvements in Health Service Organizations (I
SO 2005). These guidelines were developed through an International Workshop Agre
ement (IWA) and based on an earlier draft jointly developed by the American Soci
ety for Quality and Automotive Industry Action Group, a worldwide industry assoc
iation representing automotive organizations including Ford, DaimlerChrysler, an
d General Motors. The guidelines are based on ISO 9004:2000, supplemented with s
pecific information to aid healthcare organizations in the attainment of ISO cer
tification and the improvement of quality systems.
Baldrige Award
Japanese automobiles and electronics gained market share in the United States du
ring the 1970s because of their higher quality and lower costs. In the early 198
0s, both U.S. government and industry believed that the only way for the country
to stay competitive was to increase industry focus on quality. The Malcolm Bald
rige National Quality Award was established by Congress in 1987 to recognize U.S
. organizations for their achievements in quality. It was hoped that the award w
ould raise awareness about the importance of quality as a competitive priority a
nd help to disseminate best practices by providing examples of how to achieve qu
ality and performance excellence. The award was originally given annually to a m
aximum of three organizations in each of three categories: manufacturing, servic
e, and small business. In 1999 the categories of education and healthcare were a
dded, and in 2002 the first Baldrige Award in healthcare was given. The healthca
re category includes hospitals, health maintenance organizations, long-term care
facilities, healthcare practitioner offices, home health agencies, health insur
ance companies, and medical/dental laboratories. By 2005, 83 applications had be
en submitted in the healthcare category. Two additional categories, nonprofit an
d government, were added in 2006. The program is a cooperative effort of governm
ent and the private sector. Annual government funding is $5 million, and the act
ual evaluations are performed by a board of examiners that includes experts from
industry, aca-
39
demia, government, and nonprofits. The examiners volunteer their time to review
applications, conduct site visits, and provide applicants with feedback on their
strengths and opportunities for improvement in each of seven categories. Additi
onally, board members give presentations on quality management, performance impr
ovement, and the Baldrige Award. One of the main purposes of the award is the di
ssemination of best practices and strategies. Recipients are asked to participat
e in conferences, provide basic materials on their organizations performance stra
tegies and methods to interested parties, and answer inquiries from the media. B
aldrige Award recipients have gone above and beyond these expectations to give t
housands of presentations aimed at educating other organizations on the benefits
of using the Baldrige framework and disseminating best practices.
JIT, Leading to Lean and Agile
Just-in-time (JIT) is an inventory management strategy aimed at reducing or elim
inating inventory. It is one part of Lean manufacturing, or the Toyota Productio
n System (TPS). The goal of Lean production is to eliminate waste, of which inve
ntory is one form. JIT was the term originally used for Lean production in the U
nited States, where industry leaders noted the success of the Japanese auto manu
facturers and attempted to copy it by adopting Japanese practices. As academics
and organizations realized that Lean production was more than JIT, inventory man
agement terms like big JIT and little JIT were employed, and JIT production beca
me somewhat synonymous with Lean production. For clarity, the term JIT indicates
the inventory management strategy in this text. After World War II, Japanese in
dustry needed to rebuild and grow, and its leaders wanted to copy the assembly l
ine and mass production systems found in the United States. However, they had li
mited resources and limited storage space. At the Toyota Motor Corporation, Taii
chi Ohno and Shigeo Shingo began to develop what has become known as the Toyota
Production System, or TPS. They began by realizing that large amounts of capital
dollars are tied up in inventory in the mass production system that was at that
time typical. Ohno and Shingo sought to reduce inventory by various means, most
importantly by increasing the flow rate of product. Standardization reduced the
number of parts in inventory and the number of tools and machines needed. Proce
sses like single-minute exchange of die (SMED) allowed for quick changeovers of
tooling, increasing the amount of time that could be used for production by redu
cing setup time. As in-process inventory was reduced, large amounts of capital w
ere freed. Customer lead time and uncertainty about orders were reduced as the s
peed of product flow increased throughout the plant. Because inventory provides
a buffer for poor quality, reducing inventory forced Toyota to pay extremely clo
se attention to not only its own quality but suppliers quality as well. To discov
er the best ways to fix the system to reduce inventory, management and line work
ers needed to cooperate, and teams became an integral part of Lean.
40
When the U.S. auto industry began to be threatened by Japanese automobiles, mana
gement and scholars from the United States began to study this Japanese system. Ho
wever, what they brought back were usually the most visible techniques of the pr
ogramJIT, kanbans, quality circles rather than the underlying principles of Lean.
Not surprisingly, many of the first U.S. firms that attempted to copy this syste
m failed; however, some were successful. The Machine that Changed the World (Wom
ack, Jones, and Roos 1990), a study of Japanese, European, and American automobi
le manufacturing practices, first introduced the term Lean manufacturing and broug
ht the theory, principles, and techniques of Lean to a broad audience. Lean is b
oth a management philosophy and a strategy. Its goal is to eliminate all waste i
n the system. Although Lean production originated in manufacturing, the goal of
eliminating waste is easily applied to the service sector. Many healthcare organ
izations are using the tools and techniques associated with Lean to improve effi
ciency and effectiveness. Lean is sometimes seen as being more all encompassing
than TQM or Six Sigma. In order to be a truly Lean organization, it must have qu
ality. In order to be a quality organization, it does not necessarily need to be
Lean. On the other hand, if customers value speed of delivery and low cost, and
quality is defined as customer satisfaction, a quality focus would cause an org
anization to strive to be Lean. Either program would result in the same outcome.
Baldrige, Six Sigma, Lean, and ISO 9000
All of the above are systems or frameworks for performance improvement, and each
has a slightly different focus, tools, and techniques associated with it. Howev
er, all of these programs emphasize customer focus, process or system analysis,
teamwork, and quality, and they are all compatible. As Patrice L. Spath, of Brow
n-Spath & Associates, says, They are tools, techniques, methodologies. Jack Welch
[former GE CEO who helped make Six Sigma a household word] had a visionhe probab
ly could have used any tools or techniques, she adds. Its not the graphics you use;
its the vision of the leaders and how you align and excite organizations about a
chieving that vision (Homa-Lowry, Bertin-Epp, and Spath 2002). The importance of
the culture of the organization and managements ability to shape that culture can
not be overestimated. The successful implementation of any program or deployment
of any technique requires a culture that supports those changes. The leading ca
use of failure of new initiatives is lack of top management support or lack of b
uy-in on the part of employees. Management must truly believe that a particular
initiative will make the organization better and must demonstrate their support
in that belief, both ideologically and financially, to ensure the success of the
initiative. Employee buy-in and support will only happen when top management co
mmitment is evident. Communication and training can aid this process, but only t
rue management commitment will ensure success.
41
Service Typologies
Service typologies, or classification schemes, are primarily used to segment dif
ferent types of services to gain strategic insight into the management and posit
ioning of a particular service (Cook, Goh, and Chung 1999). The focus of these s
ervice typologies varies. A few are based on ownership type (e.g., nonprofit). A
large number are based in a customer/marketing perspective with dimensions such
as product tangibility or type of customer. Finally, some service typologies ar
e more operational in nature. Here, the focus is on the more operational typolog
ies, where typical dimensions include customer contact, capital intensity, custo
mer involvement, and the service process. Prior to the 1970s few typologies exis
ted. During the 1980s, a large number of service typologies were put forth by ex
perts in the field. However, this trend tapered off in the 1990s. Beginning with
Chase (1978), many of these typologies have level or extent of customer contact
as a distinguishing dimension of services. High customer contact services are m
ore difficult to manage and control because they are more variable in nature. Sc
hmenner (1986) extended this idea and developed a service typology similar to th
e product-process manufacturing typology of Hayes and Wheelwright (1979). The Sc
hmenner typology distinguishes services based on the degree of labor intensity a
nd the degree of customer interaction and service customization. Labor intensity
is the relative proportion of labor as opposed to capital (facility and equipme
nt) found in a process. Customer interaction is the degree to which a customer i
s involved in the service process, and service customization is the degree to wh
ich a particular service offering can be tailored for a particular customer. The
resulting matrix was divided into four quadrants, each with a different combina
tion of the distinguishing characteristics: professional service, mass service,
service shop, and service factory. Professional services have high labor intensi
ty and a high degree of customer interaction and service customization. Typicall
y, physicians and lawyers fall into this quadrant of the matrix. Mass services h
ave a high degree of labor intensity and a low degree of interaction and customi
zation. Services found in this quadrant include schools and retail firms. Servic
e shops have a low degree of labor intensity and a high degree of interaction an
d customization. Hospitals are classified in this quadrant. Finally, service fac
tories have a low degree of labor intensity and a low degree of interaction and
customization. Fast-food restaurants such as McDonalds are found in this quadrant
. From a managerial perspective, as labor intensity increases there is a greater
need to focus on workforce issues and less on facility, equipment, and technolo
gical advances. Firms with low customer interaction and customization may have a
more difficult time differentiating themselves in the market. However, costs ca
n be decreased because these organizations benefit from economies of scale and p
rocedures and processes can be highly standardized.
42
Schmenner argued that many of the successful strategic moves within service indu
stries had been toward the diagonal, as shown in Figure 2.3, and up the diagonal
toward a service-factory type of operation. Reducing labor intensity and reduci
ng interaction and customization enabled service firms to become more profitable
. Control in service operations is more difficult than in manufacturing operatio
ns, and Schmenner saw the trend toward the diagonal as a manifestation of the de
sire of many service organizations to gain more control of their operations. Mor
e recently, Schmenner (2004) redefined the axes of his matrix (Figure 2.3) based
on the theory of swift, even flow (Schmenner 2001; Schmenner and Swink 1998), w
hich asserts that a process is more productive as the stream of materials (custo
mers or information) flows more swiftly and evenly. Productivity rises as the sp
eed of flow through the process increases and falls as the variability associate
d with that process increases. In the new service process matrix, the interactio
n and customization axis is defined as variation. When viewed from an operations
standpoint, interaction and customization are sources of variability in a servi
ce process. Schmenner was careful to note that the variation is associated with
the service process rather than variety in the service output. The interaction a
nd customization axis was redefined as relative throughput time. Throughput time
is the period that starts when the customer begins or enters the service proces
s and ends after the service is delivered to the customer and she exits the syst
em. The primary concern here is not the level of labor intensity, but rather how
quickly relative to competitors the service can be delivered to the customer. R
ather than being concerned with profitability, this matrix is related to product
ivity. Examining this new matrix provides insight into service organizations. Pr
ofitable service firms can be found anywhere in the matrix. In fact, service org
anFigure 2.3 Service Process Matrix
Low Relative throughput time Service shop Service factory
Professional service
Mass service
High High Degree of variation Customer interaction and service customization
SOURCE: Adapted from Schmenner, R. W. 2004. Service Businesses and Productivity. D
ecision Sciences 35 (3): 33347.
Low
43
izations may be able to gain competitive advantage by positioning themselves whe
re others are not present. However, service organizations focused on increasing
productivity need to move toward and up the diagonal. Reducing variation or decr
easing throughput time increases the productivity of the organization. In the he
althcare arena, minute clinics and surgical centers provide examples of how heal
thcare organizations have reduced variation or increased throughput to move towa
rd and up the diagonal and increase productivity. Other trends in healthcare als
o support the ideas behind the service process matrix. For example, advanced acc
ess scheduling is aimed at increasing patient throughput, and evidence-based med
icine is aimed at reducing process variation. The service process matrix can hel
p to explain the increasing popularity of such initiatives as Lean, Six Sigma, a
nd supply chain management. Lean initiatives are typically aimed at increasing t
hroughput, whereas quality/Six Sigma initiatives are aimed at reducing variation
and eliminating errors. The rising popularity of these initiatives can be expla
ined by organizations desire to move toward and up the diagonal of the service pr
ocess matrix to increase productivity. Finally, supply chain management can be v
iewed as an organizaitons desire to increase throughput and decrease variability
in the total supply chain system.
Supply Chain Management
The term supply chain management (SCM) was first used in the early 1980s. In 2005,
the Council of Supply Chain Management Professionals agreed on the following de
finition of SCM (Council of Supply Chain Management Professionals 2006):
Supply Chain Management encompasses the planning and management of all activitie
s involved in sourcing and procurement, conversion, and all Logistics Management
activities. Importantly, it also includes coordination and collaboration with c
hannel partners, which can be suppliers, intermediaries, thirdparty service prov
iders, and customers. In essence, Supply Chain Management integrates supply and
demand management within and across companies.
This definition makes it apparent that SCM is a broad discipline, encompassing a
ctivities outside as well as inside an organization. SCM has its roots in system
s thinking. Systems thinking is based on the idea that everything affects everyth
ing else. The need for systems thinking is predicated on the fact that it is poss
ible, and even likely, to optimize one part of a system while the whole system i
s suboptimal. A current example of a suboptimal system in healthcare can be seen
in prescription drugs. In the United States, the customer can optimize his drug
purchases (minimize cost) by purchasing drugs from pharmacies located in foreig
n countries (e.g., Canada, Mexico). Often, these drugs are manufactured in the U
nited States. While the customer has minimized his costs, the total supply chain
has
44
additional costs from the extra transportation incurred by shipping drugs to Can
ada or another foreign country and then back to the United States. SCM became in
creasingly important to manufacturing organizations in the late 1990s, driven by
the need to decrease costs in response to competitive pressures and enabled by
technological advances. As manufacturing became more automated, labor costs as a
percentage of total costs decreased, and the percentage of material and supply
costs increased. Currently, 70 to 80 percent of the cost of a manufactured good
is in purchased materials and services and less than 25 percent is in labor (Bur
eau of Economic Analysis 2006). Consequently, there are fewer opportunities for
reducing the cost of goods through decreasing labor and more opportunities assoc
iated with managing the supply chain. Additionally, advances in information tech
nology allowed firms to collect and analyze the information needed to more effic
iently manage their supply chains. Beginning with the inventory management syste
ms of the 1970s, including materials requirements planning, followed by the ente
rprise resource planning systems of the 1990s, SCM was enabled by technology. As
industry moved to more sophisticated technological systems for managing the flo
w of information and goods, the ability to collect and respond to information ab
out the entire supply chain increased, and firms were able to actively manage th
eir supply chains. In the healthcare industry, interest in SCM has lagged behind
that in the manufacturing sector partly because deployment of information techn
ology has been slower, but mainly because of the labor intensity found in most h
ealthcare organizations. Labor still accounts for 50 percent of the cost of heal
thcare (Bureau of Economic Analysis 2006). There is still opportunity to reduce
labor costs. However, SCM is becoming increasingly important with the increasing
focus on reducing the costs of healthcare and the need to reduce those costs th
rough the development of more efficient and effective supply chains.
Conclusion
Service organizations in general and healthcare organizations in particular have
lagged in their adoption of process improvement philosophies, techniques, and t
ools of operations management, but they no longer have this option. Healthcare o
rganizations are facing increasing pressures from consumers, industry, and gover
nments to deliver their services in a more efficient and effective manner and ne
ed to adopt these new philosophies to remain competitive. In healthcare today, org
anizations such as IHI and AHRQ are leading the way in the development and disse
mination of tools, techniques, and programs aimed at improving the quality, safe
ty, efficiency, and effectiveness of the healthcare system. Although these tools
and techniques have been adapted for contemporary healthcare, their roots are i
n the past, and an understanding of this history (Figure 2.4) can enable organiz
ations to move successfully into the future.
FIGURE 2.4
W. Edwards Deming (U.S.) Henry Ford Mass production Genichi Taguchi Cost of vari
ation Harlan Cleveland Knowledge hierarchy Kaoru lshikawa TQM, fishbone
Important Events in Performance Improvement
Frederick Winslow Taylor Father of scientific management
Frank and Lillian Gilbreth Time and motion Walter A. Shewhart Grandfather of qua
lity movement If Japan Can, Why Cant We? Shigeo Shingo Poka-Yoke and SMED Eliyahu
M. Goldratt TOC Russell L. Ackoff Systems thinking
Institute for Healthcare Improvement James Womack TQM The Machine that JIT Chang
ed the World Avedis Robert S. Kaplan Donabedian Balanced scorecard
Venice Arsenal First moving assembly line W. Edwards Deming Father of quality mo
vement (Japan)
Baldrige Award ISO 9000
A. Erlang Queueing
Adam Smith Specialization of labor Joseph M. Juran Quality trilogy TPS
IOM report To Err Is Human Baldrige Award in Healthcare AHRQ Six Sigma SCM 100k
Lives
Florence Nightingale
1300 1925
1900
1950 CPM method PERT method
1975 Project Management Institute
2000
Henry Gantt Gantt charts
45
46
1. What is the difference between data, information, knowledge, and wisdom? Give
specific examples of each in your own organization. 2. How has operations manag
ement changed since its early days as scientific management? 3. What are the maj
or factors leading to increased interest in the use of operations management too
ls and techniques in the healthcare sector? 4. Why has ISO 9000 certification be
come important to healthcare organizations? 5. Research firms that have won the
Baldrige Award in the healthcare sector. What factors led to their success in wi
nning the award? 6. What are some of the reasons for the success of Six Sigma? 7
. What are some of the reasons for the success of Lean? 8. How are Lean initiati
ves similar to TQM and/or Six Sigma initiatives? How are they different? 9. Why
is SCM becoming increasingly important to healthcare organizations?
References
Ackoff, R. L. 1989. From Data to Wisdom. Journal of Applied Systems Analysis 16: 39
. Allen, S. 2004. Emergency Room Recovery. The Boston Globe July 18: A1. Auerbach,
A. D. 2001. Making Health Care Safer: A Critical Analysis of Patient Safety Prac
tices, In Localizing Care to High-Volume Centers. Evidence Report/Technology Asse
ssment: No. 43, AHRQ Publication No. 01-E058. Rockville, MD: Agency for Healthca
re Research and Quality. Bellis, M. 2006. Henry Ford (18631947). [Online informatio
n; retrieved 6/8/06.] http://inventors.about.com/library/inventors/blford.htm. B
ureau of Economic Analysis, U.S. Department of Commerce. 2006. Benchmark Input-Ou
tput Accounts. [Online information; retrieved 12/20/06.] http://bea.gov/bea/dn2/h
ome/benchmark.htm. Centers for Medicare & Medicaid Services (CMS), Office of the
Actuary. 2005. Data from the National Health Statistics Group: NHE Summary Incl
uding Share of GDP, CY 19602004. Baltimore: CMS [Online information; retrieved 6/
8/06.] www.cms.hhs.gov/NationalHealthExpendData/downloads/ nhegdp04.zip. Chase,
R. B. 1978. Where Does the Customer Fit in a Service Operation? Harvard Business R
eview 56 (6): 13742. Cleveland, H. 1982. Information as a Resource. The Futurist 16
(Dec.): 3439. Cohen, I. B. 1984. Florence Nightingale. Scientific American 250 (3)
: 12837. Cook, D. P., C. Goh, and C. H. Chung. 1999. Service Typologies: A State o
f the Art Survey. Production and Operations Management 8 (3): 31838.
47
Council of Supply Chain Management Professionals. 2006. Supply Chain Management/
Logistics Management Definitions. [Online information; retrieved 12/20/06.] www.c
scmp.org/Website/AboutCSCMP/Definitions/Definitions.asp. Deming, W. E. 1998. The
Deming Philosophy. [Online information; retrieved 6/9/06.] http://deming.ces.clem
son.edu/pub/den/deming_philosophy.htm. . 1994. The New Economics for Industry, Gove
rnment, Education, 2nd ed. Cambridge, MA: MIT Press. ______. 1986. Out of the Cr
isis. Cambridge, MA: Cambridge University Press. Donabedian, A. 1985. Exploratio
n in Quality Assessment and Monitoring, Volume 3. The Methods and Findings of Qu
ality Assessment and Monitoring: An Illustrated Analysis. Ann Arbor, MI: Health
Administration Press. . 1982. Exploration in Quality Assessment and Monitoring, Vol
ume 2. The Criteria and Standards of Quality. Ann Arbor, MI: Health Administrati
on Press. . 1980. Exploration in Quality Assessment and Monitoring, Volume 1. Defin
ition of Quality and Approaches to its Assessment. Ann Arbor, MI: Health Adminis
tration Press. . 1966. Evaluating the Quality of Medical Care. Milbank Memorial Fund
Quarterly 44 (3): 166206. Finnie, W. 1997. Strategies, Systems, and Organizations:
An Interview with Russell L. Ackoff. Strategy and Leadership 25 (2): 2228. Gilbre
th, F. B., and E. G. Carey. 1948. Cheaper by the Dozen. New York: T.Y. Crowell C
o. Halm, E. A., C. Lee, and M. R. Chassin. 2002. Is Volume Related to Outcome in
Health Care? A Systematic Review and Methodologic Critique of the Literature. Ann
als of Internal Medicine 137 (6): 51120. Hayes, R. H., and S. C. Wheelwright. 197
9. Link Manufacturing Process and Product Life Cycles. Harvard Business Review 57
(1): 13340. Hefkin, D. C. 1993. The Navys Quality Journey: Operational Implementat
ion of TQL. Executive Research Project CS5. Washington, D.C.: Industrial College
of the Armed Forces. Heskett, J. L. 2003. Shouldice Hospital Ltd. Harvard Busin
ess School Case 9-683068. Boston: Harvard Business School Publishing. Homa-Lowry
, J., D. Bertin-Epp, and P. L. Spath. 2002. Enhance Pursuit of Excellence by Inte
grating ISO 9000, Baldrige: National Baldrige Winners Often Use ISO as Platform.
HealthCare Benchmarks and Quality Improvement 9 (9): 2529. Institute for Healthca
re Improvement (IHI). 2006. Protecting 5 Million Lives from Harm. [Online informat
ion; retrieved 6/20/07.] www.ihi.org/IHI/Programs/ Campaign/Campaign.htm. Instit
ute of Medicine. 1999. To Err Is Human: Building a Safer Health System. Washingt
on, D.C.: National Academies Press. International Organization for Standardizati
on (ISO). 2005. IWA 1, Quality Management SystemsGuidelines for Process Improvemen
ts in Health Service Organizations. [Online white paper; retrieved 10/29/07.] www
.iso.org.
48
International Work Simplification Institute. 1968. Pioneers in Improvement and Ou
r Modern Standard of Living. IW/SI News Sept. 18. [Online information; retrieved
6/8/06.] http://gilbrethnetwork.tripod.com/bio.html. Juran, J. M. 1986. The Quali
ty Trilogy. Quality Progress 19 (8): 1924. Juran, J. M., and A. B. Godfrey. 1998.
Jurans Quality Handbook, 5th ed. New York: McGraw-Hill. Lang, W. 1950. Cheaper by
the Dozen. Beverly Hills, CA: Twentieth Century Fox Film Corporation. Levy, S.
2003. Cheaper by the Dozen. Beverly Hills, CA: 20th Century Fox. Mason, C. C. 19
80. If Japan Can, Why Cant We? New York: NBC. Mayo, E. 1933. The Human Problems o
f an Industrial Civilization. New York: MacMillan. McCormick, M. 2006. Immanuel K
ant (17241804) Metaphysics. [Online information; retrieved 6/8/06.] www.iep.utm.ed
u/k/kantmeta.htm. NationMaster.com. 2004. Venetian Arsenal. [Online information; r
etrieved 6/8/06.] www.nationmaster.com/encyclopedia/Venetian-Arsenal. Nelson, D.
1980. Frederick Taylor and the Rise of Scientific Management. Madison: Universi
ty of Wisconsin Press. NetMBA.com. 2005. Frederick Taylor and Scientific Manageme
nt. [Online information; retrieved 6/8/06.] www.netmba.com/mgmt/scientific/. Neuh
auser, D. 2003. Florence Nightingale Gets No Respect: As a Statistician That Is. Q
uality & Safety in Health Care 12 (4): 317. Nightingale, F. 1999. Florence Night
ingale: Measuring Hospital Care Outcomes. Oakbrook Terrace, IL: Joint Commission
on Accreditation of Healthcare Organizations. . 1858. Notes on Matters Affecting t
he Health, Efficiency and Hospital Administration of the British Army. Presented
by request to the Secretary of State for War. Privately printed for Miss Nighti
ngale by Harrison and Sons. Perkins, J. S. 1997. Frank B. Gilbreths Research: The
Quest of the One Best Way. THE QUEST, Newsletter of the Gilbreth Network 1 (2). [
Online information; retrieved 10/29/07.] http://gilbrethnetwork.tripod.com/qv1n2
.html. Ressler, T., and M. Ahrens. 2006. The Decision Making Book. Minneapolis,
MN: University of St. Thomas. Schmenner, R. W. 2004. Service Businesses and Produ
ctivity. Decision Sciences 35 (3): 33347. . 2001. Looking Ahead by Looking Back: Swift
, Even Flow in the History of Manufacturing. Production and Operations Management
10 (1): 8796. . 1986. How Can Service Businesses Survive and Prosper? Sloan Managemen
t Review 27 (3): 2132. Schmenner, R. W., and M. L. Swink. 1998. On Theory in Opera
tions Management. Journal of Operations Management 17 (1): 97113. Shewhart, W. A.
1931. Economic Control of Quality of Manufactured Product. New York: D. Van Nost
rand Company, Inc. Shewhart, W. A., and W. E. Deming. 1939. Statistical Method f
rom the Viewpoint of Quality Control. Washington, D.C.: The Graduate School, U.S
. Department of Agriculture.
49
Simkin, J. 2005. Henry Ford. [Online information; retrieved 6/8/06.] www.spartacus
.schoolnet.co.uk/USAford.htm. Sunol, R. 2000. Avedis Donabedian. International Jou
rnal for Quality in Health Care 12 (6): 45154. Taylor, F. W. 1911. The Principles
of Scientific Management. New York: Harper and Row. Womack, J. P., D. T. Jones,
and D. Roos. 1990. The Machine that Changed the World: Based on the Massachuset
ts Institute of Technology 5-Million Dollar 5-Year Study on the Future of the Au
tomobile. New York: Rawson Associates. Zeleny, M. 1987. Management Support System
s: Towards Integrated Knowledge Management. Human Systems Management 7 (1): 5970.
CHAPTER
3
EVIDENCE-BASED MEDICINE AND PAY-FOR-PERFORMANCE
CHAPTER OUTLINE
Operations Management in Action Overview Evidence-Based Medicine Standard and Cu
stom Patient Care Financial Gains from EBM Chronic Care Tools to Expand the Use
of EBM Public Reporting Pay-for-Performance P4P methods Issues in P4P Examples o
f P4P Bridges to Excellence Integrated Healthcare Association Premier Hospital D
emonstration Tiering Patient Choice Minnesota Advantage Vincent Valley Hospital
and Health System and P4P Using EBM and Public Reporting to Advance Health Polic
y: A Proposal Conclusion Discussion Questions References
50

ambulatory care sensitive condition (ACSC) Bridges to Excellence (BTE) care mana
gement process certificate of need (CON) clinical guideline clinical protocol cu
stom patient care evidence-based medicine (EBM) Integrated Healthcare Associatio
n (IHA) Medicare Payment Advisory Commission (MedPAC) Minnesota Advantage Patien
t Choice prevention quality indicator (PQI) pay-for-performance (P4P) Premier Ho
spital Demonstration public reporting risk adjustment standard patient care tier
ing
51

In December 2004, the Institute for Healthcare Improvement (IHI) began the 100,0
00 Lives Campaign, the first-ever national campaign to promote saving a specifie
d number of lives in hospitals by a certain date (June 14, 2006) through the imp
lementation of proven, evidence-based practices and procedures. IHI determined t
hat evidence in the medical literature indicated that by implementing six proven
interventions hospitals could prevent 100,000 deaths in the United States. The
targeted interventions (IHI 2006) were: 1. Deploy rapid-response teams by allowi
ng any staff member, regardless of position in the chain of command, to call upo
n a specialty team to examine a patient at the first sign of decline; 2. Deliver
reliable evidence-based care for acute myocardial infarction by consistently de
livering key measures, including early administration of aspirin and beta-blocke
rs, that prevent patient deaths from heart attack; 3. Prevent adverse drug event
s by implementing medication reconciliation, which requires that a list of all o
f a patients medications (even for unrelated illnesses) be compiled and reconcile
d to ensure that the patient is given (or prescribed) the right medications at t
he correct dosagesat admission, discharge, and before transfer to another care un
it; 4. Prevent central line infections by consistently delivering five interdepe
ndent, scientifically grounded steps collectively called the central line bundle;
5. Prevent surgical-site infections by reliably delivering the correct periopera
tive antibiotics, maintaining glucose levels, and avoiding shaving hair at the s
urgical site; and 6. Prevent ventilator-associated pneumonia by implementing fiv
e interdependent, scientifically grounded steps collectively called the ventilato
r bundle, for example, elevating the head of the hospital bed to between 30 and 4
5 degrees, thereby dramatically reducing mortality and length of stay in the int
ensive care unit. One of the key interventions was the implementation of rapid-r
esponse teams. Julia Herzenberg of Kaisers San Francisco hospital led its 100,000
Lives Campaign. She helped assemble her hospitals rapid-response team and says (
Kertesz 2005): Several studies have shown that patients show some instability fo
r a period prior to cardiac arrest. Now, often nothing is done. With a rapid res
ponse team, if a nurse or even a receptionist is concerned about a patient, they
52
C h a p t e r 3 : Ev i d e n c e - B a s e d M e d i c i n e a n d Pa y - Fo rPe r f o r m a n c e
53
can call up the rapid response team, which does an assessment as to whether an i
ntervention is needed. The hope is that it will decrease deaths and the number o
f people going to ICUs. More than 2,800 hospitals in the United States participa
ted in the 100,000 Lives Campaign, and in June 2006 IHI reported that it had exc
eeded its target, estimating that 123,000 needless deaths had been averted. IHI
has built on this success and has now embarked on a 5 Million Lives Campaign.
Overview
The science of medicine advanced rapidly through the latter half of the twentiet
h century with advances in pharmaceuticals, surgical techniques, and laboratory
and imaging technology and the rapid subspecialization of medicine itself. This a
ge of miracles improved health and lengthened life spans. In the mid 1960s, the f
ederal government began the Medicare and Medicaid programs, and this new source
of funding fueled the growth and expansion of the healthcare delivery system. Un
fortunately, because this growth was so explosive, many new tools and clinical a
pproaches were initiated that had little scientific merit. As these clinical app
roaches were used broadly, they became community standards. In addition, many high
ly effective, simple clinical tools and techniques were not being used consisten
tly. In response to these problems, a number of courageous clinicians began the
movement that has resulted in what is known today as evidence-based medicine (EB
M). According to the Centre for Evidence-Based Medicine (2006), Evidence-based me
dicine is the conscientious, explicit and judicious use of current best evidence
in making decisions about the care of individual patients. In almost all cases,
the broad application of EBM not only improves clinical outcomes for patients bu
t reduces costs in the system as well. This chapter reviews: The history and cur
rent status of EBM; Standard and custom care; Public reporting, pay-for-performa
nce (P4P), and tiering; and Issues in the use of P4P and EBM for changing clinic
ian behavior.
EBM is explored in depth, followed by
se principles to encourage the use of
cluded in other chapters of this book
pter concludes with an example of how
an examination of how payers are using the

EBM by clinicians. The operations tools in
are linked to achieving EBM goals. The cha
a relatively simple government
54
policy based on EBM could be implemented to drive significant cost reductions an
d quality improvements in the U.S. healthcare system. The companion web site at
ache.org/books/OpsManagement contains many additional references and, because th
ese topics are changing rapidly, will be updated frequently.
Evidence-Based Medicine
The expansion of clinical knowledge has three major phases. First, basic researc
h is done in the lab and with animal models. Second, carefully controlled clinic
al trials are conducted to demonstrate the efficacy of a new diagnostic or treat
ment methodology. Third, the clinical trial results are translated into clinical
practice. This final phase of translation is where the system breaks down. As C
laude Lenfant (2003, 869), director of the National Heart, Lung, and Blood Insti
tute, has stated:
Regardless of the reasons for this phenomenonstructural, economic or motivationalt
he result is the same: we are not reaping the full public health benefits of our
investment in research. . . And it is not just recent research results that are
not finding their way into clinical practices and public health behaviors, ther
e is plenty of evidence that old research outcomes have been lost in translation a
s well.
Lenfant cites the poor application of beta-blockers for heart attack patients as
an example. Only 62.5 percent of patients were getting this treatment 15 years
after its high level of effectiveness had been demonstrated. He cites another si
mple and inexpensive treatmentaspirin for a variety of cardiovascular diseasesthat
was being given to just 33 percent of appropriate patients as of 2000. Many oth
er studies have documented this same barrier to the translation of clinical tria
ls to extensive adoption by practitioners. The cure to this wide variation in pr
actice is the consistent application of EBM. The major tool is the clinical guid
eline (also known as a protocol), which the Institute of Medicine defines as syst
ematically developed statements to assist practitioner and patient decisions abo
ut appropriate health care for specific clinical circumstances (Timmermans and Ma
uck 2005, 18). One source for such information is the National Guideline Clearin
ghouse (2006), a comprehensive database of evidence-based clinical practice guid
elines and related documents that contains more than 4,000 guidelines. What are
the barriers to the wider application of EBM? Timmermans and Mauck (2005) postul
ate that critics of EBM question the science and trials that were used to develo
p clinical guidelines. Another perceived barrier is the challenge to professiona
l autonomythe concern that guideline use will
55
override the patients specific and unique needs. Another frequently cited critici
sm is that EBM will decrease variation so much that natural discoveries and new cl
inical approaches will not occur. However, even in the face of these criticisms,
many leading clinicians have moved to embrace EBM and consider the current chal
lenge to be discovering the optimal methods to expand its use. Most guidelines a
re developed by committees of physicians within professional associations or med
ical groups based on a review of the salient EBM literature. Once guidelines hav
e been developed, Timmermans and Mauck (2005) advocate for multidisciplinary pro
ject teams to customize the guideline and its application to specific organizati
ons. One author of this book (McLaughlin) used this approach at Hennepin County
Medical Center in Minneapolis, which now uses clinical guidelines for more than
50 percent of its inpatient care.
Standard and Custom Patient Care
One continuing criticism of EBM is that all patients are unique, and EBM is cookb
ook medicine that only applies to a few patients. EBM proponents counter this arg
ument with simple examples of well-accepted and effective clinical practices tha
t are inconsistently followed. A more productive view of the mix of the art and
science in medicine is provided by Bohmer (2005), who suggests that all healthca
re is a blend of custom and standard care. Figure 3.1 shows the four currently u
sed models that blend these two approaches. FIGURE 3.1 Four Approaches to Blendi
ng Custom and Standard Processes
(A) Separate and select
(B) Separate and accommodate
(C) Modularized
(D) Integrated
I
I
I
I
O
Customized subprocess
O
O
Sorting process
O
Reasoning process
O I Input O Output
Standard subprocess
SOURCE: Bohmer, R. M. J. 2005. Medicines Service Challenge: Blending Custom and St
andard Care. Health Care Management Review 30 (4): 32230. Used with permission.
56
Model A (separate and select) provides an initial sorting by patients themselves
. Those with standard problems are treated with standard care using EBM guidelin
es. Examples of this type of system are specialty hospitals for laser eye surger
y or walk-in clinics operating in pharmacies and retail outlets. Patients who do
not fit the providers homogenous clinical conditions are referred to other provi
ders who can deliver customized care. Model B (separate and accommodate) combine
s the two methods inside one provider organization. The Duke University Health S
ystem has developed standard protocols for its cardiac patients. Patients are in
itially sorted, and those who can be treated with the standard protocols are car
ed for by nurse practitioners using a standard care model. Cardiologists care fo
r the remainder using custom care. However, on every fourth visit to the nurse p
ractitioner, the cardiologist and nurse practitioner review the patients case tog
ether to ensure that standard care is still the best treatment approach. Model C
(modularized) is used when the clinician moves from the role of care provider t
o that of architect of care design for the patient. In this case, a number of st
andard processes are assembled to treat the patient. The Andrews Air Force Base
clinic uses this system to treat hypertension patients. After an initial evaluati
on, treatment may include weight control, diet modification, drug therapy, stres
s control, and ongoing surveillance. Each component may be provided by a separat
e professional and sometimes a separate organization. What makes the care unique
ly suited to each patient is the combination of components (Bohmer 2005, 326). Mo
del D (integrated) combines both standard care and custom care in a single organ
ization. In contrast to Model B, each patient receives a mix of both custom and
standard care based on his or her condition. Intermountain Healthcare System emp
loys this model through the use of 62 standard care processes available as proto
cols in its electronic health record. These processes cover the care of over 90%
of patients admitted in IHC hospitals (Bohmer 2005, 326). Clinicians are encourag
ed to override elements in these protocols when it is in the best interest of th
e patient. All of these overrides are collected and analyzed, and changes are ma
de to the protocol, which is an effective method to continuously improve clinica
l care. All of the tools and techniques of operations improvement included in th
e remainder of this book can be used to make standard care processes operate eff
ectively and efficiently.
Financial Gains from EBM
EBM has the potential to not only improve clinical outcomes but also to decrease
total cost in the healthcare system. The Agency for Healthcare Research and Qua
lity (2006) has identified 14 prevention quality indicators (PQIs) (Table 3.1).
This set of measures can be used with hospital inpatient dis-
57
Number 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Prevention Quality Indicator Diabetes short-term complication admission rate Per
forated appendix admission rate Diabetes long-term complication admission rate C
hronic obstructive pulmonary disease admission rate Hypertension admission rate
Congestive heart failure admission rate Low birth weight Dehydration admission r
ate Bacterial pneumonia admission rate Urinary tract infection admission rate An
gina admission without procedure Uncontrolled diabetes admission rate Adult asth
ma admission rate Rate of lower-extremity amputation among patients with diabete
s
TABLE 3.1 Prevention Quality Indicators
SOURCE: Agency for Healthcare Research and Quality (2006).
charge data to identify ambulatory care sensitive conditions (ACSCs). ACSCs are
conditions for which good ambulatory care can potentially prevent the need for h
ospitalization, or for which early intervention can prevent complications or mor
e severe disease. Each PQI can be targeted for improvement by healthcare provide
rs through the use of guidelines and a mix of standard and custom care. A study
undertaken by the Minnesota Department of Health on inpatient admissions in 2003
showed that $371 million could be saved each year if hospital admissions for AC
SCs (9 percent of all inpatient admissions) were eliminated (Gildemeister 2006).
Chronic Care
Casalino et al. (2003) studied how medical groups choose to implement EBM guidel
ines. They found four major strategies, which they named care management process
es (CMPs): 1. case management; 2. guidelines adopted by the group and available
in the chart; 3. feedback to physicians on the care they deliver compared with g
uidelinerecommended care; and 4. disease registries to track patients with chron
ic conditions. Casalino et al. (2003) surveyed 1,587 physician organizations thr
oughout the United States in 2002 and reviewed the care of patients for four chr
onic conditions: diabetes, asthma, congestive heart failure, and depression. Of
58
the 16 possible uses (4 diseases 4 CMPs) of CMPs, they found the national averag
e to be 5.1. This low rate of usage (1.27 per disease) has led health plans, pri
vate buyers, and government agencies to devise strategies to more rapidly increa
se the use of EBM and related clinical management strategies.
Tools to Expand the Use of EBM
Organizations that are outside the healthcare delivery system itself have used t
he increased acceptance of EBM as the basis for new programs focused on encourag
ing increased implementation of EBM. These programs, sometimes called value purch
asing, include public reporting of clinical results, P4P, and tiered systems of c
are.
Public Reporting
Although strongly resisted by clinicians for many years, public reporting has co
me of age. The Centers for Medicare & Medicaid Services (CMS) now report the per
formance of hospitals, long-term care facilities, and medical groups online (e.g
., www.hospitalcompare.hhs.gov). Many health plans also report performance and t
he prices of providers in their networks to assist their plan members, particula
rly those with consumer-directed health insurance products. In addition, communi
ty-based public reporting has arisen nationally with medical group reporting in
a number of states. Leading examples include: California Cooperative Healthcare
Reporting Initiative, San Francisco Massachusetts Health Quality Partners, Water
town Minnesota Community Measurement, St. Paul Wisconsin Collaborative for Healt
hcare Quality, Madison
Although it is a growing field, a number of issues surround public reporting. Th
e first and most prominent is risk adjustment. Most clinicians feel their patien
ts are sicker than average and that contemporary risk adjustment systems do not ad
equately account for this. Patient compliance is another challenging aspect of p
ublic reporting. If a doctor follows EBM guidelines for diagnosis and treatment
but the patient does not take her medication, should the doctor be given a poor
grade? One of the anticipated effects of public reporting is that patients will
use the Internet to shop for quality healthcare products as they might for an au
tomobile or television. Currently, few patients use these public systems to guid
e their buying decisions. However, clinical leaders do review the public reports
and target improvement efforts to areas where they have poor performance compar
ed with their peers.
59
Hibbard, Stockard, and Tusler (2003) studied public reporting of quality in hosp
itals in Wisconsin. One group of hospitals had its quality metrics publicly repo
rted, another group received a private report, and a third group received no dat
a. The researchers found a significant difference in the quality improvement act
ivities by those hospitals that had their quality reports made public. For examp
le, in the area of obstetric care the publicreport hospitals initiated an averag
e of 3.4 projects, the private-report hospitals had 2.5 projects, and the no-rep
ort hospitals had 2.0 projects. Shaller et al. (2003) recommend five strategies
that will be necessary to make public reporting effective and widely used. First
, consumers need to be convinced that quality problems are real and can be impro
ved; the mindset of My doctor provides high-quality caredont they all? must be overc
ome. Second, purchasers and policymakers must ensure that quality reporting is s
tandardized and universal. Third, consumers must be given quality information th
at is relevant and easy to understand and use. Fourth, dissemination methods nee
d to be optimized. Fifth, quality improvement efforts undertaken by providers ne
ed to be rewarded with systems such as P4P and tiering, as described in the foll
owing sections.
Pay-for-Performance
Another logical tool to expand the use of EBM is the financing system. Many buye
rs of healthcare are now installing P4P systems to encourage providers to delive
r EBM care.
P4P methods Rosenthal et al. (2004) identified more than 31 P4P projects coverin
g more than 20 million individuals under way in the United States. In general, P
4P systems add payments to the amount that would otherwise be paid to a provider
. To obtain these additional payments, the provider must demonstrate that he is
delivering care that meets clinical EBM goals. These clinical measures can be ei
ther process or outcome measures. Although many providers would prefer to be mea
sured on outcomes, this approach is difficult to use as some outcomes need to be
measured over many years. In addition, some providers have a small number of pa
tients in a particular clinical group so outcome results can vary dramatically.
Therefore, process measures are used for many conditions; these measures are bac
ked by extensive EBM literature. For example, a patient with diabetes whose bloo
d pressure is maintained in a normal range will experience fewer complications t
han one whose blood pressure is uncontrolled. Blood pressure can be measured and
reported at every visit, whereas complications will occur infrequently. Rosenth
al et al. (2004) also found that most of the P4P programs in place reward high-p
erforming medical groups and hospitals. This, of course, rewards these organizat
ions for past investments in EBM and systems. However,
60
if the intent of P4P is to encourage broader use of EBM, payment systems will al
so need to be developed to provide incentives for providers who improve as well
as those who are the highest performers. The Medicare Payment Advisory Commissio
n (MedPAC) advises Congress on improvements in the Medicare system. As P4P use h
as grown in commercial markets, MedPAC has made recommendations to improve Medic
ares payment systems by adding P4P (Milgate and Cheng 2006). MedPAC determined a
number of criteria that should be employed in this policy change. The commission
recommends that in a Medicare P4P system evidence-based measures should be read
ily available and data collection should not be burdensome. In addition, appropr
iate risk-adjustment tools need to be available and quality goals attainable thr
ough operations improvements such as implementing the CMPs of Casalino et al. (2
003). Using these criteria, MedPAC suggested a number of design principles. Prov
iders should be rewarded for both high performance and improvement. Although Med
PAC quotes some studies asserting that effective P4P must provide payments of 5
percent to 20 percent in additional funds for physicians and 1 percent to 4 perc
ent for hospitals, MedPAC recommends a P4P payment increase of only 1 to 2 perce
nt for Medicare because of its large presence in the market. Finally, MedPAC rec
ommends that an initial measurement set should be determined but also suggests t
hat this set should be evaluated and updated frequently.
Issues in P4P P4P is a system that has breached the wall of professional autonom
y to influence the day-to-day care of many patients. However, significant system
changes frequently result in unintended consequences no matter how well they ha
ve been designed. Gosfield (2004) enumerates a number of challenges in current v
ersions of P4P. Once P4P programs change behavior in providers, will payers ceas
e to reward providers? How will P4P incentives be allocated from one disease to
another? For example, will the P4P payments for patients with diabetes be higher
than for patients with low back pain, and by what rationale if so? Finally, whe
re does the P4P funding come fromis it just another version of a discount or with
hold? Another interesting issue is the intersection of existing payment systems
and P4P. For example, if a physician orders additional tests to meet P4P guideli
nes, will she be penalized by a health plan for excessive testing in aggregate?
Another common concern of providers is the trade-off of new revenue versus the a
dditional cost of the administrative resources needed to provide the data. Final
ly, risk adjustment and patient compliance also complicate the administration of
P4P methods. Roland (2004) has examined the effects of P4P in the United Kingdo
m and believes early signs in this system point to a number of positive and
61
negative effects. Positive changes in the system include the expansion of clinic
al computing systems, the expansion of the role of nurses, increases in clinics
that specialize in one disease, and improved health outcomes. Possible challenge
s in the new P4P system include fragmentation of care for patients with multiple
illnesses, loss of a holistic approach, reductions in quality for those conditi
ons not included in the system, and increased administrative costs. Galvin (2006
) reports on interesting initial results:
Physicians were scored on 146 indicators of quality, with clinical quality accou
nting for more than 50 percent of the total. Each point earned had a financial b
onus associated with it, and general practitioners stood to achieve compensation
amounting to 30 percent of their salary. This represented $1.8 billion in new m
oney, a 20 percent increase in the National Health Service budget. First-year re
sults are in: [General practitioners] greatly exceeded projections of their perf
ormance and achieved an eye-popping mean of more than 91 percent compliance with
clinical guidelines. This resulted in payments estimated at $700 million more t
han expected.
Examples of P4P P4P is a relatively new and growing phenomenon, so a number of e
xamples now exist but many new organizations are implementing P4P programs every
year. Three representative examples are discussed below. Updated examples will
be continuously added to this books companion web site. Bridges to Excellence Bri
dges to Excellence (BTE) is a multistate, multiemployer coalition developed by e
mployers, physicians, healthcare services researchers, and other industry expert
s (BTE 2006). The BTE program is a grantee of the Robert Wood Johnson Foundation
and has a mission of rewarding quality across the healthcare system. The Physic
ian Office Link BTE program enables physician office sites to qualify for bonuse
s based on implementation of specific processes to reduce errors and increase qu
ality. Sites can earn up to $50 per year for each patient covered by a participa
ting employer and plan. Diabetes Care Link enables physicians to achieve one- or
three-year recognition for high performance in diabetes care. Qualifying physic
ians receive up to $80 for each patient with diabetes covered by a participating
employer and plan. This program has shown significant quality gains. For partic
ipants from 1997 to 2003:
The average rate of patients with diabetes who had hemoglobin A1c (HbA1c) levels
of less than 7 percent increased from 25 percent to 46 percent, an indication t
hat more adults with diabetes are maintaining
62
proper control of HbA1c (a measure of average blood glucose over the previous th
ree months). The rate of patients with diabetes who had controlled low-density l
ipoprotein (LDL) cholesterol below 100 mg/dL rose from 17 percent to 45 percent.
The rate of patients with diabetes who were monitored for kidney disease rose f
rom 60 percent to 85 percent. Cardiac Care Link enables physicians to achieve th
ree-year recognition for high performance in cardiac care. Qualifying physicians
are eligible to receive up to $160 for each cardiac patient covered by a partic
ipating employer and plan. Both Diabetes Care Link and Cardiac Care Link offer a
suite of products and tools to engage cardiac patients in their care, achieve b
etter outcomes, and identify local physicians who meet high performance standard
s.
Integrated Healthcare Association The Integrated Healthcare Association (IHA) is
a statewide leadership group of California health plans, physician groups, and
healthcare systems, plus academic, consumer, purchaser, and pharmaceutical repre
sentatives (IHA 2006). Six participating health plans (representing more than 8
million enrollees) have agreed to participate: Aetna, Blue Cross of California,
Blue Shield of California, CIGNA HealthCare of California, Health Net, and Pacif
iCare. Each plan will use results on common performance measures while still des
igning its own physician group bonus program. IHA estimates that $100 million or
more could be paid out to physician groups in the first year, based on performa
nce results. A single scorecard incorporating patient satisfaction and clinical
measures will be developed for each physician group and made public. Future plan
s include expanded measures and increased funding. Premier Hospital Demonstratio
n A good example of hospital P4P is the Premier Hospital Demonstration. This CMS
-sponsored demonstration includes 268 hospitals that are part of the Premier all
iance of nonprofit hospitals (Kahn et al. 2006). This P4P project is unique in t
hat hospitals that participate can both gain and lose compensation based on thei
r performance in the areas of care for heart attacks, heart failure, pneumonia,
coronary artery bypass, and hip and knee replacements. This demonstration may he
lp inform any changes in Medicares payment system.
Tiering
Tiering is another method being used to encourage the spread of EBM. Tiered syst
ems present health plan members with a number of network or
63
medical clinic options that have variable premiums or copays. The health plan cr
eates the tiers based on quality and cost. In some cases, tiering is solely base
d on cost, and health plan members are directed to other public sources for qual
ity information. (See the discussion of public reporting earlier in this chapter
.)
Patient Choice A relatively mature example of tiering is Medicas Patient Choice (
Lee 2002). This program was developed in Minnesota and has been replicated in ot
her areas of the United States. In 1997, the Buyers Healthcare Action Group, an
organization of Minnesota business leaders, was frustrated by escalating healthc
are prices and the lack of focus on quality and launched Choice Plus. This first
tiered healthcare delivery network in the country became Medica Health Plans Pat
ient Choice program and is the longest-running, most wellestablished plan of its
kind. Patient Choice first divides the total universe of primary care physician
s into care systems, the rule being that each primary care physician must be in
only one care system. The care system then contracts with specialists, hospitals
, and other caregivers and submits a price bid to Patient Choice. Patient Choice
combines the price bid and the care system quality scores into an aggregate val
ue score. The care systems with the highest scores are assigned a value ranging
from $ to $$$. If a health plan member chooses to receive care in the $ system, she wi
ll pay the lowest premium. She will pay the highest premium if she chooses a $$$ s
ystem. Patient Choice believes this system achieves three goals (Medica 2006):
1. Allow providers to be accountable to their patients for their care delivery r
esults and recognize and reward those who deliver quality care; 2. Give employer
s the ability to use their purchasing dollars to obtain greater value without li
miting consumer choice; and 3. Educate consumers to be better managers of their
care and empower them to make value-based choices.
Minnesota Advantage The Minnesota Advantage Health Plan (2006) for state of Minn
esota employees builds on the success of Patient Choice. Its cost-sharing featur
es are expected to help the state better control healthcare costs while maintain
ing flexibility in access to doctors and clinics. Under Minnesota Advantage, sta
te employees share in the cost of specific medical services they obtain by payin
g varying levels of out-of-pocket costs (deductibles, office visit copayments, a
nd coinsurance). The state has placed providers into one of four cost levels dep
ending on the care system in which the provider participates and that care syste
ms total
64
cost of delivering healthcare. A care system includes primary care providers, sp
ecialty providers, and hospitals. Although the specialty providers and hospitals
may be in a number of care systems, the primary care clinics must be unique, as
in Patient Choice. The tiering system is based on the primary care clinics care
system. The amount of cost sharing paid when using healthcare services varies de
pending on the cost level of the provider chosen by the employee. The most impor
tant aspect of this tiering system is its annual firstdollar deductible (Table 3
.2). Minnesota employees have responded to this annual first-dollar deductible i
ncentive by moving their choice of providers to those in the lower-cost tiers. F
igure 3.2 shows the distribution of clinic tiers in 2004 and 2006. Providers hav
e responded to this market pressure by increasing their use of generic drugs, ex
tending office hours to minimize emergency department visits, and working with t
heir primary hospitals to reduce hospitalization costs. These operations improve
ments have moved many clinics into the lower-cost tiers. Tiering is a powerful m
otivator for providers, as it shifts market share. This is in contrast to the we
aker effects of public reporting and additional P4P revenue. A provider who find
s herself in a low-quality, high-cost tier can use the tools contained in this b
ook to identify the causes and charter projects to make the needed improvements
in cost and quality.
Vincent Valley Hospital and Health System and P4P
The leadership of Vincent Valley Hospital and Health System (VVH) was encouraged
by its board to participate in BTE (2006), an employer-sponsored program that p
rovides participating medical groups with up to $80 a year per patient with diab
etes if they are top performers in diabetes care. The specific goals VVH chose t
o target are those identified by the Institute for Clinical Systems Improvement
(2005): HbA1c of less than 7 percent LDL lower than 100 mg/dL
TABLE 3.2 Minnesota Advantage Health Plan Annual First-Dollar Deductible, 2006
Tier 1 2 3 4
Individual $30 $100 $280 $500
Family $60 $200 $560 $1,000
SOURCE: Minnesota Advantage Health Plan (2006).
65
Tier 4
Tier 3
2006 Tier 1 163 Tier 2 1,062 Tier 3 588 Tier 4 235
2004 304 568 410 339
FIGURE 3.2 Number of Primary Care Clinics in Each Payment Tier for Minnesota Adv
antage, 2004 and 2006
Tier 2
Tier 1
2004 2006 0 200 400 600 800 1,000 1,200
NOTE: A chi-square test indicates significant differences in tier membership fro
m 2004 to 2006. Membership in Tier 1 decreased significantly, while membership i
n Tier 2 increased significantly. Although membership in Tier 3 increased signif
icantly, the difference is not of practical significance. Membership in Tier 4 d
ecreased significantly. In summary, very low cost and very high cost providers m
oved to the middle, or Tier 2, level. SOURCE: Minnesota Advantage Health Plan (2
006).
Blood pressure lower than 130/80 Nonsmoking status Aspirin for patients over age
40 VVH also chose to use the Institute for Clinical Systems Improvement care gu
idelines as the basis for implementation (see companion web site). The operation
s management tools and approaches detailed in the remainder of this book were us
ed to achieve these goals. Chapter 14 describes how VVH acccomplished this.
Using EBM and Public Reporting to Advance Health Policy: A Proposal
Public policymakers face the urgent challenge of restraining healthcare costs an
d meeting the Institute of Medicines six quality aims. Yet there has been strong
public support for the policy direction of letting the marketplace drive U.S. he
alth policy aims. This implies that top-down, centralized, government-run health
care is unlikely in the United States. However, the new policy tools discussed i
n this chapter could be combined into a solution that involves minimal governmen
tal intrusion into the delivery of healthcare.
66
The government, at the state level, has a powerful tool that could be used in co
mbination with the growing tools of EBM and public reporting of qualitythe certif
icate of need (CON). In many states, agencies review and grant providers a CON f
or major capital expenditures for buildings or equipment. The provider would vio
late state law if it made the expenditure without a CON. However, because the ru
les governing the granting of CONs are complex and ineffective, CONs are not con
sidered an adequate policy tool today. CONs could become a powerful tool to cont
rol costs and increase quality if they were used in the following manner. First,
all providers would be divided into groups (e.g., primary care doctors, special
ty doctors, hospitals). Next, each provider would be ranked on costs using Medic
are data as aggregated in the various Medicare classification systems (e.g., dia
gnosisrelated groups, ambulatory care groups). Each would also be ranked by aggr
egate quality measures as publicly reported by CMS and other reporting agencies.
The providers would then receive an aggregate ranking based 50 percent on quali
ty and 50 percent on cost. A major goal of health policy reformers has been to r
educe the inflation rate of healthcare services to match the general inflation r
ate. Therefore, the state could model each provider based on the cost increases
and potential savings achieved from their quality performance. States would then
separate providers into those that met their policy goal of low inflation and t
hose that did not. When providers wished to expand, only those meeting the polic
y goals would be granted a CON. This fairly simple system would face resistance
by providers, particularly those with poor performance. It could also harm under
served communities if their sole providers scored poorly in this system. However
, this system would have a number of advantages. By using publicly available dat
a, it would be transparent and readily understood by providers, patients, and po
licymakers. By denying low-performing organizations a CON, it would provide powe
rful incentives for organizations to improve or lose market share. It would also
align rewards in the system with provider values, as all providers want to deli
ver high-quality, cost-effective care. Finally, it would require a minimal burea
ucratic role for government, making it the simple scorekeeper and granter or den
ier of CONs.
Conclusion
It is unclear how strongly the broad spread of EBM and the new policy tools will
influence the delivery of healthcare in the United States. However, it is clear
that providers with high-performing, cost-effective operations will be most lik
ely to succeed.
67
1. What are other examples of a care delivery setting with a mix of standard and
custom care? 2. Select three PQIs from Table 3.1 and go the National Guideline
Clearinghouse to find guidelines that would minimize hospital admissions for the
se conditions. What would be the challenges in implementing each of these guidel
ines? 3. Review the four public reporting systems listed earlier in the chapter.
How well do they meet the five goals of Shaller et al. for public reporting? 4.
What are three strategies to maximize P4P revenue? 5. What projects would you i
nitiate if you were in a primary care clinic that was classified as Tier 4 in th
e Minnesota Advantage program? 6. Assume you are a lobbyist for the CON policy p
roposal outlined at the end of the chapter. What would be your arguments to a sk
eptical state legislature worried about government-run healthcare?
References
Agency for Healthcare Research and Quality, U.S. Department of Health and Human
Services. 2006. Guide to Prevention Quality Indicators: Hospital Admission for A
mbulatory Care Sensitive Conditions. [Online information; retrieved 8/30/06.] ww
w.qualityindicators.ahrq.gov/downloads/pqi/ pqi_guide_v30a.pdf. Bohmer, R. M. J.
2005. Medicines Service Challenge: Blending Custom and Standard Care. Health Care
Management Review 30 (4): 32230. Bridges to Excellence. 2006. [Online information
; retrieved 8/29/06.] www. bridgestoexcellence.org. Casalino, L., R. Gillies, S.
Shortell, J. A. Schmittdiel, T. Bodenheimer, J. C. Robinson, T. Rundall, N. Osw
ald, H. Schauffler, and M. C. Wang. 2003. External Incentives, Information Techno
logy, and Organized Processes to Improve Health Care Quality for Patients with C
hronic Diseases. Journal of the American Medical Association 289 (4): 43441. Centr
e for Evidence-Based Medicine. 2006. [Online information; retrieved 8/30/06.] ww
w.cebm.net. Galvin, R. 2006. Pay-for-Performance: Too Much of a Good Thing? A Con
versation with Martin Roland. Health Affairs Web Exclusives 25: pw41219. Gildemeis
ter, S. 2006. Hospitalizations for Ambulatory Care Sensitive ConditionsIndicator o
f Opportunities to Improve the Quality of Outpatient Care? Presented at the Minne
sota Health Services Research Conference, Minneapolis, MN, March 7.
68
Gosfield, A. G. 2004. P4P Contracting: Bold Leaps or Baby Steps? Patient Safety an
d Quality Oct./Dec. [Online information; retrieved 8/29/06.] www.psqh.com/ octde
c04/gosfield.html. Hibbard, J. H., J. Stockard, and M. Tusler. 2003. Does Publici
zing Hospital Performance Stimulate Quality Improvement Efforts? Health Affairs 2
2 (2): 8494. Institute for Clinical Systems Improvement. 2005. Diabetes Mellitus,
Type 2; Management of. [Online information; retrieved 9/1/06.] hwww.icsi.org/ kno
wledge/detail.asp?catID=29&itemID=182. Institute for Healthcare Improvement. 200
6. 100,000 Lives Campaign. [Online information; retrieved 9/1/06.] www.ihi.org/IHI
/Programs/Campaign/. Integrated Healthcare Association. 2006. [Online informatio
n; retrieved 8/30/06.] www.iha.org. Kahn, C. N., T. Ault, H. Isenstein, L. Potet
z, and S. Van Gelder. 2006. Snapshot of Hospital Quality Reporting and Pay-for-Pe
rformance Under Medicare. Health Affairs 25 (1): 14862. Kertesz, L. 2005. Bringing
Evidence-Based Medicine into the Physicians Office. Coverage Nov./Dec. [Online inf
ormation; retrieved 9/1/06.] www.ahip.org/content/default.aspx?bc=31|130|136|139
58|13962. Lee, K. 2002. Tiered Provider Networks Bring Promise of Higher Quality
Care. Employee Benefit News 16 (11): 4142. Lenfant, C. 2003. Clinical Research to C
linical PracticeLost in Translation? New England Journal of Medicine 349 (9): 86874
. Medica. 2006. Patient Choice. [Online information; retrieved 8/30/06.] www.pchea
lthcare.com/. Milgate, K., and S. B. Cheng. 2006. Pay-for-Performance: The MedPAC
Perspective. Health Affairs 25 (2): 41319. Minnesota Advantage Health Plan. 2006.
[Online information; retrieved 8/30/06.] www.doer.state.mn.us/ei-segip/health.h
tm. National Guideline Clearinghouse. 2006. [Online information; retrieved 8/28/
06.] www.guideline.gov/. Roland, M. 2004. Linking Physicians Pay for Quality of Ca
reA Major Experiment in the United Kingdom. New England Journal of Medicine 351 (1
4): 144854. Rosenthal, M. B., R. Fernandopulle, H. R. Song, and B. Landon. 2004. P
aying for Quality: Providers Incentives for Quality Improvement. Health Affairs 23
(2): 12741. Shaller, D., S. Sofaer, S. D. Findlay, J. H. Hibbard, D. Lansky, and
S. Delbanco. 2003. Consumers and Quality-Driven Health Care: A Call to Action. He
alth Affairs 22 (2): 95101. Timmermans, S., and A. Mauck. 2005. The Promises and P
itfalls of Evidence-Based Medicine. Health Affairs 24 (1): 1828.
PART
II
setting goals and executing strategy
CHAPTER
4
STRATEGY AND THE BALANCED SCORECARD
CHAPTER OUTLINE
Operations Management in Action Overview Moving Strategy to Execution The Challe
nge of Execution Why Do Todays Management Tools Fail? Robert Kaplan and David Nor
ton The Balanced Scorecard in Healthcare The Balanced Scorecard as Part of a Str
ategic Management System Elements of the Balanced Scorecard System Mission and V
ision Financial Perspective Revenue Growth Cost ReductionProductivity Asset Utili
zation and Investment Strategy Risk Management Through Diversity Customer Perspe
ctive and Market Segmentation Customer measures Customers: The Value Proposition
Vincent Valley Hospital and Health Systems value proposition Internal Business P
rocess Perspective Innovation Ongoing Process Improvement 70 Post-sale Service V
incent Valley Hospital and Health System Internal Business Processes Learning an
d Growing Perspective Employee Skills and Abilities Necessary IT Employee Motiva
tion Vincent Valley Hospital and Health System Learns and Grows Linking Balanced
Scorecard Measures to Strategy Outcomes and Performance Drivers Strategy Maps V
incent Valley Hospital and Health System Strategy Maps Implementation of the Bal
anced Scorecard Linking and Communicating Targets, Resources, Initiatives, and B
udgets Displaying Results Ensuring That the Balanced Scorecard Works Modificatio
ns of the Classic Balanced Scorecard Implementation Issues
Conclusion Discussion Questions Chapter Exercise

Case Study: Saint Marys Duluth Clinic References Further Reading
Balanced Scorecard Collaborative cause-and-effect statement diagnosis-related gr
oup (DRG) full-time equivalent (FTE) employee if-then statement lagging indicato
r leading indicator performance driver perspectives quality function deployment
(QFD) relative value unit (RVU) strategy map value proposition
71
Setting Goals and E xecuting Strategy

Falls Memorial Hospital in International Falls was down to five days of cash on
hand when Mary Klimp became chief executive officer (CEO) in late 1998. On the C
anadian border in northern Minnesota, the hospital is 100 miles from any seconda
ry center and 165 miles from a tertiary care center in Duluth. The average daily
census was 3.5, and the profit margin was 3 percent. The leadership created a st
rategic plan in 1999 and began using the balanced scorecard in 2000. Without out
side consulting assistance, they put their balanced scorecard into a spreadsheet
and used it with the board, administration, medical staff, department managers,
and all staff. The scorecard was also part of the orientation of every new empl
oyee and was posted throughout the hospital. The implementation process took nin
e months, but the CEO estimates that ongoing maintenance takes no more than seve
n hours a month. Currently, Falls Memorial measures 16 different performance ind
icators on its scorecard and has made a remarkable turnaround. As of late 2003,
the hospital had nearly 100 days of cash on hand, had a profit margin of 15 perc
ent, and had undertaken extensive capital expansion. The inpatient census had mo
re than tripled to 11. Leadership attributed the improvement primarily to the ba
lanced scorecard. Mary Klimp says, My proudest balanced scorecard moment was when
I met a housekeeper in the hall who remarked that the balanced scorecard indica
ted that the hospital was doing great and that she was proud to work here. Klimp
believes that this widespread staff understanding of strategy has been an import
ant marketing tool for the hospital.
SOURCE: Mountain States Group 2003.
Overview
Most healthcare organizations have good strategic plans; what frequently fails i
s their execution. This chapter demonstrates how the balanced scorecard can be u
sed as an effective tool to consistently move strategy to execution. First, trad
itional management systems are examined and their failures explored. Next, the t
heory of the balanced scorecard and strategy mapping is reviewed and its applica
tion to healthcare organizations explained. Practical steps to implement and mai
ntain a balanced scorecard system are provided. Detailed examples from Vincent V
alley Hospital and Health System (VVH) will demonstrate the application of these
tools. The companion web site contains templates and explanatory videos that ca
n be used for student exercises or to implement a balanced scorecard in an exist
ing healthcare organization. 72
C h a p t e r 4 : St ra t e g y a n d t h e B a l a n c e d S c o re c a rd
73
This chapter gives readers a basic understanding of balanced scorecards to enabl
e them to: Explain how a balanced scorecard can be used to move strategy to acti
on. Explain how to monitor strategy from the four perspectives. Identify key ini
tiatives to achieve a strategic objective. Develop a strategy map linking initia
tives. Identify and measure leading and lagging indicators for each initiative.
Use Microsoft PowerPoint and Excel to create strategy maps and scorecards.
Moving Strategy to Execution
The Challenge of Execution
Environmental causes commonly cited for the failure of execution in healthcare o
rganizations include intense financial pressures, complex operating structures,
and cultures with multistakeholder leadership that resists change. New and redef
ined relationships of healthcare providers, particularly physicians, are being a
ccompanied by a rapid growth of medical treatment knowledge and technology. Incr
eased public scrutiny of how healthcare is delivered is leading to the associate
d rise of consumer-directed healthcare. No matter how significant these external
factors, most organizations flounder on internal factors. Mankins and Steele (2
005) studied strategy execution failures in all types of enterprises and identif
ied nine key factors that contribute to poor performance: 1. 2. 3. 4. 5. 6. 7. 8
. 9. Inadequate or unavailable resources; Poorly communicated strategy; Poorly d
efined actions required to execute the strategy; Unclear accountability for exec
ution; Organizational silos; A culture that blocks execution; Inadequate perform
ance monitoring; Inadequate consequences for failure or success; and Poor, uncom
mitted senior leadership.
These factors also plague healthcare organizations. To gain competitive advantag
e from its operations, an organization needs an effective system to move its str
ategies forward. The management systems of the past are poor tools for todays cha
llenging environment. The day-to-day world of a current healthcare leader is int
ense (Figure 4.1). Because of ever-present communications technologies (beepers,
cell phones,
74
FIGURE 4.1 The Complex World of Todays Healthcare Leader
Last years initiative
Urgent operating problems
This years new initiative
Whats on your desk today?
Employee turnover recruiting Financial performance pressure
40 e-mails 10 voice mails
personal digital assistants, Internet, wi-fi), managers float in a sea of inputs
and daily challenges. Healthcare delivery environments are susceptible to focus
ing on urgent issues rather than addressing important challenges of strategy exe
cution. And although organizations can develop effective project managers (as di
scussed in Chapter 5), they will fail to compete successfully if they do not pla
ce these projects in a broader system of strategy implementation. Fortunately, t
he balanced scorecard provides a framework and sophisticated mechanisms to move
from strategy to execution.
Why Do Todays Management Tools Fail?
Historically, most organizations have been managed with three primary tools: str
ategic plans, financial reports, and operational reports. Figure 4.2 shows the r
elationships among these tools. In this traditional system, the first step is to
create a strategic plan, which is usually updated annually. Next, a budget and
operations or project plan is created. The operational plan is sometimes referre
d to as the tactical plan; it provides a more detailed level of task description
s with timelines and expected outcomes. The organizations performance is monitore
d by senior management through the financial and operational reports. Finally, i
f deviations from expected performance are encountered, managers take corrective
action. Although theoretically easy to grasp, this management system frequently
fails for a number of common reasons. Organizations are awash in operating data
, and there is no effort to identify key metrics. The strategic plans, financial
reports, and operational reports are all created by different departments, and
each report is reviewed in different time frames, often by different managers. F
inally, none of the reports connect with one another. These are the root causes
of poor execution. If strategies are not linked to actionable items, operations wi
ll not change, nor will the financial
75
Operating statistics Strategic plan Operations Financial results Management cont
rol
FIGURE 4.2 The Traditional Theory of Management
results. In addition, strategic plans are frequently not linked to departmental
or individual goals and, therefore, reside only on the shelf in the executive su
ite. Many strategic plans contain a logic hole, meaning they lack an explanation
of how accomplishing a strategic objective will provide a specific financial or
operational outcome. For example: Strategic objective: increase the use of evid
ence-based medicine (EBM) Expected outcome: increased patient satisfaction Altho
ugh this proposition may seem correct on the surface, the logic to connect the u
se of EBM to patient satisfaction is unclear. In fact, patient satisfaction may
decrease if providers consistently meet EBM guidelines by counseling patients on
personal lifestyle issues (e.g., Will you stop smoking?or You need to lose weight).
Frequently, the time frame of strategy execution is also problematic. Financial
reports tend to be timely and accurate but only reflect the current reporting p
eriod. Unfortunately, the review of these reports does not encourage the long-te
rm strategic allocation of resources (e.g., a major capital expenditure) that ma
y require multiple-year investments. A good currentmonth financial outcome is pr
obably due to an action that occurred many months in the past. The cumulative re
sult of these problems is poor execution, leading to poor outcomes.
Robert Kaplan and David Norton
In the early 1990s, Kaplan and Norton undertook a study to examine how companies
measure their performance. The growing sophistication of company-wide informati
on systems was beginning to provide senior management with executive information
systems, which could provide sophisticated displays and dashboards of company per
formance. The original purpose of Kaplan and Nortons study was to understand and
document this trend.
76
However, their study uncovered a number of reporting practices that many leading
companies were using to measure their performance. These firms looked at their
operations from a number of perspectives that, in total, provided a balanced scor
ecard. The essential elements of this work were first reported in The Balanced Sc
orecard: Translating Strategy into Action (Kaplan and Norton 1996) and The Strat
egy-Focused Organization (Kaplan and Norton 2001). The key element of the balanc
ed scorecard is, of course, balance. An organization can be viewed from many per
spectives, but Kaplan and Norton identified four common perspectives from which
an organization must examine its operations (Figure 4.3): 1. 2. 3. 4. Financial;
Customer; Internal process and innovation; and Employee learning and growth.
As an organization is viewed from each perspective, different measures of perfor
mance are important. Every perspective in a complete balanced scorecard contains
a set of objectives, measures, targets, and actions. Each measure in each persp
ective must be linked to the organizations overall strategy. The indicators of pe
rformance in each of the four perspectives must be both leading (predicting the
future) and lagging (reporting on performance today). Indicators must also be ob
tained from inside the organization and from the external environment. Although
many think of the balanced scorecard as a reporting technique, its true power co
mes from its ability to link strategy to action. Balanced scorecard practitioner
s develop strategy maps that link projects and actions to outcomes in a series o
f maps. These maps display the theory of the company and can be evaluated and fine
-tuned with many of the quantiFIGURE 4.3 The Four Perspectives in the Balanced S
corecard
Financial stakeholders
Operations and strategic plan
Customers
Operations
Employees
77
tative techniques described in Chapter 6. Kaplan and Norton (2001) have expanded
their system of constructing strategy maps in The Strategy-Focused Organization
.
The Balanced Scorecard in Healthcare
The balanced scorecard and its variations have been adopted by leading healthcar
e organizations, many of which are listed by the Balanced Scorecard Collaborativ
e at www.bscol.com. The Collaborative is dedicated to the worldwide awareness, u
se, enhancement, and integrity of the balanced scorecard as a value-added manage
ment process. It has given its Hall of Fame award to two healthcare organization
s: Duke Childrens Hospital (Durham, NC) and Saint Marys Duluth Clinic (Duluth, MN)
. Inamdar and Kaplan (2002) reviewed the use of the balanced scorecard in health
care and concluded that many healthcare organizations are effectively using scor
ecards to improve their competitive marketing position, financial results, and c
ustomer satisfaction.
The Balanced Scorecard as Part of a Strategic Management System
Although it does not substitute for a complete strategic management system, the
balanced scorecard is a key element in such a system and provides an effective t
ool to move an organizations strategy and vision into action. The development of
a balanced scorecard leads to the clarification of strategy and communicates and
links strategic measures throughout an organization. Organizational leaders can
plan projects, set targets, and align strategic initiatives during the creation
of the balanced scorecard. If used properly, the balanced scorecard can also en
hance strategic feedback and learning.
Elements of the Balanced Scorecard System
A complete balanced scorecard system will have the following elements, explained
in detail below: Organizational mission, vision, and strategy Perspectives - Fi
nancial - Customer - Internal business process - Learning and growing Strategy m
aps Strategic alignmenttop to bottom Processes for identifying targets, resources
, initiatives, and budgets Feedback and the strategic learning process
78
Mission and Vision
The balanced scorecard system presupposes that an organization has an effective
mission, vision, and strategy in place. For example, the mission of VVH is to pro
vide high-quality, cost-effective healthcare to our community. Its vision is: With
in five years, we will be financially sound and will be considered the place to
receive high-quality care by the majority of the residents of our community. To a
ccomplish this vision, VVH has identified six specific strategies: 1. 2. 3. 4. 5
. Recruit five new physicians; Expand the outpatient imaging center; Revise the
VVH web site to allow self-scheduling by patients; Increase the volume of obstet
ric care; Renegotiate health plan contracts to include performance incentives; a
nd 6. Improve emergency department operations and patient satisfaction. Box 4.1
lists many common healthcare organizational strategies.
BOX 4.1 Common Healthcare Organizational Strategies
In a study on the use of the balanced scorecard in healthcare delivery organizat

ions, Inamdar and Kaplan (2002) identified the following strategies that connect
to each organizations scorecard: Achieve financial strength. Develop reputation
or brand image. Grow patient volume through market analysis and customer focus.
Achieve operational excellence and demonstrate value through improved measuremen
t systems. Form strategic alliances and partnerships, especially with physicians
. Develop infrastructure to offer and integrate across the continuum of care thr
ough enhancing information technology capabilities.
Zelman, Pink, and Matthias (2003) found that rural hospitals were using balanced
scorecards to support the following strategies: Modernize facilities Recruit
d retain clinical staff Improve cash flow Recruit and retain physicians Improve
inpatient census Upgrade financial and information systems Improve Medicaid reim
bursement Improve the collection process
79
The VVH example is used throughout this chapter to demonstrate the use of the ba
lanced scorecard. The two strategies that will be examined in depth are increasi
ng the volume of obstetric care and improving emergency department operations an
d patient satisfaction. With an effective strategic plan in place, the next step
is to begin evaluating its implementation as viewed from each of the four persp
ectives (financial, customer, operational, and learning and growing). Placing a
perspective at the top of a balanced scorecard strategy map means that results i
n this perspective contain the final outcomes desired by an organization. In mos
t organizations, the financial view is the top-most perspective. Therefore, the
initiatives undertaken in the other three perspectives should result in positive
financial performance for the organization. No margin, no mission is still a true
statement for nonprofit healthcare organizations. Nonprofit healthcare institut
ions need operating margins to provide financial stability and capital. However,
some organizations prefer to position the customer (patient) as the top perspec
tive. In that case, the initiatives undertaken in the other three perspectives w
ill result in positive patient outcomes. (Modifications to the classic balanced
scorecard are discussed at the end of this chapter.)
Financial Perspective
Although the other three perspectives and their associated areas of activity sho
uld lead to outstanding financial performance, there are initiatives that can be
undertaken within the financial perspective by themselves. Although the focus o
f this book is not directly on healthcare finance, some general strategies shoul
d always be under consideration by an organization. If an organization is in a g
rowth mode, the focus should be on increasing revenue to accommodate this growth
. If it is operating in a relatively stable environment, the organization may ch
oose to emphasize profitability. If the organization is both stable and profitab
le, the focus can shift to investmentin both physical assets and human capital. A
nother major strategy in the financial domain is the diversification of both rev
enues and expenditures to minimize financial risk.
Revenue growth An organization that is in a growth mode should be engaged in dev
eloping and deploying new products. Imaging centers are a current example of the
rapid introduction of ever more effective technologies. Another growth strategy
is to find new uses for an existing technology. For example, VVH is planning to
use its existing web site to allow patients to schedule their own appointments,
expecting that this will increase ambulatory care revenue. A growing organizati
on will seek new customers and markets. An example is placing primary care clini
cs where the population is experiencing rapid growth. Developing partnerships is
another growth strategy. The newest growth strategy in healthcare is consumer-d
irected pricing. If an organization
80
can price a consumer-sensitive service such as magnetic resonance imaging scans
correctly, it will increase both volume and revenue. Public reporting of costs a
nd quality were explored in depth in Chapter 3.
Cost reductionproductivity In most cases, cost reductions will improve financial
performance. Chapters 6 through 13 provide tools for improving processes and red
ucing costs. However, other important strategies should also be evaluated. The f
irst is to redefine the channel mix. In other words, deliver services in another
mode. The use of e-mail and group visits are two examples redefining channel mi
x in primary care. The advanced use of new communication and information technol
ogies (ITs) will allow creative healthcare organizations to rethink the channels
they use to deliver services. A second, frequently overlooked, cost management
strategy is to reduce overhead costs. All nonoperating costs should be scrutiniz
ed carefully to ensure that they are contributing to the achievement of organiza
tional outcomes. Asset utilization and investment strategy Balance sheet managem
ent is also part of the financial perspective. Managing working capital assets s
uch as accounts payable, inventory, and accounts receivable is part of this pers
pective. One of the most complex processes in healthcare is that of sending and
collecting a bill. These systems are amenable to the process improvement tools i
n Chapters 6 through 11. Another balance sheet strategy is to improve asset allo
cations. Does one invest more in buildings or equipment? Can an IT system be sha
red by more than one department? The analysis of these questions is well documen
ted in Healthcare Finance: An Introduction to Accounting and Financial Managemen
t (Gapenski 2005); the initiatives and projects that surround such tradeoffs sho
uld be part of an organizations balanced scorecard. Risk management through diver
sity The final key financial strategy is to minimize risk by increasing diversit
y. The more diverse an organizations revenue sources, the less likely it is that
a significant change in any one source will have a major impact on the organizat
ion. Diversification of payers is difficult to achieve but should always be atte
mpted. Product lines can be expanded in both clinical areas (e.g., emergency, bi
rthing center, internal medicine) and delivery models (e.g., inpatient, ambulato
ry, stand alone). Geographic expansion is a classic diversification strategy. Ho
wever, the greatest opportunity for diversification today may be innovative reta
il strategies to attract market share from the newly empowered consumer. Box 4.2
lists many common metrics used to measure performance from the financial perspe
ctive.
81

Percent of budgetrevenue Percent of budgetexpense Days in accounts receivable Days
of cash on hand Collection rate Return on assets Expense per relative value uni
t (RVU) Cost per surgical case Case-mix index Payer mix Growth, revenue, expense
, and profitproduct line Growth, revenue, expense, and profitdepartment Growth, re
venue, and cost per adjusted patient day Growth, revenue, and cost per physician
full-time equivalent (FTE) Price competitiveness on selected services Research
grant revenue
BOX 4.2 Metrics to Measure Performance from the Financial Perspective
Customer Perspective and Market Segmentation
The second perspective is to view an organizations operations from the customers p
oint of view. In most healthcare operations, the customer is the patient. Integr
ated health organizations, however, may have insurance or health plan delivery v
ehicles; their customers are then employers or the government. Many hospitals an
d clinics also consider their community, in total, as the customer. The physicia
n could also be seen as the customer in many hospital organizations. Once the ge
neral customers are identified, it is helpful to segment them into smaller group
s and determine the value proposition that will be delivered to each customer segm
ent. Example market segments are patients with chronic illnesses (e.g., diabetes
, congestive heart failure), obstetric care, sports medicine, cancer care, emerg
ency care, Medicaid patients, small employers, and referring primary care physic
ians.
Customer measures Once market segments have been determined, a number of traditi
onal measures of marketplace performance may be applied, the most prominent bein
g market share. Customers should be individually tracked and measured in terms o
f both retention and acquisition, as it is always easier to retain an existing c
ustomer than to attract a new one. Customer satisfaction and profitability are a
lso useful measures. Box 4.3 displays a number of common customer metrics.
82
BOX 4.3 Metrics to Measure Performance from the Customer Perspective

Patient care volumes - By service, type, and physician - Turnovernew patients and
those exiting the system Physician - Referral and admission rates - Satisfactio
n - Availability of resources (e.g., operating suite time) Market share by produ
ct line Clinical measures - Readmission rates - Complication rates - Compliance
with evidence-based guidelines - Medical errors Customer service - Patient satis
faction - Waiting time - Cleanlinessambience - Ease of navigation - Parking - Bil
ling complaints Reputation Price comparisons relative to competitors
Customers: The value proposition Organizations create value to retain old custom
ers and attract new ones. Each market segment may require products to have diffe
rent attributes to maximize the particular value proposition. For example, it ma
y be important to be a price leader for outpatient imaging, as many patients wil
l be paying for this service via a healthcare savings account. For another segme
ntemergency services, for examplespeed of delivery may be critical. The personal r
elationship of provider to patient may be important in primary care but not as i
mportant in anesthesiology. Image and reputation are particularly strong influen
ces in consumer behavior and can be a competitive advantage for specialty health
care services. Careful understanding of the value proposition in an organization
will lead to effective metrics and strategy maps in the balanced scorecard syst
em. Chapter 13 offers a number of examples of how the value proposition operates
in competitive tiered systems. Vincent Valley Hospital and Health Systems value
proposition VVH has developed a value proposition for its obstetric services. It
s market segment is pregnant women, aged 18 to 35. VVH believes the product attrib
utes for this market should be:
83
quick access to care; warm and welcoming facilities; customer relations characte
rized by strong and personal relationships with nurses, midwives, and doctors; a
nd an image of high-quality care that will by supported by an excellent system f
or referrals and air transport for high-risk deliveries. VVH determined that the
following metrics would be used to measure each of these attributes: the time f
rom arrival to care in the obstetric suite; a patient survey of facility attribu
tes; a patient survey of satisfaction with staff; and percent of high-risk newbo
rns referred and transported, and the clinical outcomes of these patients.
The main value proposition for emergency care was identified as reduced waiting
time. Following internal studies, competitive benchmarking, and patient focus gr
oups, it was determined that VVHs goal would be to have fewer than 10 percent of
its emergency department patients wait more than 30 minutes for care.
Internal Business Process Perspective
The third perspective in the balanced scorecard is that of internal business pro
cesses or operationsthe primary focus of this book. According to Kaplan and Norto
n (1996), the internal business process perspective has three major components:
innovation, ongoing process improvement, and post-sale service.
Innovation A well-functioning healthcare organization will have a purposeful inn
ovation process. Unfortunately, many health organizations today can only be char
acterized as reactionary. They respond to new reimbursement rules, government ma
ndates, or technologies introduced through the medical staff. Bringing thoughtfu
l innovation into the life cycle is one of the most pressing challenges contempo
rary organizations face. The first step in an organized innovation process is to
identify a potential market segment. Then, two primary questions need to be ans
wered: (1) What benefits will customers value in tomorrows market? and (2) How ca
n the organization innovate to deliver these benefits? Once these questions have
been researched and answered, product creation can commence. Quality function d
eployment (Chapter 8) can be a useful tool for new product or service developmen
t. If a new service is on the clinical leading edge, it may require additional r
esearch and testing. A more mainstream service could
84
require competitor research and review of the clinical literature. The principle
s of project management (Chapter 5) should be used throughout this process until
the new service is operational and stable. Standard innovation measures used in
many industries outside healthcare include percent of sales from new products,
percent of sales from proprietary products, new product introductions per year,
time to develop new products, and time to break even. Healthcare operations tend
toward stability (bordering on being rigid), and therefore, a major challenge i
s simply ensuring that all clinical staff use the latest and most effective diag
nostic and treatment methodologies. However, in the coming era of consumer-direc
ted healthcare, those organizations with a well-functioning new product developm
ent process will clearly have a competitive advantage.
Ongoing process improvement The case for process improvement and operations exce
llence is made throughout this book. The project management system (Chapter 5) a
nd process improvement tools (Chapters 6 through 11) are key to these activities
. The strategic effect of process improvement and maintaining gains is discussed
in Chapter 14. Post-sale service The final aspect of the operations perspective
is the post-sales area, an area poorly executed in most healthcare delivery org
anizations. Sadly, the most common post-sale contact with a patient may be an un
decipherable or incorrect bill. Good post-service systems provide patients with
follow-up information on the service they received. Patients with chronic diseas
es should be contacted periodically with reminders on diet, medication use, and
the need to schedule follow-up visits. An outstanding post-sale system also find
s opportunities for improvement in the service as well as possible innovations f
or the future. Open-ended survey questions, such as From your perspective, how co
uld our organization improve? or How else can we serve your healthcare needs? can p
oint to opportunities for improvement and innovation. Box 4.4 lists common metri
cs used to measure operational performance.
BOX 4.4 Metrics to Measure Performance from the Operational Perspective

Average length of staycase-mix adjusted FTE/adjusted patient day FTE/diagnosis-re
lated group (DRG) FTE/RVU FTE/clinic visit Waiting time inside clinical systems
Access time to appointments
85

Percent value-added time Utilization of resources (e.g., operating room, imaging
suite) Patients leaving emergency department without being seen Operating room
cancellations Admitting process performance Billing system performance Medicatio
n errors Nosocomial infections National Quality Forum (2002) never events
BOX 4.4 Metrics to Measure Performance from the Operational Perspective (continu
ed)
Vincent Valley Hospital and Health System internal business processes VVH decide
d to execute four major projects to move its birthing center and emergency depar
tment strategies forward. The birthing center projects were to remodel and redec
orate labor and delivery suites, contract with a regional health system for emer
gency transport of high-risk deliveries, and begin predelivery tours of labor an
d delivery facilities by nursing staff. The emergency department project was to
execute a Lean analysis and kaizen event to improve patient flow.
Learning and Growing Perspective
The final perspective from which to view an organization is learning and growing
. To effectively execute a strategy, employees must be motivated and have the ne
cessary tools to succeed. Therefore, a successful organization will make substan
tial investments in this aspect of their operations. Kaplan and Norton (1996) id
entified three critical aspects of learning and growing: employee skills and abi
lities, necessary IT, and employee motivation.
Employee skills and abilities Although employees in healthcare usually come to t
heir jobs with general training in their technical field, continuous updating of
skills is necessary. Some healthcare organizations are effective in ensuring th
at clinical skills are updated but neglect training in other vital processes (e.
g., purchasing systems, organization-wide strategies). A good measure of the att
ention paid to this area is the number of classes conducted by the organization
or an outside contractor. Another important measure is the breadth of employee o
ccupations attending these classes. For example, do all employees from doctors t
o housekeepers attend organization-wide training? Necessary IT Most healthcare w
orkers are knowledge workers. The more immediately and conveniently information is
available to them, the more effectively they
86
will be able to perform their jobs. Process redesign projects frequently use IT
as a resource for automation and information retrieval. Measures of automation i
nclude number of employees having easy access to IT systems, percentage of indiv
idual jobs that have an automation component, and speed of installation of new I
T capabilities.
Employee motivation A progressive culture and motivated employees are clearly co

mpetitive advantages; therefore, it is important to monitor these with some freq
uency. Measures of employee satisfaction include:
Involvement with decisions Recognition for doing a good job Access to suffi
information to do the job well Active encouragement of creativity and initiativ
e Support for staff-level functions Overall satisfaction with the organization T
urnover Absenteeism Training hours per employee
Data for many of these measures are typically collected with employee surveys. T
hese three aspects of learning and growingemployee skills, IT, and motivationall c
ontribute to employee satisfaction. A satisfied employee is also productive and
will remain with the organization. Employee satisfaction, productivity, and loya
lty make outstanding organizational performance possible.
Vincent Valley Hospital and Health System learns and grows VVH realized its empl
oyees needed new skills to successfully execute some of its projects, so it enga
ged training firms to provide classes for all staff (Table 4.1).
TABLE 4.1 VVH Improvement Projects and Associated Training
Project Begin predelivery tours of labor and delivery facilities by nursing staf
f Execute a Lean analysis and kaizen event to improve patient flow in the emerge
ncy department
Employees Involved Obstetric nursing and support staff Managers and key clinicia
ns in the emergency department
Training Customer service and sales Lean tools (Chapter 9)
87
Linking Balanced Scorecard Measures to Strategy
Once expected objectives and their related measures are determined for each pers
pective, the initiatives to meet these goals must be developed. An initiative ca
n be a simple action or a large project. However, it is important to logically l
ink each initiative to the desired outcome through a series of cause-and-effect
statements. These are usually constructed as if-then statements that tie each init
iative together and contribute to the outcome, for example: If the wait time in
the emergency department is decreased, then the patient will be more satisfied.
If an admitting process is improved through the use of automation, then the fina
l collection rate will improve. If an optically scanned wrist band is used in co
njunction with an electronic health record, then medication errors will decline.
Each initiative should have measures associated with it, and every measure sele
cted for a balanced scorecard should be an element in a chain of causeeffect rel
ationships that communicate the organizations strategy.
Outcomes and Performance Drivers
Selecting appropriate measures for each initiative is critical. There are two ba
sic types of indicators. Outcome indicators, familiar to most managers, are also
termed lagging indicators because they result from earlier actions. Outcome indic
ators tend to be generic instead of tightly focused. Healthcare operations examp
les include profitability, market share, and patient satisfaction. The other typ
e of indicator is a performance driver, or leading indicator. These indicators pre
dict the future and are specific to an initiative and the organizations strategy.
For example, a performance driver measure could be waiting time in the emergenc
y department. A drop in waiting time should predict an improvement in the outcom
e indicator, patient satisfaction. A common pitfall in developing indicators is
the use of measures associated with the improvement project rather than with the
process improvement. For example, the fact that a project to improve patient fl
ow in a department is 88 percent complete is not as good an indicator as a measu
re of the actual change in patient flow, a 12 percent reduction in waiting time.
Outcome measures are always preferred, but in some cases, they may be difficult
or impossible to obtain. Because the number of balanced scorecard measures shou
ld be limited, identifying measures that are indicators for a complex process is
sometimes useful. For example, a seemingly simple indicator such as time to next
appointment for patient scheduling will track many complex processes within an o
rganization.
88
Strategy Maps
As discussed, a set of initiatives should be linked together by if-then statemen
ts to achieve a desired outcome. Both outcome and performance driver indicators
should be determined for each initiative. These can be displayed graphically in
a strategy map, which may be most helpfully organized into the four perspectives
with learning and growing at the bottom and financial at the top. A general str
ategy map for any organization would include: If employees have skills, tools, a
nd motivation, then they will improve operations. If operations are improved and
marketing is also improved, then customers will buy more products and services.
If customers buy more products and services and operations are run efficiently,
then the organizations financial performance will improve. Figure 4.4 shows a st
rategy map in which these general initiatives are indicated. The strategy map is
enhanced if each initiative also contains the strategic objective, measure used
, and target results that the organization hopes to achieve. Each causal pathway
from initiative to initiative needs to be as clear and quantitative as possible
.
FIGURE 4.4 General Strategy Map
Financial Improve financial results
Customers
Improve marketing and customer service
Business Processes
Improve operations
Learning and Growing
Provide employees with skills, tools, and motivation
89
Vincent Valley Hospital and Health System Strategy Maps
VVH has two major areas of strategic focusthe birthing center and the emergency d
epartment. Figure 4.5 displays the strategy map for the birthing center. Recall
that VVH had decided to execute three major projects in this area. Other initiat
ives needed for the successful execution of each project are identified on the m
ap. For instance, for nursing staff to successfully lead expectant mothers on to
urs of labor and delivery suites, they will need to participate in a customer se
rvice training program. After the tours begin, the birthing center will measure
potential patients satisfaction to ensure that the tours are being conducted effe
ctively. Once patients deliver their babies in VVHs obstetric unit, they will aga
in be surveyed on their experience with special questions on the effect of each
of these major projects. These leading satisfaction indicators should predict th
e lagging indicators of increased market share and net revenue. The second major
strategy for VVH was to improve patient flow in the emergency department. Figur
e 4.6 shows the strategy map for the emergency department. The first required st
eps in this strategy are forming a project team (Chapter 5) and learning how to
use Lean process improvement tools (Chapter 9). Then, the team can begin analyzi
ng patient flow and implementing changes to
Financial
Increase net revenue of obstetric product line Goal = 10%
FIGURE 4.5 VVH Birthing Center Strategy Map
Measure market share Goal = 5% increase Customers Measure patient satisfaction (
facilities) Goal 90% satisfaction ^ Measure patient satisfaction (perceived clin
ical quality) Goal 90% satisfaction ^ Measure patient satisfaction (high touch)
Goal 90% satisfaction ^
Business Processes
Remodel obstetric suite Goal = complete by November 1
Contract for emergency transportation Goal = 10 runs/month
Begin tours and survey Goal = patient satisfaction 90% ^
Customer service training Goal = 90% average passing score
90
FIGURE 4.6 VVH Emergency Department Strategy Map
Financial
Increase net revenue of emergency department production line Goal = 10%
Customers
Measure patient wait time Goal 30 minutes ^
Measure patient share Goal = 5% increase
Business Processes
Do project on patient flow and make changes Goal = Value stream increased by 30%
Learn Lean process improvement tools Goal = Complete by December 1
improve flow. VVH has set a goal of reducing the amount of nonvalueadded time by 30
percent. Once this is accomplished, waiting time for 90 percent of patients sho
uld not exceed 30 minutes. A reduced waiting time should result in patients bein
g more satisfied and, hence, a growth in market share and increased net revenue.
Following are more formal cause-and-effect statements: If emergency department
staff undertakes educational activities to learn project management and Lean, th
en they can effectively execute a patient flow improvement project. If a patient
flow project is undertaken and nonvalue-added time is reduced by 30 percent, the
n the waiting time for 90 percent of the patients should never exceed 30 minutes
. If the waiting time for most patients never exceeds 30 minutes, then they will
be highly satisfied and this will increase the number of patients and VVHs marke
t share. If the emergency department market share increases, then net revenue wi
ll increase. The companion web site, ache.org/books/OpsManagement, contains a do
wnloadable strategy map and linked scorecard. It also has a number of videos
91
that demonstrate how to use and modify these tools for both student and practiti
oner use.
Implementation of the Balanced Scorecard
Linking and Communicating
The balanced scorecard can be used at many different levels in an organization.
However, departmental scorecards should link to the divisional, and ultimately t
he corporate, level. Each scorecard should be linked upward and downward. For ex
ample, an obstetric initiative to increase revenue from normal childbirth will b
e linked to the corporate-level objective of overall increased revenue. Sometime
s it is difficult to specifically link a departmental strategy map to corporate
objectives. In this case, the department head must make a more general linkage b
y stating how a departmental initiative will influence a particular corporate go
al. For example, improving the quality of the hospital laboratory testing system
will generally influence the corporate objective that patients should perceive
that the hospital provides the highest level of quality care. The development an
d operation of scorecards at each level of an organization requires disciplined
communication, which can be an incentive for action. Balanced scorecards can als
o be used to communicate with an organizations external stakeholders. A well-impl
emented balanced scorecard system will be integrated with individual employee go
als and the organizations performance management system.
Targets, Resources, Initiatives, and Budgets
A balanced scorecard strategy map consists of a series of linked initiatives. Ea
ch initiative should have a quantitative measure and a target. Initiatives can r
eside in one department, but they are frequently cross-departmental. Many initia
tives are projects, and the process for successful project management (Chapter 5
) should be followed. A well-implemented balanced scorecard will also link caref
ully to an organizations budget, particularly if initiatives and projects are exp
ected to consume considerable operating or capital resources. The use of the bal
anced scorecard does not obviate the use of additional operating statistics. Man
y other operating and financial measures still need to be collected and analyzed
. If the performance of any of these measures deviates substantially from its ex
pected target, a new strategy and initiative may be needed. For example, most he
althcare organizations carefully track and monitor their accounts receivable. If
this financial measure is within industry norms, it probably will not appear on
an organizations balanced scorecard. However, if the accounts receivable balance
drifts over time and
92
begins to exceed expectations, a balanced scorecard initiative may be started to
address the problem.
Displaying Results
The actual scorecard tracks and communicates the results of each initiative. (Ch
apter 7 provides examples of many types of visual display.) A challenge for most
organizations is to collect the data to display in the scorecard. Because the s
corecard should have fewer than 20 measures, a simple solution is to assign this
responsibility to one individual who develops efficient methods to collect the
data and determines effective methods to display them. The companion web site in
cludes a straightforward balanced scorecard, built in Excel, that can store and
display one year of data. Figure 4.7 shows the cover worksheet of this scorecard
.
Ensuring that the Balanced Scorecard Works
The explicit purpose of the balanced scorecard is to ensure the successful execu
tion of an organizations strategy. But what if it does not achieve the desired re
sults? There are two possible causes for this problem. The first, most obvious,
problem is that the initiative itself is not achieving its targeted results. For
example, VVHs patient flow project may not be able to decrease nonvalued-added ti
me by 30 percent. Therefore, VVH may have to employ another tactic, such as enga
ging a consultant. It will be important to monitor this measure frequently and p
ost it on the scorecard. The second, more complex, problem occurs when the succe
ssful execution of an initiative does not cause the achievement of the next link
ed target. For example, although patient satisfaction scores in VVHs obstetric de
partment increased to target levels, VVH may not gain market share. The solution
to this problem is to reconsider the cause-and-effect relationships. VVH may ne
ed to add an advertising campaign as an initiative to begin to shift market shar
e. The results and strategy map for an organization should be reviewed at least
quarterly and must be reviewed and revised annually, usually as part of the budg
eting process. The most advanced way to review the results of a balanced scoreca
rd is to use quantitative tools such as correlation and regression. Pineno (2002
) developed an incremental modeling approach to link the following performance m
easures as part of a hospital balanced scorecard: financial performance, efficie
ncy, liquidity, capital, process quality, outcomes, clinical quality, number of
hospital partners, housekeeping, information use, coordination of care, communit
y linkages, readmissions, length of stay, complications, surgery rates, and proc
edure rates. By varying elements in this linked balanced scorecard model, financ
ial performance can be predicted.
93
700,000
Net revenue
YTD Actual Goal 2,877,842 3,266,667
600,000 500,000 400,000 300,000 200,000 100,000 0 Jan
120%
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Patient satisfaction
YTD Actual Goal
100% 80% 60%
FIGURE 4.7 Balanced Scorecard Template (Contained on book companion web site)
88% 90%
40% 20% 0% Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
160
Increase in admissions
YTD Actual Goal
140 120 100 80 60 40 20 0 Jan
102% 100% 98% 96% 94% 92% 90% 88% 86% 84% 82% Jan
6.0%
83 100
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Customer service training tests
YTD Actual Goal
94% 90%
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Profit
YTD Actual Goal
5.0% 4.0% 3.0%
3.2% 3.0%
2.0% 1.0% 0.0% Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
35%
FT admission (%)
YTD Actual Goal
30% 25% 20%
16% 30%
15% 10% 5% 0% Jan 250 Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Cost/unit
YTD Actual Goal
200 150
134 115
100 50 0 Jan
120% 100% 80% 60%
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Six Sigma training tests
YTD Actual Goal
86% 90%
40% 20% 0% Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
94
Modifications of the Classic Balanced Scorecard
The balanced scorecard has been modified by many healthcare organizations, most
commonly by placing the customer or patient at the top of the strategy map (Figu
re 4.8). Finance then becomes a means to achieve superior patient outcomes and s
atisfaction. Other modifications have been made by leading healthcare organizati
ons. For example, the Mayo Clinic has added new perspectivesincluding mutual resp
ect and diversity, social commitment, and external environmental assessmentsto it
s scorecard (Curtright and Stolp-Smith 2000).
Implementation Issues
Two common challenges occur when implementing balanced scorecards: (1) determina
tion and development of metrics, and (2) initiative prioritization. The balanced
scorecard is a quantitative tool and, therefore, requires data systems that gen
erate timely information for inclusion. Each initiative on a strategy map should
have quantitative measures, which should represent an even mix of leading and l
agging measures. Each initiative should have a target as well. However, setting
targets is an art: Too timid a goal will not move the organization forward, and
too aggressive a goal will be discouraging.
FIGURE 4.8 Rotated General Strategy Map
Customers
Improve patient results and satisfaction
Business Processes
Improve operations
Financial
Improve availability of financial resources
Provide employees with skills, tools, and motivation
95
Niven (2002) suggests that a number of sources should be used to construct targe
ts. These include internal company operating data, executive interviews, interna
l and external strategic assessments, customer research, industry averages, and
benchmarking data. Targets can be incremental based on current operating results
(e.g., increase productivity in a nursing unit by 10 percent in the next 12 mon
ths), or they can be stretch goals, which are possible but will require extraordin
ary effort to achieve (e.g., improve compliance with evidence-based guidelines f
or 98 percent of patients with diabetes). Too many measures and initiatives will
confuse a scorecard, and even the most sophisticated organizations therefore li
mit their measures to 20 or fewer. The second major implementation challenge is
the prioritization of initiatives. Most organizations do not lack for current in
itiatives or projects, and influential leaders in the organization often propose
many more. Niven (2002, 190) suggests a methodology to manage this phenomenon:
1. 2. 3. 4. Inventory all current initiatives taking place within the organizati
on. Map those initiatives to the objectives of the balanced scorecard. Consider
eliminating nonstrategic initiatives, and develop missing initiatives. Prioritiz
e the remaining initiatives.
Most organizations will never be able to achieve perfect alignment with their ba
lanced scorecard goals for all of their initiatives. However, the closer the ali
gnment, the more likely the organizations strategic objectives will be achieved.
Conclusion
This text is about how to get things done. The balanced scorecard provides a pow
erful tool to that end because it: links strategy to action in the form of initi
atives; provides a comprehensive communication tool inside and outside an organi
zation; and is quantitatively based, providing a tool for ongoing strategy analy
sis and improvement.
1. What other indicators might be used in each of the four perspectives for publ
ic health agencies? For health plans? 2. If you were to add another perspective,
what would it be? Draw a strategy map of a healthcare delivery organization and
include this perspective.
96
3. How do you manage the
appear on a strategy map
ts balanced scorecard to
involve the customer or
scorecard?
other operations of an
or balanced scorecard?
a corporate scorecard?
patient in identifying
organization, those that do not

4. How would a department link i
5. What methods could be used to
the key elements of the balanced
Chapter Exercise
View the web videos on ache.org/books/OpsManagement and download the PowerPoint
strategy map and Excel worksheet. Develop a strategy map and balanced scorecard
for a primary care dental clinic. Do Internet research to determine the challeng
es facing primary care dentistry, and develop a strategy map for success in this
environment. Make sure that the strategy map includes at least eight initiative
s and that they touch the four perspectives. Include targets, and be sure the me
trics are a mix of leading and lagging indicators. Develop a plan to periodicall
y review your map to ascertain its effectiveness.
Case Study: St. Marys Duluth Clinic
An innovative leader in healthcare for northeastern Minnesota and northwestern W
isconsin, St. Marys Duluth Clinic Health System (SMDC) encompasses 20 clinics, a
350-bed tertiary medical center, two community hospitals, and a specialty care f
acility. SMDCs medical team of more than 400 physicians and 200 allied healthcare
providers work with an experienced staff of more than 6,000 to provide primary
care, specialty services, and medical technology to families in their own commun
ities. SMDC began using the balanced scorecard when it was a newly merged organi
zation facing decreased revenues as a result of financial constraints in the hea
lthcare environment, such as the Balanced Budget Act. Since then, the balanced s
corecard has become the cornerstone for all strategic decisions, providing direc
tion for daily decision making. Not only has SMDC made the balanced scorecard th
e framework for all leadership sessions, but it has also used the balanced score
card to align its service lines and regional community clinics, link its budget
to the strategy, and disseminate strategic awareness to every employee. In addit
ion to improved fiscal strength, SMDCs patients are also feeling better; they now
enjoy easier access to the primary care clinics, and their overall satisfaction
with both hospitals and clinics continues to climb. With such significant gains
, SMDC has created a promising prognosis for its own institution and for those i
t serves.
FIGURE 4.9
Corporate Strategy Map with Measures: Duluth Clinic Focus, Fiscal Years 20062008
FY 0608 Corporate Strategy Map with MeasuresDuluth Clinic Focus
Vision: SMDC will be the best place to work and the best place to receive care
We will pursue our vision through a focus on Quality, Safety, and Value
C2 Clinical Expertise Effective Care Safe Care C3 Customer Value Efficient Care
Customer To achieve our Quality Vision, how should we appear to our internal and
external customers?
Clinical Excellence P1 Right patient, right care, right process Management Excel
lence P6 Optimized physican and staff productivity Monthly RVU volume
C1 Quality Relationships Timely Care Patient-centered Care Equitable Care Pt Sat
isfaction score Pt Perception overall quality of doctor care
Service Excellence
Internal Processes To satisfy our customers, at which operational and quality pr
ocesses must we excel?
Pts on continuous opioid treatment who have opioid agreements in force (%) Pts o
verdue for INR check (%) P5 Design and implement coordinated care models to effe
ctively manage disease processes Pts with optimal diabetes control (%) Hypertens
ion measure Well-child measure
P2 Provide easy, timely, coordinated access to healthcare services
Specialty sections meeting access standard (%) Primary care open access
P4 Create direct access to specific programs and services
P7 Excel in efficient and effective operations
P3 Capital investment emphasis in technology and facilities to support clinical
and financial objectives Performance to epic implementation plan Performance to
technology replacement plan
Learning and Growth How will we sustain our ability to change and improve as a s
ystem?
L1 Create an environment that has an expectation of consistently delivering qual
ity healthcare
L2 Emphasize entity and system missions
L3 Recruit and develop people to outstanding levels of performance Physicians re
cruited in targeted areas (%) Clinical education section meetings held (#)
L4 Engage physician leaders as partners in the success with SMDC Sections conduc
ting monthly section/site meetings (%) Physician leaders attending leadership se
ssions (%)
F1 Achieve a 3% operating margin by FY08 to sustain our mission and achieve our
vision DC margin fully allocated DC margin without indirect allocations F5 Captu
re earned revenue F3 Restructure regional clinic business and clinical models F4
Redesign Duluth and Superior primary care strategy
Financial To financially sustain our mission, on what must we focus?

*F2 Grow key specialty services and programs and stimulate demand for primary ca
re
*Cancer, Cardiovascular, Digestive, Neuromusculoskeletal
Procedure volume for targeted areas Surgery volume New DC Main specialty patient
s in target growth areas (#) Encounters in all sections (#) New Duluth/Sup prima
ry care patients (#)
97
SOURCE: St. Marys/Duluth Clinic Health System. Used with permission.
98
Examination of the strategy map shows that SMDC has rotated the financial perspe
ctive to the bottom and put the customer at the top (Figure 4.9). In addition, m
any of the initiatives mirror the improvement goals from Crossing the Quality Ch
asm (Institute of Medicine 2001), as discussed in Chapter 1. SMDC has only 19 in
itiatives on its strategy map, indicating that the organization is focused on th
ose items that are key to moving it forward.
References
Curtright, J. W., and S. C. Stolp-Smith. 2000. Strategic Performance Management:
Development of a Performance Measurement System at the Mayo Clinic. Journal of He
althcare Management 45 (1): 5868. Gapenski, L. C. 2005. Healthcare Finance: An In
troduction to Accounting and Financial Management. Chicago: Health Administratio
n Press. Inamdar, N., and R. S. Kaplan. 2002. Applying the Balanced Scorecard in
Healthcare Provider Organizations. Journal of Healthcare Management 47 (3): 17995.
Institute of Medicine. 2001. Crossing the Quality ChasmA New Health System for t
he 21st Century. Washington, D.C.: National Academies Press. Kaplan, R. S., and
D. P. Norton. 2001. The Strategy-Focused Organization: How Balanced Scorecard Co
mpanies Thrive in the New Business Environment. Boston: Harvard Business School
Press. . 1996. The Balanced ScorecardTranslating Strategy into Action. Boston: Harva
rd Business School Press. Mankins, M. C., and R. Steele. 2005. Turning Great Stra
tegy into Great Performance. Harvard Business Review 83 (7): 6472. Mountain States
Group. 2003. Balanced Scorecards for Small Rural Hospitals: Concept Overview an
d Implementation Guidance. Publication No. ORHP00346. Health Resources and Servi
ces Administration, U.S. Department of Health and Human Services. [Online inform
ation; retrieved 8/15/06.] http:// tasc.ruralhealth.hrsa.gov/documents/Final%20B
SC%20Manual%20 edits%2010.18F.pdf. National Quality Forum. 2002. Serious Reporta
ble Events in Healthcare: A National Quality Forum Consensus Report. Publication
No. NQFCR01-02. Washington, D.C.: National Quality Forum. Niven, P. R. 2002. Ba
lanced Scorecard Step-by-Step: Maximizing Performance and Maintaining Results. N
ew York: John Wiley & Sons. Pineno, C. J. 2002. The Balanced Scorecard: An Increm
ental Approach Model to Health Care Management. Journal of Health Care Finance 28
(4): 6980. Zelman, W. N., G. H. Pink, and C. B. Matthias. 2003. Use of the Balanc
ed Scorecard in Health Care. Journal of Health Care Finance 29 (4): 115.
99
Further Reading
Bilkhu-Thompson, M. K. 2003. A Process Evaluation of a Health Care Balanced Score
card. Journal of Health Care Finance 30 (2): 3764. Cleverley, W. O., and J. O. Cle
verley. 2005. Scorecards and Dashboards. Healthcare Financial Management 59 (7): 6
469. Gumbus, A., B. Lyons, and D. E. Bellhouse. 2002. Journey to Destination 2005.
Strategic Finance 84 (2): 4650. MacStravic, S. 1999. A Really Balanced Scorecard. H
ealth Forum Journal 42 (3): 6467. Meyers, S. 2004. Data in, Safety out. Trustee 57
(7): 1219. Tarantino, D. P. 2003. Using the Balanced Scorecard as a Performance Ma
nagement Tool. Physician Executive 29 (5): 6972. Wyatt, J. 2004. Scorecards, Dashbo
ards, and KPIs: Keys to Integrated Performance Measurement. Healthcare Financial
Management 58 (2): 7680.
CHAPTER
5
PROJECT MANAGEMENT
CHAPTER OUTLINE
Operations Management in Action Overview Definition of a Project Project Selecti
on and Chartering Project Selection Project Charter Stakeholder identification a
nd dialogue Feasibility analysis Project charter document Project Scope and Work
Breakdown Tools Scope Work Breakdown Structure Scheduling Network Diagrams and
Gantt Charts Slack and the Critical Path Crashing the Project Project Control Mo
nitoring Progress Earned Value Analysis Change Control Communications Risk Manag
ement Quality Management, Procurement, the Project Management Office, and Projec
t Closure Quality Management Procurement Contracting Selecting a vendor Payment
based on earned value The Project Management Office Project Closure The Project
Manager and Project Team Team Skills Team Structure and Authority Team Meetings
Dialogue Leadership Conclusion Discussion Questions Chapter Exercises References
100

actual cost of work performed (ACWP) budgeted cost of work performed (BCWP) budg
eted cost of work scheduled (BCWS) cost performance index (CPI) critical index (
CI) critical path method earned value analysis estimate at completion (EAC) Gant
t chart Microsoft Project mitigation plan network diagram program evaluation and
review technique (PERT) procurement project charter project closure project con
trol Project Management Body of Knowledge (PMBOK) Project Management Professiona
l (PMP) RASIC chart request for information (RFI) request for proposal (RFP) ris
k management schedule performance index (SPI) slack stakeholder statement of wor
k (SOW) strengths, weaknesses, opportunities, and threats analysis variance at c
ompletion work breakdown structure (WBS)
101

The risk of poor project management can be seen in many organizations struggling
to implement electronic health records (EHRs). One of the most publicized failu
res of a large-scale project was a $34 million EHR project at a major West Coast
teaching hospital. Although this hospital had an outstanding reputation and del
ivered leading-edge care, the medical staff voted to turn off their new system a
fter three months and return to paper charting. One leading physician said, A tas
k that once took three minutes to scribble shorthand at the patients bedside sudd
enly devoured 30 to 40 minutes. Whos got five extra hours in a day?
SOURCE: Connolly 2005.
Overview
Everyone manages projects, whether painting a bedroom at home or adding a 100-be
d wing to a hospital. This chapter provides grounding in the science of project
management. The major topics covered include: Selecting and chartering projects;
Using stakeholder analysis to set project requirements; Developing a work break
down structure and schedule; Using Microsoft Project to develop project plans an
d monitor cost, schedule, and earned value; Managing project communications, cha
nge control, and risk; and Creating and leading project teams. After reading thi
s chapter and completing the associated exercises, readers should be able to: Cr
eate a project charter with a detailed plan for costs, schedule, scope, and perf
ormance; Monitor the progress of a project, make changes as required, communicat
e to stakeholders, and manage risks; and Develop the skills to successfully lead
a project team. If everyone manages projects, why is there a need to devote a c
hapter in a healthcare operations book to this topic? The answer lies in the que
stion: Although everyone has life experiences in project management, few healthc
are professionals take the time to understand and practice the science and disci
pline of project management. The ability to successfully move a project forward
while meeting time and budget goals is a distinguishing characteristic of a high
-quality, highly competitive healthcare organization.
102
C h a p t e r 5 : Pro j e c t M a n a g e m e n t
103
Effective project management provides an opportunity for progressive healthcare
organizations to quickly develop new clinical services, fix major operating prob
lems, reduce expenses, and provide new consumer-directed products to their patie
nts. The problems with poor project management became apparent in the defense in
dustry after World War II. Many new weapons systems were wildly over budget, wer
e late, and did not perform as expected. The automated baggage conveyor system a
t Denver International Airport is another frequently cited example of poor proje
ct management. In 2005, after 10 years of malfunctions and high maintenance cost
s, it was turned off and baggage is now handled manually. A response to this pro
blem was the gradual development of project management as a discipline, culminat
ing in the establishment of the Project Management Institute (PMI) in 1969 (www.
PMI.org). Today, PMI has more than 110,000 members and more than 50,000 of those
are certified as Project Management Professionals (PMPs). PMI members have deve
loped the Project Management Body of Knowledge (PMBOK) (PMI 2004), which details
best practices for successful project management. Much as evidence-based medici
ne delineates the most effective methods to care for specific clinical condition
s, the PMBOK provides sciencebased, field-tested guidelines for successful proje
ct management. This chapter is based on PMBOK principles as applied to healthcar
e. Healthcare professionals who spend much of their time leading projects should
consider using resources available through PMI; for some, PMP certification may
be appropriate.
Definition of a Project
A project is a one-time set of activities that culminates in a desired outcome.
Therefore, activities that occur repeatedly, for example, making appointments fo
r patients in a clinic, are not projects. However, the installation of new softw
are to upgrade this capability would be a project. A major process improvement e
ffort to reduce phone hold time for patients would also qualify as a project. Sl
ack (2005) provides a useful tool for determining the need for formal project ma
nagement (Figure 5.1). Operating issues arise frequently; when simple, they can
be fixed immediately by operating staff. More complicated problems can be addres
sed by using the tools detailed in Chapter 6. However, projects that are complex
and have high organizational value need the discipline of formal project manage
ment. Many of the strategic initiatives on an organizations balanced scorecard sh
ould use the project management methodology. A well-managed project will include
a specified scope of work, expected outcomes and performance levels, budget, an
d detailed work breakdown tied to a schedule. It will also include a formal chan
ge procedure, communications plan, and plan to deal with risk. Finally, all good
projects will include a project conclusion process and plan for redeployment of
staff.
104
FIGURE 5.1 When to Use Project Management
Complex
Level of detail and problem solving
Simple
Find it , Fix it Problem-solving process
Project management
SOURCE: Slack (2005). Used with permission.
Many high-performing organizations will also have a formal executivelevel charte
ring process for projects and a project management office to monitor enterprisewide project activities. Some healthcare organizations (e.g., health plans) can
have a substantial share of their operating resources invested in projects at an
y one time. For effective execution of a project, PMI recommends that three elem
ents must be present. A project charter begins the project and addresses stakeho
lder needs. A project scope statement identifies project outcomes, timelines, an
d budget in detail. Finally, to execute the project, a project plan must be deve
loped; the plan includes scope management, work breakdown, schedule management,
cost management, quality control, staffing management, communications, risk mana
gement, procurement, and close-out process. Figure 5.2 displays the relationship
s among these elements.
Project Selection and Chartering
Project Selection
Most organizations have many projects vying for attention, funding, and senior e
xecutive support. The annual budget and strategic planning process serves as a u
seful vehicle for prioritizing projects in many organizations. The balanced scor
ecard (Chapter 4) will help guide the identification of worthwhile strategic pro
jects. Other external forces (e.g., new Medicare rules) or clinical innovations
(e.g., new imaging technologies), however, will conspire to present an organizat
ions leadership with a list of projects too long for successful implementation. N
iven (2005) provides a useful tool to prioritize projects (Table 5.1).
105
Initiation and charter
Scope requirements
Project plan
Scope management and work breakdown Stakeholders Schedule management Cost manage
ment Quality control Communications Risk management
FIGURE 5.2 Complete Project Management Process
Procurement Close-out process
Criteria Linkage to strategy Financial gain Project cost Key personnel required
Time to complete Affects other projects Total
Project A Weight points 45% 7 15% 10% 10% 10% 10% 5 5 8 8 3
Project A score 3.15 0.75 0.50 0.80 0.80 0.30 6.30
Project B points 1 10 10 10 10 10
Project B score 0.45 1.5 1.0 1.0 1.0 1.0 5. 95
TABLE 5.1 Project Prioritization Matrix
SOURCE: Niven, P. R. 2002. Balanced Scorecard Step by Step. Figure 7.2: Prioriti
zing Balanced Scorecard Initiatives, page 194. Used with permission of John Wile
y & Sons, Inc.
106
To use this tool, each potential project should be scored by a senior planning g
roup based on how well it fits into the organizations strategy, financial benefit
and cost, need for key personnel, time required, and positive effect on other p
rojects. A scale of one (low) to ten (high) is usually used. Each criterion is a
lso weighted; the scores are multiplied by their weight for each criterion and s
ummed over all of the criteria. In Table 5.1, Project A has a higher total score
due to its importance to the organizations strategy. Such a ranking methodology
helps organizations avoid committing resources to projects that may have a power
ful internal champion but do not advance the organizations overall strategy. This
matrix can be modified with other categories and weights based on an organizati
ons current needs. Another, less quantitative methodology for evaluating clinical
quality improvement projects has been suggested by Mosser (2005): How frequentl
y does the disease or condition occur? What is the likelihood of success? (For e
xample, is there broad evidencebased clinical research to support a new care pro
cess?) How large is the performance gap (today versus the ideal) to be closed? T
he answers to these questions will guide the prioritization of clinical quality
improvement projects.
Project Charter
Once a project is identified for implementation, it needs to be chartered. Four
factors interact to constrain the execution of a project charter: time, cost, sc
ope, and performance. A successful project will have a scope that specifies the
resulting performance level, how much time it will take, and its budgeted cost.
A change in any one of these factors will affect the other three. This can be ex
pressed mathematically as follows: Scope = f (Time, Cost, Performance)
Similarly: Time = f (Cost, Scope, Performance)
and so on. Figure 5.3 demonstrates these relationships graphically. Here, the ar
ea of the triangle is a measure of the scope of the project. The length of each
side of the triangle indicates the amount time, money, or performance needed in
the project. Because each side of the triangle is connected, changing any of the
se parameters affects the others. Figure 5.4 shows this same
107
Cost = f (Performance, Time, Scope)
Performance Scope
Cost
FIGURE 5.3 Relationship of Project Scope to Performance Level, Time, and Cost
Time
project with an increase in required performance level and shortened timelines.
With the same scope, this new project will incur additional costs. Although it is
difficult to specifically and exactly determine the relationship between all fou
r factors, the successful project manager understands this general relationship
well and communicates it to project sponsors. A useful analogy is the balloon: I
f you push hard on part of it, a different part will bulge out. The classic proj
ect management dilemma is an increase in scope without additional time or fundin
g (sometimes termed scope creep). Many project failures are directly attributable
to ignoring this unyielding formula.
Stakeholder identification and dialogue The first step in developing a project c
harter is to identify the stakeholders, in general, anyone who has a stake in th
e outcome of the project. Key stakeholders on a project include the project mana
ger, customers, users, project team members, any contracted organizations involv
ed, project sponsor, those who can influence the project, and the project manage
ment office, if one exists in the organization. The project manager is the indiv
idual held accountable for the projects success and, therefore, forms the core of
the stakeholder group. The customer or user of the service or product is an imp
ortant stakeholder who will influence and help determine the performance of the
final product. Even if project team members serve on the project in a limited pa
rt-time role, the success of the project reflects on them and, therefore, they b
ecome stakeholders. A common
FIGURE 5.4 Project with Increased Performance Requirement and Shortened Schedule
Performance Scope
Cost
Time
108
contracting relationship in healthcare involves large information technology (IT
) installations provided through an outside vendor, which would also be included
as a project stakeholder. A project should always have a sponsor with enough ex
ecutive-level influence to clear roadblocks as the project progresses; hence, su
ch individuals need to be included in the stakeholder group. A project may be ai
ded or hindered by many individuals or organizations that are not directly part
of it; a global systems analysis should be performed (Chapter 1) to identify whi
ch of these should be included as stakeholders. Once stakeholders have been iden
tified, they need to be interviewed by the project manager to develop the projec
t charter. If an important stakeholder is not available, the project manager sho
uld interview someone who represents her interests. At this point, it is importa
nt to differentiate between the needs and wants of the stakeholders; however, en
ough detail needs to be gathered to construct the project charter. When the proj
ect team is organized, it need not include all stakeholders, but the team should
be vigilant in attempting to meet all stakeholder needs. The project team shoul
d also be cognizant of the culture of the organization, sometimes defined as how
things get done around here. Projects that challenge an organizations culture will
encounter frequent difficulties. The project charter is the document that formal
ly authorizes a project. The project charter provides the project manager with t
he authority to apply organizational resources to project activities (PMI 2004, 8
1). A project initiator, or sponsor external to the project, issues the charter
and signs it to authorize the start of the project.
Feasibility analysis An important part of the project charter is determining the
projects feasibility. Because the project should have already had an initial pri
oritization by the senior management team, the link to the organizations strategy
has likely already been made. However, this link should be documented in the fe
asibility analysis. The operational and technical feasibility should also be exa
mined. For example, if a new clinical project requires the construction of new f
acilities, this may impede its execution. An initial schedule should also be con
sidered, as a needed completion date may be clearly impossible. Finally, both fi
nancial benefit and marketplace demand should be considered here. For details on
conducting a financial feasibility analysis, see Gapenski (2005). All elements
of the feasibility analysis should be included in the project charter document.
Project charter document The project charter authorizes the project and serves a
s an executive summary. A formal charter document should be constructed with the
following elements:
Project mission statement; Project purpose or justification and linkage to strat
egic goals;
109
High-level requirements that satisfy customers, sponsors, and other stakeholders
; Assigned project manager and authority level; Summary milestones; Stakeholder
influences; Functional organizations and their participation; Organizational, en
vironmental, and external assumptions and constraints; Financial business case,
budget, and return on investment (ROI); and Project sponsor with approval signat
ure. A project charter template is contained on the companion web site. The init
ial description of the scope of the project is found in the Requirements, Mileston
es, and Financial sections of the project charter. A project charter can be illu
strated with an example from Vincent Valley Hospital and Health System (VVH). Th
e hospital operates a primary care clinic (Riverview Clinic) in the south suburb
an area of Bakersville. Recently, the three largest health plans in the area ins
tituted pay-for-performance programs in the areas of diabetes, asthma, congestiv
e heart failure, and generic drug use. The health plans will pay primary care cl
inics bonuses if they achieve specific levels of performance in these areas. Riv
erview staff have decided to embark on a project to increase their use of generi
c drugs; their project charter is displayed in Figure 5.5. FIGURE 5.5 Project Ch
arter for VVH Generic Drug Project
Project Mission Statement This project will increase the level of generic drug p
rescriptions to lower the costs to our patients and increase reimbursements to t
he clinic. Project Purpose and Justification Health plans in Bakersville have be
gun to provide additional funding to clinics that meet pay-for-performance guide
lines. Although a number of chronic conditions are covered by these new payment
systems, it is felt that generic drug use should be the first project executed b
ecause it is likely to be accomplished in a reasonable time frame with the maxim
um financial benefit to our patients and the clinic. Once this project has been
executed, the clinic will move on to more complex clinical conditions. The team
will be able to incorporate what they have learned about some of the barriers to
success and methods to succeed on pay-for-performance projects. This project is
a part of the larger VVH strategic initiative of maximizing pay-forperformance
reimbursement. High-Level Requirements Once completed, a new prescribing process
will Continue to meet patients clinical needs and provide high-quality care; and
Increase generic drug use by 4 percent from baseline within six months.
110
FIGURE 5.5 Project Charter for VVH Generic Drug Project (continued)
Assigned Project Manager and Authority Level Sally Humphries, RN, will be the pr
oject manager. Sally has authority to make changes in budget, time, scope, and p
erformance by 10 percent. Any larger change requires approval from the clinic op
erating board. Summary Milestones The project will commence on January 1. A syst
em to identify approved generic drugs will be available on February 15. The syst
em will go live on March 15.
Stakeholder Influences The following stakeholders will influence the project: Cl
inicians will strive to provide the best care for their patients. Patients will
need to understand the benefits of this new system. Clinic staff will need train
ing and support tools. Health plans should be a partner in this project as part
of the supply chain. Pharmaceutical firms should provide clinical information on
the efficacy of certain generic drugs.
Functional Organizations and Their Participation Clinic management staff will le
ad. Compcare EHR vendor will perform software modifications. VVH IT department w
ill support. VVH main pharmacy department will support.
Organizational, Environmental, and External Assumptions and Constraints Success
depends on appropriate substitution of generic for brand-name drugs. Patients ne
ed to understand the benefits of this change. Health plans need to continue to f
und this project over a number of years. IT modifications need to be approved ra
pidly by the VVH central IT department.
Financial Business CaseROI The project budget is $61,000 for personnel. Software
modifications are included in the master VVH contract and, therefore, have no di
rect cost to this project. If the 4 percent increase in generic drug use is achi
eved, the two-year revenue increase should be approximately $75,000. Project Spo
nsor with Approval Signature Dr. Jim Anderson, Clinic President
James Anderson, MD
111
Project Scope and Work Breakdown
Once a project has been chartered, the detailed work of planning can begin.
Tools
At this point, the project manager should consider acquiring two important tools
. The first is the lowest of low tech, the humble three-ring binder. All project
s need a continuous record of progress reports, team meetings, approved changes,
and so on. A complex project will require many binders, but they will prove inv
aluable to the project manager. The classic organization of the binders is by da
te, so the first pages will be the project charter. Of course, if the organizati
on has an effective imaging and document management system, this can substitute
for the binders. The second tool is project management software. Although many o
ptions are available, the market leader is Microsoft Project, which is referred
to throughout the remainder of the chapter. Microsoft Project is part of the Mic
rosoft Office suite and may already be on many computers. If not, a demonstratio
n copy can be downloaded from Microsoft. The companion web site for this book pr
ovides additional explanation and videos related to the use of Project, along wi
th detailed illustrations of the softwares use for the Riverview Clinic generic d
rug project. Go to ache.org/books.OpsManagement, Chapter 5, to view a video on s
tarting Project and setting its global parameters. Project management software i
s not essential for small projects, but it is helpful and almost required for an
y project that lasts longer than six months and involves a large team of individ
uals. Although the Riverview Clinic generic drug project is relatively small, Pr
oject software is used to manage it to provide an illustration of the programs ap
plicability.
Scope
The project scope determines what falls inside and outside the scope of a projec
t. The starting point for developing a scope document is the project charter, al
though the scope statement is much more detailed than the description contained
in the project charter. The project manager will want to revisit many of the sam
e stakeholders to acquire more detailed inputs and requirements. A simple method
ology suggested by Lewis (2000) is to interview stakeholders and ask them to lis
t the three most important outcomes of the project, which can be combined into p
roject objectives. Lewis also suggests that objectives be specific, achievable,
measurable, and comprehensible to stakeholders. In addition, they should be stat
ed in terms of time-limited deliverables. The objective improve the quality of care
to patients with diabetes is a poor one. Improve the rate of foot exams by 25 per
cent in one year for patients with diabetes is a much better objective.
112
Austin and Boxerman (2003) provide a detailed interview methodology for clinical
IT projects in Information Systems for Health Care Management. Quality function
deployment is also a useful tool for specifying objectives and customer require
ments (Chapter 8). In scope creation, it is important to avoid expanding the sco
pe of the project: While we are at it, we might as well ____. These ideas, sometim
es called gold plating, tend to be some of the most dangerous in the world of proj
ect management. The scope document should also provide detailed requirements and
descriptions of expected outcomes. A good scope document is also specific about
project boundaries. The Riverside Clinic project scope document might include a
statement that the project does not include access to online pharmacological da
tabases. The scope document should specify deliverables such as implementation o
f a new process, installation of a new piece of equipment, or presentation of a
paper report. The project organization is also specified in the scope document,
including the project manager, team members, and specific relationships to all p
arts of the organization. An initial evaluation of potential risks to the projec
t should be enumerated in the scope document. The schedule length and milestones
should be more detailed in the scope statement than in the charter. As discusse
d in the next section, however, the final schedule will be developed based on th
e work breakdown structure. Finally, the scope document should include methods t
o monitor progress and make changes where necessary, including the formal approv
als required.
Work Breakdown Structure
The second major component of the scope document is the work breakdown structure
(WBS), considered the engine of the project because it determines how its goals
will be accomplished. The WBS lists the tasks that need to be accomplished, inc
luding an estimate of the resources required (e.g., staff time, services, equipm
ent). For complex projects, the WBS is a hierarchy of major tasks, subtasks, and
work packages. Figure 5.6 demonstrates this graphically. The size of each task
should be constructed carefully. A task should not be so small that monitoring w
ould consume a disproportionate share of the task itself. However, an overly lar
ge task cannot be effectively monitored and should be divided into subtasks and
then work packages. The task should have enough detail associated with it that t
he individual responsible, the cost, and duration can be identified. A reasonabl
e guideline is that a task should have a duration of one to three weeks to be ef
fectively monitored. The completion of some tasks is critical to the success of
the project. These tasks should be identified as milestones. The completion of mil
estones provides a convenient shorthand method to communicate overall project pr
ogress to stakeholders.
113
Project
FIGURE 5.6 General Format for WBS
Task 2 Task 3
Task 1
Subtask 1.1
Subtask 1.2
Subtask 1.3
Subtask 2.1
Subtask 3.1
Work package 1.1.1
Work package 2.1.1
Work package 2.1.2
Work package 1.1.2
Work package 1.1.3
NOTE: This type of diagram can be easily generated in Microsoft Word and other M
icrosoft Office products by using the commands Insert Picture Organization Chart
.
The WBS can be developed by the project team itself or with the help of outside
experts who have executed similar projects. At this point in the project, the WB
S is the best estimate of how the project will be executed. It is almost always
inaccurate in some way, so the formal control and change procedures described in
this section are essential to successful project management. After the WBS has
been constructed, the resources required and estimated time for each element mus
t be determined. Estimating the time a task will require is an art. It is best d
one by a team of individuals. Any previous experiences and data can be helpful.
One group process that has proved useful is the program evaluation and review te
chnique (PERT) time estimation. Team members individually estimate the time a ta
sk will take as best, worst, and most likely. After averaging the teams response
for each of the times, the final PERT time estimate is: Best + (4 Most likely) +
Worst 6
Estimated task time =
After a number of meetings, the Riverview Clinic team determined that the generi
c drug project included three major tasks, each with two subtasks, that needed t
o be accomplished to meet the goals of the project. These subtasks are: Develop
a clinical strategy that maintains quality care with the increased use of generi
cs; Develop a system to inform clinicians of approved generics;
114
Update systems to ensure that timely patient medication lists are available to c
linicians; Develop and deploy a staff education plan; Develop a system to monito
r performance; and Develop and begin to use patient education materials. The WBS
is displayed in Figure 5.7. In actuality, these actions represent the higher-le
vel tasks for this project. For a project of this scope to proceed effectively,
many more subtasks, perhaps 50 to 100, would be required; to illustrate the prin
ciples of project management, this WBS has been held to higher-level tasks. It i
s important to note that the time estimate for each task is the total time neede
d to accomplish a task, not the calendar time it will takea three-day task can be
accomplished in three days by one person or in one day by three people. The nex
t step is to determine what resources are needed to accomplish these tasks. Rive
rview Clinic has decided that this project will be accomplished by four existing
employees and the purchase of consulting time from VVHs IT supplier. The individ
uals involved are: Tom Simpson, clinic administrator Dr. Betsey Thompson, family
physician Sally Humphries, RN, nursing supervisor Cindy Tang, billing manager B
ill Onku, IT vendor support consultant
The Project software provides a convenient window to enter these individuals and
their cost per hour. The program also provides higher levels of detail, such as
the hours an individual can devote to the project and actual calendar days when
they will be available. When using clinicians, the project manager should consi
der the revenue per hour from these individuals as opposed to their salaries FIG
URE 5.7 WBS for Riverview Generic Drug Project
Management and administration
Generic drug project
Systems
Training
Develop clinical strategy (10 days)
Develop monitoring system (27 days)
Identify approved generic drugs (22 days)
Supply current patient medication list (33 days)
Train staff (17 days)
Provide patient education (9 days)
115
FIGURE 5.8 Resources for the Riverview Clinic Generic Drug Project
and benefits, because most organizations will lose this revenue if the clinician
has a busy practice. Figure 5.8 shows the Project window for the Riverview staf
f who will work on the generic drug project. A video on ache.org/books/ OpsManag
ement shows how to enter these data; find it in Chapter 5 on the site. Team memb
ers should be clear about their accountability for each task. A functional respo
nsibility chart (sometimes called RASIC) is helpful; the Riverview project RASIC
is displayed in Figure 5.9. The RASIC diagram is a matrix of team members and t
asks from the WBS. For each task, one individual is responsible (R) for ensuring
that the task is completed. Other team members may need to approve (A) the comp
letion of the task. Additional team members may work on the task as well, so the
y are considered support (S.) The obligation to inform (I) other team members he
lps a team communicate effectively. Finally, some team members need to be consul
ted (C) as a task is being implemented.
Scheduling
Network Diagrams and Gantt Charts
Because the WBS was developed without a specific sequence, the next step is sche
duling each task to accomplish the total project. First, the logical order of th
e tasks must be determined. For example, the Riverview project team determined t
hat the system to identify appropriate generic drugs must be developed before th
e training of staff and patients can begin. Other constraints must also be consi
dered in the schedule, including required start or completion dates and resource
availability. Two tools are used to visually display the schedule. The first is
a network diagram that connects each task in precedence order. This is essentia
lly a process map (Chapter 6) where the process is performed only once. However,
network diagrams never have paths that return to the beginning (as happens freq
uently in process maps). A practical way to develop an initial network diagram i
s to place each task on a sticky note and begin arranging them on a
116
FIGURE 5.9 Riverview Clinic Generic Drug Project (RASIC)
WBS Task
Clinic Board of Directors
A
Lead M.D.
Betsey Thompson R
Administrator
Tom Simpson
Project Manager
Sally Humphries
Billing
Cindy Tang I
IT
Bill Onku I
Develop clinical strategy System to identify approved generics Updated medicatio
n lists Patient education Staff education Monitoring system
C
C
A
R
S
R
S
R
I
S
I
S
A
S A
S R R
R C C
I S S
I I C
A
C
NOTE: R = responsible; A = approval; S = support; I = inform; C = consult.
set of flip charts until they meet the logical and date constraints. The tasks c
an then be entered into a project management software system. Figure 5.10 is the
network diagram developed by the team for the Riverview Clinic generic drug pro
ject. This schedule can be entered into Project to generate a similar diagram. A
nother common scheduling tool is the Gantt chart, which lists each task on the l
eft side of the page with a bar indicating the start and end times. The Gantt ch
art for the Riverview project, generated by Project, is shown in Figure 5.11. Ea
ch bar indicates the duration of the task, and the small arrows connecting the b
ars indicate the predecessorsuccessor FIGURE 5.10 Network Diagram for Riverview G
eneric Drug Project
Start
Updated medication lists System to identify approved generics Patient education
Implement
Develop clinical strategy
Staff education
Monitoring system
117
FIGURE 5.11 Riverview Generic Drug Project Gantt Chart
relationship of the tasks. Chapter 5 on the companion web site includes videos o
n how to enter tasks into Project and establish their precedence, which will gen
erate the schedule. Also shown are the many ways in which the schedule and staff
assignments can be viewed in Project. The next step is to assign resources to e
ach task. Figure 5.12 shows how the resources are assigned for each day in the p
roject. Care must be taken when assigning resources, as no person works 100 perc
ent of the time. A practical limit is 80 percent, so it is helpful that Project
generates a resource use graph (Figure 5.13) for each individual. If any single
individual is allocated at more than 80 percent in any time period, the schedule
may need to be adjusted to FIGURE 5.12 Riverview Generic Drug Project, Tasks wi
th Resources Assigned
118
FIGURE 5.13 Resource Use Graph for Tom Simpson
NOTE: As Toms assignment represents more than 80 percent of his time for some wee
ks, the schedule or assignment should be revised.
reduce this allocation. Adjusting the schedule to accommodate this constraint is
known as resource leveling. Go to ache.org/books/OpsManagement, Chapter 5, to see
a video on how to use Project to adjust staff time. A final review of this init
ial schedule is made to assess how many tasks are being performed in parallel (s
imultaneously). A project with few parallel tasks will take longer to accomplish
than one with many parallel tasks. Another consideration may be date constraint
s. Examples include a task that cannot begin until a certain date because of sta
ff availability, or a task that must be complete by a certain date to meet an ex
ternally imposed deadline (e.g., new Medicare billing policy). The Project softw
are provides tools to set these constraints inside the schedule.
Slack and the Critical Path
To optimize a schedule, the project manager must pay attention to slack in the s
chedule and to the critical path. If a task takes three days but does not need t
o be completed for five days, there would be two days of slack. The critical pat
h is the longest sequence of tasks with no slack, or the shortest possible compl
etion time of the project.
119
Slack is determined by the early finish and late finish dates. The early finish
date is the earliest date that a task could possibly be completed, based on the
early finish dates of predecessors. The late finish date is the latest date that
a task can be completed without delaying the finish of the project; it is based
on the late start and late finish dates of successor tasks. The difference betw
een early finish and late finish dates determines the amount of slack. For criti
cal path tasks (which have no slack), the early finish and late finish dates are
identical. Tasks with slack can start later based on the amount of slack they h
ave available. In other words, if a task takes three days and the earliest the t
ask could be completed is day 18, based on its predecessors, and a late finish d
ate of day 30, based on it successors, the slack for this task is 12 days; this
task could start as late as day 27 without affecting the completion date of the
project. The critical path, which determines the duration of a project, is the c
onnected path through a project of critical tasks. Go to Chapter 5 on ache.org/b
ooks/OpsManagement for a video that provides a simple way to understand this con
cept. Calculating slack and the critical path can be complex and time consuming
(see Moder, Phillips, and Davis [1995] for a more detailed explanation). Fortuna
tely, Project does this automatically. In some cases (e.g., a basic clinical res
earch project), it is difficult to estimate the duration of tasks. If a project
has many tasks with high variability in their expected durations, the PERT estim
ating system should be used. PERT employs probabilistic task times to estimate s
lack and critical paths. PERT is good for time estimation prior to the start of
the project, but the critical path method is better suited for project managemen
t once a project has begun. It is not particularly useful to have a range of sta
rt dates for a taskwhat is really important is when a task should have started an
d whether the project is ahead or behind. Although Project provides a PERT sched
uling function, the use of PERT is infrequent in healthcare and beyond the scope
of this book. Figure 5.14 displays a Gantt chart for the Riverview generic drug
project with both slack and critical path calculated.
Crashing the Project
Consider the following scenario. The clinic president has been notified by one h
ealth plan that if the Riverview generic drug project is implemented by March 1,
the clinic will receive a $20,000 bonus. He asks the project manager to conside
r speeding up, or crashing, the project. The term project crashing has negative asso
ciations, as the thought of a computer crashing stirs up dire images. However, a
crashed project is simply one that has been sped up. Crashing a project require
s reducing the length of the critical path, which can be done by any of the foll
owing: Shortening the duration of work on a task on the critical path. Changing
a task constraint to allow for more scheduling flexibility.
120
FIGURE 5.14 Gantt Chart for Riverview Generic Drug Project with Slack and Critic
al Path Calculated
Breaking a critical task into smaller tasks that can be worked on simultaneously
by different resources. Revising task dependencies to allow more scheduling fle
xibility. Setting lead time between dependent tasks where applicable. Scheduling
overtime. Assigning additional resources to work on critical-path tasks. Loweri
ng performance goals (not recommended without strong stakeholder consultation) (
Microsoft Corp. 2003). The scope, time, duration, and performance relationships
need to be considered in a crashed project. A crashed project has a high risk of
costing more than the original schedule predicted, so the formal change procedu
re, discussed in the next section, should be used.
Project Control
It would be convenient if every projects schedule and costs occurred according to
the initial project plan. However, because this is almost never the case, an ef
fective project monitoring and change control system needs to be operating throu
ghout the life of a project.
Monitoring Progress
The first important monitoring element is a system to measure schedule completio
n, cost, and expected performance against the initial plan. Microsoft Project pr
ovides a number of tools to assist the project manager. After the plans initial s
cope document, WBS, staffing, and budget have been determined, they are saved as
the baseline plan. Any changes during the project can be compared to this initial
baseline.
121
On a disciplined time basis (e.g., once per week), the project manager needs to
receive a progress report from each task managerthe individual designated as respo
nsible on the RASIC chart (Figure 5.9)regarding schedule completion and cost. The
enterprise version of Project contains some helpful tools to automate this somet
imes-tedious data-gathering task. Figure 5.15 shows a Project report on the prog
ress of the generic drug project after three weeks. Go to Chapter 5 on ache.org/
books/OpsManagement for a video that shows how to generate progress reports.
Earned Value Analysis
For large projects, earned value analysis provides a comprehensive vehicle to mo
nitor progress. This tool provides a way to combine the monitoring of both sched
ule and cost and is particularly useful in the early stages of a project. The fi
rst step in an earned value analysis is to determine a status date, usually clos
e to the project teams meeting date. Once the status date is determined, three fu
ndamental values are calculated for each task. The first is the budgeted cost of
tasks as scheduled in the project plan, based on the costs of resources that ha
ve been planned to complete the task. Called the budgeted cost of work scheduled
(BCWS), this is the baseline expected cost up to the status date. The actual co
st of completing all or some portion of the task, up to the status date, is the
actual cost of work performed (ACWP). The final variable is the value of the wor
k performed by the status date, measured in currency. This is literally the valu
e earned by the work performed, called the budgeted cost of work performed (BCWP
). BCWP is also called the earned value. For example, assume three people have b
een assigned to a task for a week, and all have been paid at $1,000/week for a t
otal of FIGURE 5.15 Status of Riverview Generic Drug Project at Three Weeks
122
$3,000. However, one of the workers has only accomplished 50 percent of what he
was assigned to do, so the value of his work is only $500. Therefore, $2,500 is
the BCWP, or earned value. In this case, the ACWP is $3,000. Project performance
can be analyzed with earned value analysis (Microsoft Corp. 2003). Assume a tas
k has a budgeted cost (BCWS) of $1,000 and, by the status date, is 40 percent co
mplete. The earned value (BCWP) is $400, but the scheduled value (BCWS) at the s
tatus date is $500. This indicates that the task is behind scheduleless value has
been earned than was planned. Assume the tasks actual cost (ACWP) at the status
date is $600, perhaps because a more expensive resource was assigned to the task
. Therefore, the task is also over budgetmore cost has been incurred than was pla
nned. The earlier in a projects life cycle discrepancies between ACWP, BCWP, and
BCWS can be identified, the sooner steps can be taken to remedy the problem. Ear
ned value analysis is a powerful monitoring tool, especially for large and compl
ex projects. More detail and examples of this technique can be found in Lewis (2
000). The Project program can perform an earned value analysis like that display
ed in Figure 5.16 for the Riverview generic drug project. The BCWP is less than
the BCWS for three tasks, indicating that they are behind schedule. However, the
AWCP is the same as the BWCP, so it does not appear that costs are currently a
problem. Three final columns are included in this report. The first is the proje
ction for the estimate at completion (EAC), which is the expected total cost of
a task or project based on performance as of the status date. EAC is also called
forecast at completion, calculated as: EAC = ACWP + (BAC BCWP)/CPI
The CPI is the cost performance indexthe ratio of budgeted, or baseline, costs of
work performed to actual costs of work performed (BCWP/ACWP). The budget at com
pletion (BAC) shows an estimate of the total project cost. The variance at compl
etion shows the difference between BAC and EAC. In Project, the EAC is the Total
Cost field and the BAC is the Baseline Cost field from the associated baseline.
In addition to the CPI, a schedule performance index (SPI) can be calculated: S
PI = BCWP/BCWS
The two can be combined into one numberthe critical index (CI)which indicates the
overall performance of a project:
123
FIGURE 5.16 Riverview Generic Drug Project Earned Value Analysis
CI = CPI SPI
If the CI dips below 0.8 or above 1.2, the project may be in serious trouble, an
d active intervention is needed by the project team.
Change Control
The project manager should have a status meeting at least once a month, and pref
erably more frequently. At this meeting, the project team should review the actu
al status of the project based on task completion, expenses, personnel utilizati
on, and progress toward expected project outcomes. The majority of time in these
meetings should be devoted to problem solving, not reporting. Once deviations a
re detected, their source and causes must be determined by the team. For major o
r complex deviations, diagnostic tools such as fishbone diagrams (Chapter 6) can
be used. Three courses of action are now available: Ignore the deviation if it
is small, take corrective action to remedy the problem, or modify the plan by us
ing the formal change procedure developed in the project charter and scope docum
ent. One major cause of deviations is an event that occurs outside the project.
The environment will always be changing during a projects execution, and modifica
tions of the projects scope or performance level may be necessary. For example, t
he application of a new clinical breakthrough may take priority over projects th
at improve support systems, or a competitor may initiate a new service that requ
ires a response. Using a formal change mechanism is one of the key characteristi
cs of high-performing project managers. It is human nature to resist communicati
ng a schedule or cost problem to project sponsors and stakeholders. However, the
consequences of this inaction can be significant, if not fatal, to large
124
projects. The change process also forces all parties involved in a project to su
bject themselves to disciplined analysis of options and creates disincentives fo
r scope creep. Changes to the initial plan should be documented in writing and s
igned off on by the project sponsor as appropriate. They should be included in t
he project records (three-ring binders or equivalent). The Riverview project cha
rter (and subsequent scope document) states that changes in plan of less than 10
percent can be made by project manager Sally Humphries. Therefore, she could ad
just the schedule by up to 4.9 days, the cost by up to $6,100, and the performan
ce goal by 0.4 percent. For deviations greater than these amounts, Sally would n
eed the clinic board to review and sign off. The companion web site contains pro
ject change documentation and a sign-off template.
Communications
A formal communication plan was developed as part of scope creation. Communicati
ons to both internal and external stakeholders are critical to the success of a
project. Many communications media can be used, including simple oral briefings,
e-mails, and formal reports. A reasonable contemporary mix used by many organiz
ations is a web-based intranet that contains detailed information on the project
. An e-mail is sent to stakeholders periodically, with a summary progress report
and links back to the web site for more detailed information. A sophisticated c
ommunications plan will be fine-tuned to meet stakeholder needs and interests an
d will communicate only those issues of interest to each stakeholder. As part of
the communications strategy, feedback from stakeholders should always be solici
ted, as changes in the project plan may affect the stakeholder in ways unknown t
o the project manager. The project update communications should contain informat
ion gathered from quantitative reports. At a minimum, these communications shoul
d provide progress against baseline on schedule, cost, scope, and expected perfo
rmance. Any changes to project baseline, as well as to the approval process, sho
uld be noted. Any issues that need resolution, or those that are being resolved,
should also be noted. The expected completion date is always of interest to all
stakeholders and should be a prominent part of any project plan communication.
Risk Management
Comprehensive prospective risk management is another characteristic of successfu
l projects. Like many other aspects of project management, it takes dis-
125
cipline to develop a risk management plan at the beginning of a project and to u
pdate it continuously as the project progresses. A risk is an event that will ca
use the project to have a substantial deviation from planned schedule, cost, sco
pe, or performance. The most direct way to develop a risk management plan is to
begin with the WBS. Each task in the WBS should be assessed for risks, both know
n and unknown. Risks can occur for each task in its performance, duration, or co
st; if a project has 50 tasks, it will have 150 potential risks. A number of tec
hniques can be used to identify risks, but the most straightforward is a brainst
orming exercise by the project team. (Some of the tools found in Chapter 6e.g., m
ind mapping, root cause analysis, force field analysiscould also be used in risk
assessment.) Another useful technique is to interview stakeholders to identify r
isks to the project as viewed from the stakeholders perspective. The organizations
strategic plan is also a resource, as the plan will frequently contain a streng
ths, weaknesses, opportunities, and threats analysis. The Weaknesses and Threats
sections may contain clues as to potential risks to a task within a project. On
ce risks have been identified for each task in the WBS structure, the project te
am should also assign a risk probability to each. Those risks with the highest p
robability, or likelihood, of occurring during the project should be analyzed in
depth and a risk management strategy devised. The failure mode and effects anal
ysis method (Chapter 6) can also be used for a more rigorous risk analysis. A qu
antitative analysis can also be conducted for tasks that have high risk or are c
ritical to project execution. If data can be collected for similar tasks in mult
iple circumstances, probability distributions can be created and used for simula
tion and modeling. An example of this technique would be remodeling space in an
older building. If an organization reviewed a number of recent remodeling projec
ts, it might determine that the average cost per square foot of remodeled space
is $200/foot with a normal distribution. This information could be used as the b
asis of a Monte Carlo simulation (Chapter 10) or as part of a decision tree (Cha
pter 6). The results of these simulations would provide the project manager with
quantitative boundaries on the possible risks associated with the task and proj
ect and would be useful in constructing mitigation strategies. Tasks with the fo
llowing characteristics should be looked at closely, as they may be high risk: H
ave a long duration Are subject to highly variable estimates of duration Depend
on external organizations Require a unique resource (e.g., a physician who is on
call) Are likely to be affected by external governmental or payer policies
126
The management strategy for each identified risk should have three components. F
irst, risk avoidance initiatives should be identified. It is always better to av
oid an adverse event than to have to deal with its consequences. An example of a
risk avoidance strategy is to provide mentoring to a young team member who has
responsibility for key tasks in the project plan. The second element of the risk
management strategy is to develop a mitigation plan. An example of a mitigation
response would be to bring additional people and financial resources to a task.
Another might be to call on the project sponsor to help break an organizational
logjam. Third, a project team may decide to transfer the risk to an insurance e
ntity. This strategy is common in construction projects through the use of bondi
ng for contractors. All identified risks and their management plans should be ou
tlined in a risk register, a listing of each task, identified risks, and prevent
ion and mitigation plans. This risk management plan should be updated throughout
the life of the project. The Riverview project team identified three serious ri
sks, which are listed in Table 5.2 with their mitigation plans.
Quality Management, Procurement, the Project Management Office, and Project Clos
ure
Quality Management
The majority of focus in this chapter has been on managing the scope, cost, and
schedule of a project. The performance, or quality, of an operational project is
the fourth key element in successful project management. In general, quality ca
n be defined as meeting specified performance levels with minimal variation and
waste. TABLE 5.2 Risk Mitigation Plan for Riverview Generic Drug Project
Risk Generic drug use decreases quality
Mitigation Plan Assistance will be sought from VVH hospital pharmacy Pharmaceuti
cal firms Health plans IT vendor has specialists on call who will be flown to Ri
verview Clinic Assistance will be sought from VVH IT department Contingency fund
ing has been earmarked in clinic budget
Computer systems do not work
Software modifications are more expensive than budgeted
127
The fundamental tools for accomplishing these goals are described in Chapters 6
and 8. More advanced techniques for reducing variation in outcomes can be found
in Chapter 8 (quality and Six Sigma), and Chapter 9 discusses tools for waste re
duction (Lean). Throughout the life of a project, the project team should monito
r the expected quality of the final product. Individual tasks that are part of a
quality management function within a project should be created in the WBS. For
example, one task in the Riverside generic drug project is to develop a monitori
ng system. This system will not only track the use of generic drugs but will als
o ascertain whether their use has any negative clinical effects.
Procurement
Many projects depend on outside vendors and contractors, so a procurement system
integrated with an organizations project management system is essential. A purch
asing or procurement department can be helpful in this process. Procurement staf
f will have developed templates for many of the processes described in the follo
wing sections. They will also have knowledge of the latest legal constraints an
organization may face. However, the most useful attribute of the procurement dep
artment may be the frequency with which it executes the purchasing cycle. By per
forming this task frequently, its staff has developed expertise in the process a
nd is aware of common pitfalls to avoid.
Contracting Once an organization has decided to contract with a vendor for a por
tion of a project, three basic types of contracting are available. The fixed pri
ce contract is a lump sum for the performance of specified tasks. Fixed price co
ntracts sometimes contain incentives for early delivery. Cost reimbursement cont
racts provide payment to the vendor based on the vendors direct and indirect cost
s of delivering the service for a specified task. It is important to clearly doc
ument in advance how the vendor will calculate its costs. The most open-ended ty
pe of contract is known as time and materials. Here, the task itself may be poor
ly defined, and the contractor is reimbursed for her actual time, materials, and
overhead. A time and materialstype contract is commonly used for remodeling an o
lder building, where the contractor is not certain of what she will find in the
walls. Great caution and monitoring are needed when an organization uses this ty
pe of contracting. Any contract should contain a statement of work (SOW). The SO
W contains a detailed scope statement, including WBS, for the work that will be
performed by the contractor. It also includes expected quantity and quality leve
ls, performance data, task durations, work locations, and other details that wil
l be used to monitor the work of the contractor.
128
Selecting a vendor Once a preliminary SOW has been developed, the organization w
ill solicit proposals and select a vendor. A useful first step is to issue a req
uest for information (RFI) to as many possible vendors as the project team can i
dentify. The RFIs generate responses from vendors on their products and experien
ce with similar organizations. Based on these responses, the number of feasible
vendors can be reduced to manageable set. A more formal request for proposal (RF
P) can then be issued. The RFP will ask for a detailed proposal, or bid. The fol
lowing criteria should be considered in awarding the contract:
Does the vendor clearly understand the organizations requirements? What is the to
tal cost? Does the vendor have the capability and correct technical approach to
deliver the requested service? Does the vendor have a management approach to mon
itor successful execution of the SOW? Can the vendor provide maintenance or meet
future requirements and changes? Does the vendor provide references from client
s that are similar to the contracting organization? Does the vendor assert intel
lectual or proprietary property rights in the products it supplies?
Payment based on earned value Vendor performance should be monitored and payment
s made based on the contract as described previously. The most sophisticated pay
ment system for contractors employs earned value. This can be a useful tool in m
aking payments, particularly to consultants or IT service vendors. Many of these
vendors will initially request that customers pay them based on the hours their
staff has worked; when using the earned value mechanism, the contractor is paid
only for work actually accomplished. The Riverview generic drug project task of
providing updated patient medication lists is expected to require 132 hours of
Bill Onkus time, at $130/hour. Therefore, the BCWS of this task is $17,160. The p
roject team initially determined that this task would require 48 modifications t
o the existing EHR system. At the three-week status meeting, project manager Sal
ly Humphries reports that the task is 10 percent complete; however, Bill has onl
y completed 2 of the 48 modifications, or 4.1 percent. Although Bills employer se
nt Riverside Clinic an invoice for 10 percent of his time ($1,716), the clinic c
ould reasonably make the case that it would pay only the BCWP, or earned value.
The earned value here is 4.1 percent, or $703.56. Of course, payment based on ea
rned value must be a part of the contract with the vendor.
129
The Project Management Office
Many organizations outside the healthcare industry (e.g., architects, consultant
s) are primarily project oriented. Such organizations have a centralized project
management office (PMO) to oversee the work of their staff. Because healthcare
delivery organizations are primarily operational, the majority do not use this s
tructure. However, departments within large hospitals and clinics, such as IT an
d quality, have begun to use a centralized project office approach. In addition,
some organizations have designated and trained project leaders in Six Sigma or
Lean techniques. These project leaders are assigned from a central PMO. PMOs pro
vide a central structure to monitor progress on all projects in an organization
and reallocate resources as needed when projects encounter problems. They also p
rovide a resource for the training and development of project managers.
Project Closure
A successful project should have an organized closure process, which will includ
e a formal stakeholder presentation and approval process. In addition, the proje
ct sponsor should sign off at project completion to signify that performance lev
els have been achieved and all deliverables have been received. During the close
-out process, special attention should be paid to project staff, who will be int
erested in their next assignment. A disciplined handoff of staff from one projec
t to the next will allow successful completion of the closure process. All docum
ents related to the project should be indexed and stored. This can be helpful if
outside vendors have participated in the project and a contract dispute arises
in the future. Historical documents can also provide a good starting point for t
he next version of a project. The project team should have a final session to id
entify lessons learnedboth good and badin the execution of the project. These less
ons should be included in the project documentation and shared with other projec
t managers within the organization.
The Project Manager and Project Team
The project managers role is pivotal to the success of any project. Selecting, de
veloping, and nurturing high-functioning team members are also critical. The pro
ject managers team skills include both running effective meetings and facilitatin
g optimal dialogue within these meetings.
Team Skills
A project manager can take on many roles within a project. In many smaller healt
hcare projects, the project manager is actually the person who accomplishes
130
many of the project tasks. In other projects, the project managers job is solely
leadership and the management of the many individuals performing the tasks. Slac
k (2005) provides a useful matrix to determine what role a project manager shoul
d assume in projects of varying size (Table 5.3).
Team Structure and Authority
The members and structure of a project team may be selected by the project manag
er, but in many cases are outlined by the project sponsor and other members of s
enior management. It is important to formally document the team makeup and how t
eam members were assigned in the project charter and scope documents. Care shoul
d be taken to avoid overscheduling team members, as all members must have the ne
eded time available to work on the project. A number of key issues need to be ad
dressed as the project team is formed. The most important is the project managers
level of authority to make decisions. Can the project manager commit resources,
or must he ask senior management or department heads each time a new resource i
s needed? Is the budget controlled by the project manager, or does a central fin
ancial authority control it? Finally, is administrative support available to the
team, or do the project team members need to perform these tasks themselves?
Team Meetings
A weekly or biweekly project team meeting is highly recommended to keep a projec
t on schedule. At this meeting, the projects progress can be monitored and discus
sed and actions initiated to resolve deviations and problems. All good team meet
ings include comprehensive agendas and a complete set of minutes. Minutes should
be action oriented (e.g., The schedule slippage for task 17 will be resolved by
assigning additional resources from the temporary pool.). In addition, the indivi
dual accountable for following through on the issue should be identified. If the
meetings deliberations and actions are confidential, everyone on the team should
be aware of the policy and adhere to it uniformly. TABLE 5.3 Project Managers Ro
le Based on Effort and Duration of a Project
Variable
Small Project
Medium Project 3602,400 hours 36 months Manage and do some
Large Project 2,40010,000 hours 612+ months Manage and lead
Effort range 40360 hours Duration 5 days3 months Project leader role Doer with some
help
SOURCE: Slack (2005). Used with permission.
131
The decision-making process should be clear and understood by all team members.
In some situations, all major decisions will be made by the project manager. In
others, team members may have veto power if they represent a major department th
at may need to commit resources. Some major decisions may need to be reviewed an
d approved by individuals external to the project team. The use of data and anal
ytical techniques is strongly encouraged as a part of the decision process. Team
members need to take responsibility for the success of the team. They can demon
strate this behavior by following through on commitments, contributing to the di
scussion, actively listening, and giving and accepting feedback. Everyone on a t
eam should feel that she has a voice, and the project manager needs to lead the
meeting in such a way as to balance the air time between team members. This means
politely and artfully interrupting the wordy team member and summarizing his poi
nt; it also means calling on the silent team member to solicit input. At the end
of a meeting, it is useful to evaluate the meeting itself. The project manager
and team can spend a few minutes reviewing questions such as: Did we accomplish
our purpose? Did we take steps to maintain our gains? Did we document actions, r
esults, and ideas? Did we work together successfully? Did we share our results w
ith others? Did we recognize everyones contribution and celebrate our achievement
s?
Dialogue
The core of the project team meeting is the conversation that occurs among team
members. A productive team will engage in dialogue, not unfocused discussion or
debate. The study of dialogue has advanced, and much is now known about how effe
ctive dialogue fuels productive teams. Yankelovich (1999) identifies three disti
nctive features that differentiate productive dialogue from mere discussion. The
first is equality within the team and the absence of coercive influences. As mo
st project teams are made up of members from many parts of an organization, equa
lity and camaraderie are important to make team meetings enjoyable and productiv
e. A second feature of productive dialogue is the ability of team members to lis
ten with empathy. The ability to think like another member of a team helps broad
en the dialogue and keeps any disagreement focused on issues rather than persona
lities. The third key feature of productive dialogue is the ability to bring ass
umptions into the open. Once assumptions are on the floor, they can be considere
d with respect, debated, and resolved honestly.
132
Losada and Heaphy (2004) provide a specific formulation for the content of dialo
gue in high-performing teams. They studied project team interactions in detail o
ver many meetings and rated their performance based on the metrics of profitabil
ity, customer satisfaction, and 360-degree reviews. The teams were then ranked a
s high, medium, and low performers. Those teams with the highest performance wer
e those with high Nexi. Losada and Heaphy (2004) define Nexi as strong and sustain
ed interlocking behaviors by team members. In other words, a conversation with c
risp interactions between team members clearly focused on a team goal produced h
igh Nexi. The relationship of Nexi to team performance is displayed in Figure 5.
17. Three other characteristics were also measured: inquiry versus advocacy, ext
ernal versus internal viewpoint, and positive versus negative comments. Low-perf
orming teams were characterized by high levels of advocacy (Let me convince you w
hy my idea is so good) versus inquiry (Can you please explain your idea in more de
pth?). High-performing teams had an equal mix of advocacy and inquiry. Low-perfor
ming teams had a high degree of internal focus (How can we do that with our curre
nt IT department?), whereas high-performing teams mixed in an equal blend of exte
rnal environmental review (What are similar firms doing with their IT function?).
Finally, optimism and good cheer triumphed in this study. Highperforming teams h
ad a 6:1 ratio of positive (Great idea!) to negative comments (We tried that before
and it didnt work). Figure 5.18 demonstrates the linkage of these three dialogue
dimensions to the connecting Nexi for high-performing teams. FIGURE 5.17 Relatio
nship of Nexi to Team Performance
35 30 Number of Nexi per 50-minute meeting 25 20 15 10 5 0 Low Medium High Perfo
rmance: Profitability, customer satisfaction, 360 assessments Connectivity
SOURCE: Losada, M., and E. Heaphy. 2004. The Role of Positivity and Connectivity
in the Performance of Business Teams. American Behavioral Scientist 47 (6): 748.
Copyright 2008 by Sage Publications, Inc. Reprinted with permission of Sage Publ
ications, Inc.
133
Performance
FIGURE 5.18 Dialogue Attributes of HighPerforming Teams
Connectivity
Inquiry Advocacy
External Internal
Positivity Negativity
SOURCE: Losada, M., and E. Heaphy. 2004. The Role of Positivity and Connectivity
in the Performance of Business Teams. American Behavioral Scientist 47 (6): 752.
Copyright 2008 by Sage Publications, Inc. Reprinted with permission of Sage Publ
ications, Inc.
Leadership
Although this book is not primarily concerned with leadership, clearly the proje
ct manager must be able to lead a project forward. Effective project leadership
requires the following skills: The ability to think critically using complex inf
ormation; The strategic capability to take a long-term view of the organization;
A systems view of the organization and its environment (Chapter 1); The ability
to create and lead change; An understanding of oneself to permit positive inter
actions, conflict resolution, and effective communication; The ability to mentor
and develop employees into high-performing teams; and The ability to develop a
performance-based culture. Two widely known models are relevant in healthcare en
vironments and provide a foundation for developing these skills. Kotter (1996) d
escribes an eightstep method to lead change in an organization, with an emphasis
on employee engagement and communication. Heifetz (1994) presents a different,
complementary view of leadership. He distinguishes between routine technical pro
blems, which can be solved through expertise, and adaptive problems, such as maj
or health system changes, which require innovative approaches. His strategies fo
r confronting these problems include a focus on values, an understanding of the
stress caused by adaptive problems, and inclusion of stakeholders in solutions.
134
Conclusion
This chapter provides a basic introduction to the science and discipline of proj
ect management. The field is finding a home in healthcare IT departments and has
a history in construction projects. Successful healthcare organizations of the
future will use this rigorous methodology to make significant changes and improv
ements throughout their operations.
1. Who should constitute members of the project team, key stakeholders, and proj
ect sponsors in a clinical project in a physicians office? Support your choices.
In a hospital? Again, support your choices. 2. Identify five common risks in hea
lthcare clinical projects and develop contingency responses for each.
Chapter Exercises
1. Download the project charter and project schedule from ache.org/books/OpsMana
gement. Complete the missing portions of the charter. Develop a risk assessment
and mitigation plan. Add tasks to the schedule for those areas that require more
specificity. Apply resources to each task, determine the critical path, and det
ermine a method to crash the project to reduce its total duration by 20 percent.
2. Review the Institute for Healthcare Improvement web site and identify and se
lect a quality improvement project (www.ihi.org/IHI/Results/ ImprovementStories/
). Although you will not know all the details of the organization that executed
this project, create a charter document for your chosen project. 3. For the proj
ect identified above, create a feasible WBS and project schedule. Enter the sche
dule into Microsoft Project. 4. Conduct a meeting with colleagues. Have an obser
ver measure Nexi, advocacy and inquiry comments, internal and external comments,
and positive and negative comments. Develop a plan to move toward higher perfor
mance based on the research findings of Losada and Heaphy.
References
Austin, C. J., and S. B. Boxerman. 2003. Information Systems for Health Care Man
agement, 6th ed. Chicago: Health Administration Press.
135
Connolly, C. 2005. Cedars-Sinai Doctors Cling to Pen and Paper. Washington Post Ma
rch 21, A01. Gapenski, L. C. 2005. Health Care Finance: An Introduction to Accou
nting and Financial Management. Chicago: Health Administration Press. Heifetz, R
. A. 1994. Leadership Without Easy Answers. Boston: Belknap Press. Kotter, J. P.
1996. Leading Change. Boston: Harvard Business School Press. Lewis, J. P. 2000.
The Project Managers Desk Reference, 2nd ed. Boston: McGraw Hill. Losada, M., an
d E. Heaphy. 2004. The Role of Positivity and Connectivity in the Performance of
Business Teams. American Behavioral Scientist 47 (6): 74065. Microsoft Corp. 2003.
Microsoft Office Project Professional, Help Screens. Redmond, WA: Microsoft Cor
p. Moder, J. J., C. R. Phillips, and E. W. Davis. 1995. Project Management with
CPM, Pert & Precedence Diagramming, 3rd ed. Middleton, WI: Blitz Publishing. Mos
ser, G. 2005. Personal communication, July 10. Niven, P. R. 2005. Balanced Score
card Diagnostics: Maintaining Maximum Performance. New York: John Wiley & Sons. . 2
002. Balanced Scorecard Step-by-Step: Maximizing Performance and Maintaining Res
ults. New York: John Wiley & Sons. Project Management Institute. 2004. A Guide t
o the Project Management Body of Knowledge. Newtown Square, PA: Project Manageme
nt Institute. Slack, M. P. 2005. Personal communication, August 15. Yankelovich,
D. 1999. The Magic of Dialogue: Transforming Conflict into Cooperation, 2nd ed.
New York: Simon and Schuster.
PART
III
performance improvement tools, techniques, and programs
CHAPTER
6
TOOLS FOR PROBLEM SOLVING AND DECISION MAKING
CHAPTER OUTLINE
Operations Management in Action Overview Decision-Making Framework Framing Mappi
ng Techniques Mind Mapping Process Mapping Measures of process performance Activ
ity and Role Lane Mapping Service Blueprinting Problem Identification Tools Root
-Cause Analysis Five whys technique Cause-and-effect diagram Failure Mode and Ef
fects Analysis Theory of Constraints Analytical Tools Optimization Linear progra
mming Sensitivity analysis Decision Analysis Implementation: Force Field Analysi
s Conclusion Discussion Questions Chapter Exercises References
138

activity mapping advanced access scheduling capacity utilization cause-and-effec
t diagram cycle time decision analysis decision tree define-measure-analyze-impr
ovecontrol (DMAIC) diagnosis-related group (DRG) failure mode and effects analys
is (FMEA) fishbone diagram, see cause-andeffect diagram five whys flowchart forc
e field analysis Institute for Healthcare Improvement (IHI) Ishikawa diagram, se
e cause-andeffect diagram Joint Commission (formerly the Joint Commission on Acc
reditation of Healthcare Organizations) Lean linear programming mathematical pro
gramming Microsoft Excel Solver mind mapping optimization Palisade Precision Tre
e Pareto analysis Plan-do-check-act (PDCA) process mapping risk analysis root-ca
use analysis (RCA) sensitivity analysis service blueprinting Six Sigma slack var
iables theory of constraints (TOC) throughput time value-added time
139

At Allegheny General Hospital in Pittsburgh, the two intensive care units had be
en averaging about 5.5 infections per 1,000 patient days, mostly bloodstream inf
ections from catheters. That infection rate was a bit higher than the Pittsburgh
average but a bit lower than the national average, says Dr. Richard Shannon, ch
airman of medicine at Allegheny General. Over the prior 12 months, 37 patients,
already some of the sickest people in the hospital, had 49 infections. Of those,
51 percent died. Dr. Shannon and the staff in the two unitsdoctors, residents, a
nd nursesapplied the Toyota root-cause analysis system, investigating each new in
fection immediately. Their main conclusion was that femoral intravenous lines, i
nserted into an artery near the groin, had a particularly high rate of infection
s. The team made an all-out effort to replace these lines with less risky ones i
n the arm or near the collarbone. Dr. Shannon, who oversees the two units, gave
the directive to keep femoral lines to an absolute minimum. The result was a 90
percent decrease in the number of infections after just 90 days of using the new
procedures.
SOURCE: Adapted from Wysocki 2004.
Overview
This chapter introduces the basic tools associated with problem solving and deci
sion making. Much of the work of healthcare professionals is just that making dec
isions and solving problemsand in an ever-changing world that work must be accomp
lished well and quickly. A structured approach can enable efficient, effective p
roblem solving and decision making. Major topics in this chapter include: The de
cision-making process, with a focus on framing the problem or issue; Mapping tec
hniques, including mind mapping, process mapping, activity mapping, and service
blueprinting; Problem identification tools, including root-cause analysis (RCA),
failure mode and effects analysis (FMEA), and the theory of constraints (TOC);
Analytical tools such as optimization using linear programming and decision anal
ysis; and Force field analysis to address implementation issues. 140
C h a p t e r 6 : To o l s f o r Pro b l e m S o l v i n g a n d D e c i s i o n
M a k i n g
141
This chapter gives readers a basic understanding of various problemsolving tools
and techniques to enable them to: Frame questions or problems; Analyze the prob
lem and various solutions to it; and Implement those solutions. The tools and te
chniques outlined in this chapter should provide a basis for tackling difficult,
complicated problems.
Decision-Making Framework
A structured, rational approach to problem solving and decision making includes
the following steps: Identifying and framing the issue or problem; Generating or
determining possible courses of action and evaluating those alternatives; Choos
ing and implementing the best solution or alternative; and Reviewing and reflect
ing on the previous steps and outcomes. Decision Traps: The Ten Barriers to Bril
liant Decision-Making and How to Overcome Them (Russo and Schoemaker 1989) outli
nes these steps (Table 6.1) and the barriers encountered in decision making (Box
6.1). The plan-do-check-act process for continuous improvement (Chapters 8 and
9), the define-measure-analyze-improve-control process of Six Sigma (Chapter 8),
and the outline for analysis (Chapter 11) all follow the same basic steps as ou
tlined in the decision-making process. The tools and techniques found in this bo
ok can be used to help in gathering the right information to make optimal decisi
ons and learn from those decisions, as well as in the process of making those de
cisions. Often, the learning step in the decision-making process is neglected, b
ut it should not be. It is important to evaluate and analyze both the decision m
ade and the process(es) used in coming to the decision to ensure learning and en
able continuous improvement.
Framing
The frame of a problem or decision encompasses the assumptions, attitudes, and p
reconceived limits that an individual or a team brings to the analyses. These as
sumptions can limit the ability to solve the problem by reducing or eliminating
creativity and causing the decision maker(s) to overlook possibilities. Alternat
ively, these assumptions can aid in problem solving by eliminating wildly improb
able paths, but they usually hinder finding the best solution or finding a possi
ble solution.
142
TABLE 6.1 Decision Elements and Activities
Framing
Typical amount of time: 5%
Recommended amount of time: 20% Gathering intelligence Typical amount of time: 4
5%
Recommended amount of time: 35% Coming to conclusions
Typical amount of time: 40%
Recommended amount of time: 25% Learning from feedback
Typical amount of time: 10% Recommended amount of time: 20%
Structuring the question. This means defining what must be decided and determini
ng in a preliminary way what criteria would cause you to prefer one option over
another. In framing, good decision makers think about the viewpoint from which t
hey and others will look at the issue and decide which aspects they consider imp
ortant and which they do not. Thus, they inevitably simplify the world. Seeking
both the knowable facts and the reasonable estimates of unknowables that you will
need to make the decision. Good decision makers manage intelligence gathering wi
th deliberate effort to avoid such failings as overconfidence in what they curre
ntly believe and the tendency to seek information that confirms their biases. So
und framing and good intelligence dont guarantee a wise decision. People cannot c
onsistently make good decisions using seat-of-the-pants judgment alone, even wit
h excellent data in front of them. A systematic approach forces you to examine m
any aspects and often leads to better decisions than hours of unorganized thinki
ng would. Everyone needs to establish a system for learning from the results of
past decisions. This usually means keeping track of what you expected would happ
en, systematically guarding against self-serving explanations, then making sure
you review the lessons your feedback has produced the next time a similar decisi
on comes along.
SOURCE: Decision Traps by J. Edward Russo and Paul J. H. Schoemaker, copyright 1
989 by J. Edward Russo and Paul J. H. Schoemaker. Used by permission of Doubleda
y, a division of Random House, Inc.
Millions of dollars and working hours are wasted in finding solutions to the wro
ng problems. An ill-defined problem or mistaken premise can eliminate promising
solutions before they can even be considered. People tend to identify convenient
problems and find solutions that are familiar to them rather than looking more
deeply. People also have a tendency to want to do something; quick and decisive
action is seen as necessary in todays rapidly changing environment. Leaping to th
e solutions before taking the time to properly frame the problem will usually re
sult in suboptimal solutions.
M a k i n g
143
Framing the Question Plunging inBeginning to gather information and reach conclus
ions without first taking a few minutes to think about the crux of the issue your
e facing. Frame blindnessSetting out to solve the wrong problem because you have
created a mental framework for your decision with little thought, which causes y
ou to overlook the best options or lose sight of important objectives. Lack of f
rame controlFailing to consciously define the problem in more ways than one or be
ing unduly influenced by the frames of others. Gathering Intelligence Overconfid
ence in your judgmentFailing to correct key factual information because you are t
oo sure of your assumptions and opinions. Shortsighted shortcutsRelying inappropr
iately on rules of thumb, such as implicitly trusting the most readily available i
nformation or anchoring too much on convenient facts. Coming to Conclusions Shoo
ting from the hipBelieving you can keep straight in your head all the information
youve discovered, and therefore you wing it rather than following a systematic pro
cedure. Group failureAssuming that with many smart people involved, good choices
will follow automatically, and therefore you fail to manage the group decision p
rocess. Learning/Failing to Learn from Feedback Fooling yourself about feedbackFa
iling to interpret the evidence from past outcomes for what it really says, eith
er because youre protecting your ego or because you are tricked by hindsight. Not
keeping trackAssuming that experience will make its lessons available automatica
lly, and therefore you fail to keep systematic records to track results of your
decisions and fail to analyze these results in ways that will reveal their true
lessons. Failure to audit your decision processYou fail to create an organized ap
proach to understanding your own decision making, so you remain constantly expos
ed to all the aforementioned mistakes.
SOURCE: Decision Traps by J. Edward Russo and Paul J. H. Schoemaker, copyright 1
989 by J. Edward Russo and Paul J. H. Schoemaker. Used by permission of Doubleda
y, a division of Random House, Inc.
BOX 6.1 The Ten Barriers to Brilliant Decision Making and the Key Elements They
Fall Into
144
Potential solutions to any problem are directly related to, and limited by, the
definition of the problem itself. However, framing the problem can be difficult,
as it requires an understanding of the problem. If the problem is well understo
od, the solution is more likely to be obvious; when framing a problem, it is the
refore important to be expansive, solicit many different viewpoints, and conside
r many possible scenarios, causes, and solutions. The tools outlined in this cha
pter are designed to help with this process.
Mapping Techniques
Mind Mapping
Tony Buzan is credited with developing the mind-mapping technique (Buzan 1991; B
uzan and Buzan 1994). Mind mapping develops thoughts and ideas in a nonlinear fa
shion and typically uses pictures or phrases to organize and develop thoughts. I
n this structured brainstorming technique, ideas are organized on a map and the co
nnections between them are made explicit. Mind mapping can be an effective techn
ique for problem solving because it is not necessary to think linearly. Making c
onnections that are not obvious or linear can lead to innovative solutions. Mind
mapping starts with the issue to be addressed placed in the center of the map.
Ideas on causes, solutions, and so on radiate from the central theme. Questions
such as who, what, where, why, when, and how are often helpful for problem solvi
ng. Figure 6.1 illustrates a mind map related to high accounts receivables.
Process Mapping
A process map, or flowchart, is a graphic depiction of a process showing inputs,
outputs, and steps in the process. Depending on the purpose of the map, it can
be high level or detailed. Figure 6.2 shows a high-level process map for the Riv
erview Clinic, and Figure 6.3 shows a more detailed map of the check-in process
at the clinic. A process map gives a clear picture of what activities are carrie
d out as part of the process, where the activity happens, and how activities are
performed. Typically, process maps are used to understand and optimize a proces
s. The process is commonly charted from the viewpoint of the material, informati
on, or customer being processed (often the patient in healthcare) or the worker
carrying out the work. Process mapping is one of the seven basic quality tools a
nd an integral part of most improvement initiatives, including Six Sigma, Lean,
balanced scorecard, RCA, and FMEA. The steps for creating a process map or flowc
hart are: 1. Assemble and train the team. The team should consist of people from
all areas and levels within the process of interest to ensure that the real pro
cess is captured.
M a k i n g
145
Doctor coding
Incorrect coding
Procedure not covered
FIGURE 6.1 Mind Map: High Accounts Receivables
Data entry error
Procedure not medically necessary Incorrect information
Electronic medical records
Documentation problems
Insufficient information Slow payment or no payment No insurance
Complicated system
Slow billing
Claim denied
New computer systems
High accounts receivable
Type of insurance
Medicare/ Medicaid
Funding
Missing revenue
Private insurance
Identify and fix systematic problems
NOTE: Diagram created in Inspiration by Inspiration Software, Inc.
2. Determine the boundaries of the process (where does it start and end?) and th
e level of detail desired. The level of detail desired, or needed, will depend o
n the question or problem the team is addressing. 3. Brainstorm the major proces
s tasks and subtasks. List them and arrange them in order. (Sticky notes are oft
en helpful here.)
146
FIGURE 6.2 Riverview Clinic HighLevel Process Map
Patient arrives
Patient check-in Wait
Move to examining room
Nurse does preliminary exam
Wait
Physician exam and consultation
Wait
Visit complete
FIGURE 6.3 Riverview Clinic Detailed Process Map: Patient Check-in
Patient arrives
Line? Wait No Patient Existing type New Medical information Information Same HIP
AA on file No HIPAA forms
Yes
Changed Insurance information
Wait
Move to examining room
4. Create a formal chart. Once an initial flowchart has been generated, the char
t can be formally drawn using standard symbols for process mapping (Figure 6.4).
This can be done most efficiently using software such as Microsoft Visio. Note
that when first developing a flowchart it is more important to obtain an accurat
e picture of the process than to worry about using the correct symbols. 5. Make
corrections. The formal flowchart should be checked for accuracy by all relevant
personnel. Often, inaccuracies will be found in the flowchart that need to be c
orrected in this step. 6. Determine additional data needs. Depending on the purp
ose of the flowchart, data may need to be collected, or more information may nee
d to be added. Often, data on process performance are collected and added to the
flowchart.
M a k i n g
147
An oval is used to Block arrows show inputs/outputs are used to show to the proc
ess or start/ transports. end of the process.
A rectangle is used to show a task or activity.
FIGURE 6.4 Standard Flowchart Symbols
Feedback loop A diamond is used to show those points in the process where a choi
ce can be made or alternative paths can be followed.
Arrows show the direction of flow of the process.
D shapes are used to show delays.
End
A triangle shows inventory. For services, they can also indicate customer waitin
g.
Measures of process performance Measures of process performance include throughp
ut time, cycle time, and percentage of value-added time (Chapter 9). Another imp
ortant measure of process, subprocess, task, or resource performance is capacity
utilization. Capacity is the maximum possible amount of output (goods or servic
es) that a process or resource can produce or transform. Capacity measures can b
e based on outputs or on the availability of inputs. For example, if a hospital
food service can provide, at most, 1,000 meals in one day, the food service has
a capacity of 1,000 meals/day. If all magnetic resonance images (MRIs) take one
hour to perform, the MRI machine would have a capacity of 24 MRIs/day. The choic
e of appropriate capacity measure varies with the situation. Ideally, demand and
capacity are perfectly matched. If demand is greater than capacity, some custom
ers will not be served. If capacity is greater than demand, resources will be un
derutilized. In reality, perfectly matching demand and capacity can be difficult
to accomplish because of fluctuations in demand. In a manufacturing environment
, inventory can be used to compensate for demand fluctuations. In a service envi
ronment, this is not possible; therefore, excess capacity or a flexible workforc
e is often required to meet demand fluctuations. Advanced-access scheduling (Cha
pters 9 and 12) is one way for healthcare operations to more closely match capac
ity and demand.
148
Capacity utilization is the percentage of time that a resource (worker, equipmen
t, space, etc.) or process is actually busy producing or transforming output. If
the hospital food service only provides 800 meals/day, the capacity utilization
is 80 percent. If the MRI machine only operates 18 hours/day, the capacity util
ization is 75 percent (18/24 100). Generally, higher capacity utilization is bet
ter, but caution must be used in evaluation. If the hospital food service had a
goal of 95 percent capacity utilization, it could meet that goal by producing 95
0 meals/day, even if only 800 meals/day were actually consumed and 150 meals wer
e discarded. Obviously, this would not be an effective use of resources, but foo
d service would have met its goal. Typically, the more costly the resource, the
greater the importance of maximizing capacity utilization. For example, in a hos
pital emergency department the most costly resource is often the physician, and
other resources (nurses, housekeeping, clerical staff, etc.) are less expensive.
In this case, maximizing the utilization of the physicians is more important th
an maximizing the utilization of the other resources. It is often more economica
l to underutilize less expensive resources to maximize the utilization of more e
xpensive resources. Simulation (Chapter 11) can be used to help determine the mo
st effective use of various types of resources.
Activity and Role Lane Mapping
Activity and role lane mapping can be a useful exercise to include in process ma
pping. List the process activities and the roles involved, and ask who performs
the activity now. Then, take the role out of the activity so that nurse records v
ital signs becomes record vital signs and record this as is shown in Table 6.2. Thi
s type of analysis can help with process redesign and streamlining.
Service Blueprinting
Service blueprinting (Shostack 1984) is a special form of process mapping (as is
value stream mapping, covered in Chapter 9). Service blueprinting begins TABLE
6.2 Activity and Role Lane Mapping
Role Activity Take insurance information Move patient Record vital signs Take hi
story Examine patient Write pathology request Deliver pathology request Clerk x
x x x x x x x x Nurse Porter Doctor
x
M a k i n g
149
with mapping the process from the point of view of the customer. The typical pur
pose of a service blueprint is to identify points where the service might fail t
o satisfy the customer and then redesign or add controls to the system to reduce
or eliminate the possibility of failure. The service blueprint separates onstag
e actions (those visible to the customer) and backstage actions and support proc
esses (those not visible to the customer). A service blueprint specifies the lin
e of interaction, where the customer and service provider come together, and the
line of visibility, that is, what the customer sees or experiences, the tangibl
e evidence that influences perceptions of the quality of service (Figure 6.5).
Problem Identification Tools
Root-Cause Analysis
Root-cause analysis (RCA) is a generic term used to describe structured, step-by
-step techniques for problem solving. It aims to determine and correct the ultim
ate cause(s) of a problem, not just the visible symptoms, to ensure that the pro
blem does not happen again. RCA consists of determining what happened, why it ha
ppened, and what can be done to prevent it from happening again. The Joint Commi
ssion (2005b) requires all accredited organizations to conduct an RCA of any sen
tinel event (an unexpected occurrence involving death or serious physical or psy
chological injury, or the risk thereof) and provides some tools (Joint Commissio
n 2005a) to help an organization conduct that analysis. These tools are not only
useful for sentinel events and Joint Commission requirements, but they also pro
vide a framework for any RCA. A variety of commercial software is also available
for RCAs. FIGURE 6.5 Service Blueprint
Customer actions
Customer gives prescription to clerk Clerk enters data Clerk gives prescription
to pharmacist Pharmacist fills prescription
Line of interaction
Customer receives medicine Clerk gives medicine to customer
Onstage actions
Clerk retrieves medicine Pharmacist gives medicine to clerk
Backstage actions
Line of visibility
150
Although an RCA can be done in many different ways, it is always based on asking
why something happened, again and again, until the ultimate cause is found. Typ
ically, something in the system or process, rather than human error, is found to
be the ultimate cause. The five whys technique and cause-and-effect diagrams ar
e examples of tools used in RCA.
Five whys technique The five whys technique is simple but powerful. It consists
of asking why the condition occurred, noting the answer, and then asking why for
each answer over and over (five times is a good guide) until the root causes are
identified. Often, the reason for a problem is only a symptom of the real cause t
his technique can help eliminate the focus on symptoms, discover the root cause,
and point the way to eliminating the root cause and ensuring that the problem d
oes not occur again. For example:
A patient received the wrong medication. Why? The doctor prescribed the wrong me
dication. Why? There was information missing from the patients chart. Why? The pa
tients most recent lab test results had not been entered into the chart. Why? The
lab technician sent the results, but they were in transit and the patients recor
d had not been updated. The root cause here is the time lag between the test and
data entry. Identifying the root cause leads to different possible solutions to
the problem than simply concluding that the doctor made a mistake. The system c
ould be changed to increase the speed with which lab results are recorded or, at
least, a note could be made on the chart that lab tests have been ordered but n
ot yet recorded.
Cause-and-effect diagram Using only the five whys technique for an RCA can be li
miting. There is an assumption that the effect is a result of a single cause at
each level of why. Often, a set of causes is related to an effect. A cause-and-e
ffect diagram can overcome these limits. One of the seven basic quality tools, t
his type of graphic is used to explore and display all of the potential causes o
f a problem. The cause-andeffect diagram is sometimes called an Ishikawa diagram
(after its inventor, Kaoru Ishikawa [1985]) or a fishbone diagram (because it l
ooks like the skeleton of a fish). Typically, a team uses a cause-and-effect dia
gram to investigate and eliminate a problem. The problem should be stated or fra
med as clearly as possible, including who is involved and where and when the pro
blem
M a k i n g
151
Service (Four Ps) Policies Procedures People Plant/technology
Manufacturing (Six Ms) Machines Methods Materials Measurements Mother nature (en
vironment) Manpower (people)
BOX 6.2 Typical Causeand-Effect Diagram Categories
occurs, to ensure that everyone on the team is attempting to solve the same prob
lem. The problem, or outcome of interest, is the head of the fish. The rest of the
fishbone consists of a horizontal line leading to the problem statement and sev
eral branches, or fishbones, vertical to the main line. The branches represent dif
ferent categories of causes. The categories chosen may vary according to the pro
blem, but there are some common choices (Box 6.2). Possible causes are attached
to the appropriate branches. Each possible cause is examined to determine if the
re is a deeper cause behind it (stage c in Figure 6.6); subcauses are attached a
s more bones. In the final diagram, causes are arranged according to relationshi
ps and distance from the effect. This can help in identifying areas to focus on
and comparing the relative importance of different causes. Cause-and-effect diag
rams can also be drawn as tree diagrams. From a single outcome, or trunk, branch
es extend to represent major categories of inputs or causes that create that sin
gle outcome. These large branches then lead to smaller and smaller branches of c
auses all the way down to twigs at the ends. A process-type cause-and-effect dia
gram (Figure 6.7) can be used to investigate causes of problems at each step in
a process. A process RCA is similar to a Failure Mode and Effect Analysis (FMEA)
, but less quantitative in nature. An example from Vincent Valley Hospital and H
ealth System (VVH) illustrates the cause-and-effect diagramming process. The hos
pital identified excessive waiting time as a problem, and a team was assembled t
o address the issue. The problem is placed in the head of the fish, as shown in
Figure 6.6, stage a. Next, branches are drawn off the large arrow representing t
he main categories of potential causes. Typical categories are shown in Figure 6
.6, stage b, but the chosen categories should suit the particular situation. The
n, all of the possible causes inside each main category are identified. Each of
these causes should be thoroughly explored to identify the causes of causes. Thi
s process continues, branching off into more and more causes of causes, until ev
ery possible cause has been identified (stage c in Figure 6.6).
152
FIGURE 6.6 Cause-andEffect Example
(A) Waiting time
(B) Mother nature (environment) Methods
Waiting time Manpower (people)
Machines
(C)
Mother nature (environment) Transport arrives late
Methods Disorganized files Excessive paperwork HIPAA regulations Bureaucracy Wai
ting time Incorrect referrals
Lack of treatment rooms Old innercity building
Corridor blocked Poor scheduling Lack of technology Process takes too long
Poor maintenance Unexpected patients Elevators broken Original appointment misse
d Wheelchairs unavailable Machines Wrong patients Manpower (people)
Sick Staff not available Late
Much of the value gained from building a cause-and-effect diagram comes from goi
ng through the exercise with a team of people. A common and deeper understanding
of the problem develops, enabling ideas for further investigation. Once the cau
se-and-effect diagram is complete, an assessment of the possible causes and thei
r relative importance should be made. Obvious, easily fixable causes can be deal
t with quickly. Data may need to be collected to assess the more complex possibl
e causes and solutions. A Pareto analysis (Chapter 7) of the various causes is o
ften used to separate the vital few from the trivial many. Building a cause-andeffect diagram is not necessarily a onetime exercise. The diagram can be used as
a working document and updated as more data are collected and various solutions
are tried.
M a k i n g
153
No forms
Pager does not work Technician unavailable
FIGURE 6.7 Process-Type Cause-andEffect Diagram
Wrong test Order in wrong place Doctor orders test
Dispatcher busy Secretary calls dispatcher Wrong information Dispatcher sends to
technician Long time to obtain test results
Phone busy Undecipherable handwriting
Failure Mode and Effects Analysis
The failure mode and effects analysis (FMEA) process was originally developed by
the U.S. military in the late 1940s, originally aimed at equipment failure. Mor
e recently, FMEA has been adopted by many service industries, including healthca
re, to evaluate process failure. Hospitals accredited by the Joint Commission ar
e required to conduct at least one FMEA or similar proactive analysis annually (
Joint Commission 2001). Where RCA is used to examine the underlying causes of a
particular event or failure, FMEA is used to identify the ways in which a proces
s (or piece of equipment) might potentially fail, and its goal is eliminating or
reducing the severity of such a potential failure. A typical FMEA consists of t
he following steps: 1. Identify the process to be analyzed. Typically, this is a
principal process for the organization. 2. Assemble and train the team. Process
es usually cross functional boundaries; therefore, the analysis should be perfor
med by a team of relevant personnel. No one person or persons from a single func
tional area will have the knowledge needed to perform the analysis. 3. Develop a
detailed process flowchart, including all steps in the process. 4. Identify eac
h step (or function) in the process. 5. Identify potential failures (or failure
modes) at each step in the process. Note that there may be more than one potenti
al failure at each step. 6. Determine the worst potential consequence (or effect
) of each possible failure.
154
7. Identify the cause(s) (contributory factors) of each potential failure. An RC
A can be helpful in this step. Note that there may be more than one cause for ea
ch potential failure. 8. Identify any failure controls that are currently present.
A control reduces the likelihood that causes or failures will occur, reduces th
e severity of an effect, or enables the occurrence of a cause or failure to be d
etected before it leads to the adverse effect. 9. Rate the severity of each effe
ct (on a scale of 1 to 10, with 10 being the most severe). This rating should re
flect the impact of any controls that reduce the severity of the effect. 10. Rat
e the likelihood (occurrence score) that each cause will occur (on a scale of 1
to 10, with 10 being certain to occur). This rating should reflect the impact of
any controls that reduce the likelihood of occurrence. 11. Rate the effectivene
ss of each control (on a scale of 1 to 10, with 1 being an error-free detection
system). 12. Multiply the three ratings by one another to obtain the risk priori
ty number (RPN) for each cause or contributory factor. 13. Use the RPNs to prior
itize problems for corrective action. All causes that result in an effect with a
severity of 10 should be high on the priority list, regardless of RPN. 14. Deve
lop an improvement plan to address the targeted causes (who, when, how assessed,
etc.). Figure 6.8 shows an example FMEA for patient falls from the Institute fo
r Healthcare Improvement (IHI). IHI has an online interactive tool for FMEA and
offers many real-world examples that can be used as a basis for FMEAs in other o
rganizations (IHI 2005). The Veterans Administration National Center for Patient
Safety (2006) has developed a less complex FMEA process based on rating only th
e severity and probability of occurrence and using the resulting number to prior
itize problem areas.
Theory of Constraints
The theory of constraints (TOC) was first described in the business novel The Go
al (Goldratt and Cox 1986). The TOC maintains that every organization is subject
to at least one constraint that limits it from moving toward or achieving its g
oal. For many organizations, the goal is to make money now as well as in the fut
ure. Some healthcare organizations may have a different, but still identifiable,
goal. Eliminating or alleviating the constraint can enable the organization to
come closer to its goal. Constraints can be physical (e.g., the capacity of a ma
chine) or nonphysical (e.g., an organizational procedure). Five steps are involv
ed in the TOC: 1. Identify the constraint or bottleneck. What is the limiting fa
ctor stopping the system or process from achieving the goal?
M a k i n g
155
FIGURE 6.8 Patient Falls FMEA
SOURCE: IHI (2005). This material was accessed from the Institute for Healthcare
Improvements web site, IHI.org. Web page address: http://www.ihi.org/ihi/workspa
ce/tools/fmea/ViewTool.aspx?ToolId=1248.
2. Exploit the constraint. Determine how to get the maximum performance out of t
he constraint without major system changes or capital improvements. 3. Subordina
te everything else to the constraint. Other nonbottleneck resources (or steps in
the process) should be synchronized to match the
156
output of the constraint. Idleness at a nonbottleneck resource costs nothing, an
d nonbottlenecks should never produce more than can be consumed by the bottlenec
k resource. For example, if the operating room is a bottleneck and there is a su
rgical ward associated with it, a traditional view might encourage filling the w
ard. However, nothing would be gained (and operational losses would be incurred)
by putting more patients on the ward than the operating room could process. Thu
s, the TOC solution is to lower ward occupancy to match the operating rooms throu
ghput, even if resources (heating, lighting, fixed staff costs, etc.) seem to be
wasted. 4. Elevate the constraint. Do something (expend capital, hire more peop
le, etc.) to increase the capacity of the constraining resource until it is no l
onger the constraint. Something else will be the new constraint. 5. Repeat the p
rocess for the new constraint. The process must be reapplied, perhaps many times
. It is important not to let inertia become a constraint. Many constraints are o
f an organizations own makingthey are the entrenched rules, policies, and procedur
es that have developed over time. TOC defines three operational measurements for
organizations: 1. Throughput: the rate at which the system generates money. Thi
s is selling price minus the cost of raw materials. Labor costs are part of oper
ating expense rather than throughput. 2. Inventory: the money the system invests
in things it will sell. This includes inventories and buildings, land, and equi
pment. 3. Operating expense: the money the system spends turning inventory into
throughput. This includes what would typically be called overhead. The following
four measurements are then used to identify results for the organization: Net p
rofit = Throughput Operating expense Return on investment (ROI) = (Throughput Op
erating expense)/Inventory Productivity = Throughput/Operating expense Turnover
= Throughput/Inventory These measurements can help employees make local decision
s. A decision that results in increasing throughput, decreasing inventory, or de
creasing the operating expense will generally be a good decision for the organiz
ation. The TOC has been used in healthcare at both a macro and microlevel to an
alyze and improve systems. Wolstenholme (1999) developed a model that showed tha
t providing additional intermediate care enabled hospitals to
M a k i n g
157
more quickly discharge patients and was more effective at increasing overall thr
oughput than providing more hospital beds. Womack and Flowers (1999) describe th
e use of TOC in the 366th Medical Group to decrease waiting time for routine pri
mary care appointments. Initially, the waiting time was reduced from an average
of 17 days to an average of 4.5 days without any significant dollar investment.
They also found that elevating the constraint of waiting time, at a cost of less
than $200,000, could significantly increase capacity. The Radcliff Infirmary in
Oxford, England, used the TOC to improve waiting times for ophthalmology (Reid
2000). That organization experienced a 100 percent reduction in cancellations as
well as an increase in throughput of 20 percent without adding resources or a m
ajor reengineering of the process. Another way to manage constraints in a system
is to acknowledge that there will always be a bottleneck and determine where it
should be. Designing the system so that the bottleneck can be best managed or c
ontrolled can be a powerful way to deal with the bottleneck.
Analytical Tools
Optimization
Optimization, or mathematical programming, is a technique used to determine the
optimal allocation of limited resources given a desired goal. For example, the r
esources might be people, money, or equipment. Of all possible resource allocati
on(s), the goal or objective is to find the allocation(s) that maximizes or mini
mizes some numerical quantity such as profit or cost. Optimization problems are
classified as linear or nonlinear depending on whether the problem is linear wit
h respect to the variables. In many cases, it is not practically possible to det
ermine an exact solution for optimization problems; a variety of software packag
es offer algorithms to find good solutions. Optimization models have three basic
elements: 1. An objective functionthe quantity that needs to be minimized or max
imized; 2. The controllable inputs or decision variables that affect the value o
f the objective function; and 3. Constraints that limit the values that the deci
sion variables can take on. A solution where all of the constraints are satisfie
d is called a feasible solution. Most algorithms used to solve these types of pr
oblems begin by finding feasible solutions and trying to improve on them until a
maximum or minimum is found. Healthcare organizations need to maintain financia
l viability while working within various constraints on their resources. Optimiz
ation techniques can
158
TABLE 6.3 DRG Linear Programming Problem Data
Resources Diagnostic services (hours) ICU bed days Routine bed days Nursing care
(hours) Margin Minimum cases
Respiratory
Coronary Surgery
Birth/ Delivery
Alcohol/ Drug Abuse
Available
7 1 5 50 $400.00 15
10 2.5 7 88 $2,500.00 10
2 0.5 2 27 $300.00 20
1 0 7 50 $50.00 10
325 55 420 3,800
help these organizations make the best allocation decision. An example of how li
near programming could be used in a healthcare organization using Microsoft Exce
l Solver follows.
Linear programming VVH wants to determine the optimal case mix for diagnosis rel
ated groups (DRGs) that will maximize profits. Limited resources (space, qualifi
ed employees, etc.) are available to service the various DRGs, and minimum level
s of service must be offered for each DRG (Table 6.3) Table 6.3 shows that the r
espiratory DRG (DRGr) requires 7 hours of diagnostic services, 1 intensive care
unit (ICU) bed day, 5 routine bed days, and 50 hours of nursing care. The profit
for DRGr is $400, and the minimum service level is 15 cases. The goal is to max
imize profit, and the objective function is:
($400 DRGr) + ($2,500 DRGcs) + ($300 DRGbd) + ($50 DRGada) The constraints are a
s follows. Diagnostic services: (7 DRGr) + (10 DRGcs) + (2 DRGbd) + (1 DRGada) 3
25 ICU bed days: (1 DRGr) + (2.5 DRGcs) + (0.5 DRGbd) 55 (2) (1)
M a k i n g
159
Routine bed days: (5 DRGr) + (7 DRGcs) + (2 DRGbd) + (7 DRGada) 420 Nursing care
: (50 DRGr) + (88 DRGcs) + (27 DRGbd) + (50 DRGada) 3,800 Respiratory minimum ca
se level: DRGr 15 Coronary surgery minimum case level: DRGcs 10 Birth/delivery m
inimum case level: DRGbd 20 Alcohol/drug abuse minimum case level: DRGada 10 (8)
(7) (6) (5) (4) (3)
Figure 6.9 shows the Excel Solver setup of this problem. Solver (Figure 6.10) fi
nds that the hospital should service 15 DRGr cases, 12 DRGcs cases, 20 DRGbd cas
es, and 29 DRGada cases. The total profit at the optimal case mix is: (15 $400)
+ (12 $2,500) + (20 $300) + (29 $50) = $43,450 Information relating to the resou
rce constraints is found in the computer solution (Figure 6.10). The amounts rep
orted as slack provide a measure of resource utilization. All available ICU bed
days and hours of nursing care will be used. However, 17 routine bed days and al
most 31 hours of diagnostic services will be unused. VVH may want to consider el
iminating some hours of diagnostic services. Constraints 5 through 8 relate to t
he minimum service level for each DRG category. Slack, or surplus, values repres
ent services that should be provided in excess of a minimum level. Only the mini
mum levels for birth/delivery
160
FIGURE 6.9 Excel Solver Setup for DRG Linear Programming Problem
and respiratory care should be provided. However, two additional coronary surger
y and 19 alcohol/drug abuse cases should be taken.
Sensitivity analysis Sensitivity analysis (Figure 6.11) examines the effect of v
arying the assumptions, or input variables, on the output of a model. Here, a se
nsitivity analyFIGURE 6.10 Excel Solver Solution for DRG Linear Programming Prob
lem
Target Cell (Max) Cell $I$9 Margin Total Name Original Value $ 3,250.00 $ Final
Value 43,454.00
Adjustable Cells Cell $B$13 $C$13 $D$13 $E$13 Name Optimal Cases Respiratory Opt
imal Cases Coronary Surgery Optimal Cases Birth/Delivery Optimal Cases Alcohol/D
rug Abuse Original Value 1 1 1 1 Final Value 15 12 20 29.08
Constraints Cell $I$4 $I$5 $I$6 $I$7 $E$13 $D$13 $C$13 $B$13 Name Diagnostic Ser
vices (hours) Total ICU Bed Days Total Routine Bed Days Total Nursing Care (hour
s) Total Optimal Cases Alcohol/Drug Abuse Optimal Cases Birth/Delivery Optimal C
ases Coronary Surgery Optimal Cases Respiratory Cell Value 294.08 55 402.56 3800
29.08 20 12 15 Formula $I$4 =$G$4 $I$5 =$G$5 $I$6 =$G$6 $I$7 =$G$7 $E$13 =$E$11
$D$13 =$D$11 $C$13 =$C$11 $B$13 =$B$11 ^ ^ ^ ^ ^ ^ ^ ^ Status Not Binding Bindi
ng Not Binding Binding Not Binding Binding Not Binding Binding Slack 30.92 0 17.
44 0 19.08 0 2 0
M a k i n g
161
Adjustable Cells Cell $B$13 $C$13 $D$13 $E$13 Constraints Cell $I$4 $I$5 $I$6 $I
$7 Name Diagnostic Services (hours) Total ICU Bed Days Total Routine Bed Days To
tal Nursing Care (hours) Total Final Value 294.08 55 402.56 3800 Lagrange Multip
lier 0 964.8 0 1 Name Optimal Cases Respiratory Optimal Cases Coronary Surgery O
ptimal Cases Birth/Delivery Optimal Cases Alcohol/Drug Abuse Final Value 15 12 2
0 29.08 Reduced Gradient 614.8 0 209.4 0
FIGURE 6.11 Sensitivity Analysis for DRG Linear Programming Problem
sis is used to analyze the allocation and utilization of resources (diagnostic s
ervice hours, ICU bed days, routine bed days, nursing care) in relation to the o
bjective function (total profit). Shadow prices (the Lagrange multiplier in Figu
re 6.11) show the dollar effect on total profit of adding or deleting one unit o
f the resource. This allows the organization to weigh the relative benefits of a
dding more resources. In this example, adding one ICU bed day would increase tot
al profit by $964.80, and adding one hour of nursing care would increase total p
rofit by $1. If the cost of either of these options is less than the additional
profit, the hospital should increase those resources. Because there is slack in
routine bed days and diagnostic services, adding more of either of these resourc
es would not change the total profit; there is already an excess of these resour
ces. Shadow price information is also presented for the DRG minimum service leve
l requirements (the reduced gradient in Figure 6.11). The shadow price is a nega
tive $614.80 for DRGr; total profit will decrease by $614.80 for each case taken
above the minimum level required in the DRGr category. The DRGr category has a
higher profit ($400) than the DRGada and, without this analysis, the hospital mi
ght have mistakenly tried to serve more DGRr cases to the detriment of DRGada ca
ses. Optimization can also allow organizations to run what if analyses. For exampl
e, if a hospital wants to investigate the possibility of increasing beds in its
ICU, perhaps by decreasing routine beds, it could use optimization to analyze th
e available choices.
Decision Analysis
Decision analysis is a process for examining and evaluating decisions in a struc
tured manner. A decision tree is a graphic representation of the order
162
of events in a decision process. It is a structured process that enables an eval
uation of the risks and rewards of choosing a particular course of action. In co
nstructing a decision tree, events are linked from left to right in the order in
which they would occur. Three types of events, represented by nodes, can occur:
decision or choice events (squares), chance events (circles), and outcomes (tri
angles). Probabilities of chance events occurring and benefits or costs for even
t choices and outcomes are associated with each branch extending from a node. Th
e result is a tree structure with branches for each event extending to the right
. A simple example will help to illustrate this process. A health maintenance or
ganization (HMO) is considering the economic benefits of a preventive flu vaccin
ation program. If the program is not offered, the estimated cost to the HMO if t
here is a flu outbreak is $8 million with a probability of 0.4 (40 percent), and
$12 million with a probability of 0.6 (60 percent). The program is estimated to
cost $7 million, and the probability of a flu outbreak occurring is 0.7 (70 per
cent). If a flu outbreak does occur and the HMO offers the program afterward, it
will still cost the organization $7 million, but the resulting costs to the HMO
would be reduced to $4 million with a probability of 0.4 (40 percent) or $6 mil
lion with a probability of 0.6 (60 percent). What should the HMO decide? The dec
ision tree for the HMO vaccination program is shown in Figure 6.12. FIGURE 6.12
HMO Vaccination Program Decision Tree 1
Flu outbreak Program No flu outbreak Program HMO vaccination decision Flu outbre
ak C A
B
No program
No program
D No flu outbreak
NOTE: The tree diagrams in Figures 6.12 to 6.15 were drawn with the help of Prec
isionTree, a software product of Palisade Corp., Ithaca, NY: www.palisade.com.
M a k i n g
163
70.0% Flu outbreak $0 Program $7,000,000 2 30.0% No flu outbreak $0 Program HMO v
accination decision $7,000,000 1 70.0% Flu outbreak $0 60.0% A $6,000,000 5 40.0%
B $4,000,000 60.0% C $1 2,000,000 6 40.0% D $8,000,000
FIGURE 6.13 HMO Vaccination Program Decision Tree 2
4
No program $0
3
No program $0
30.0% No flu outbreak $0
The probability estimates for each of the chance nodes, benefits (in this case c
osts) of each decision branch, and outcome branch are added to the tree (Figure
6.13). The value of a node can be calculated once the values for all subsequent
nodes are found. The value of a decision node is the largest value of any branch
out of that node. The assumption is that the decision that maximizes the benefi
ts will be made. The value of a chance node is the expected value of the branche
s out of that node. Working from right to left, the value of all nodes in the tr
ee can be calculated. The expected value of chance node 6 is [0.6 (12)] + [0.4 (8)
] = 10.4. The expected value of chance node 5 is [0.6 (6)] + [0.4 (4)] = 5.2. The ex
pected value of the secondary vaccination program is 7 + (5.2) = 12.2, and the expe
cted value of not implementing the secondary vaccination program is 10.4. Therefo
re, at decision node 4 the choice would be to not implement the secondary vaccin
ation program. At chance node 3 (no initial vaccination program), the expected v
alue is 0.7 (10.4) + 0.3 0 = 7.28. The expected value at chance node 2 is 0.7 0 +
0.3 0 = 0, and the expected value of the initial vaccination program branch is 7
+ 0 = 7. Therefore, at decision node 1 the choice would be to implement the initi
al vaccination program at a cost of $7 million. In contrast, choosing not to imp
lement the initial vaccination program has an expected cost of $7.28 million (Fi
gure 6.14).
164
FIGURE 6.14 HMO Vaccination Program Decision Tree 3
Choose this path because expected costs of $7 million are less than $7.28 millio
n. Program 7
70.0% Flu outbreak 30.0% No flu outbreak 0 Program 7 Flu 7 0 60.0% A Costs 12.2 6
HMO vaccination decision
Vaccination program #1 7 Flu outbreak 70.0% 0 No program Flu 7.28
40.0% B 4 Vaccination program #2 60.0% 10.4 C 12 No program 0 Costs 10.4 40.0% D
0
30.0% No flu outbreak 0
8
Choose this path because expected costs of $10.4 million are less than $12.2 mil
lion.
A risk analysis on this decision analysis can be conducted (Table 6.4). Choosing
to implement the vaccination program results in a cost of $7 million with a pro
bability of 1. Choosing not to implement the initial vaccination program results
in a cost of $12 million with a probability of 0.42, $8 million with a probabil
ity of 0.28, and no cost with a probability of 0.3. Choosing not to implement th
e vaccination program would be less costly 30 percent of the time, but 70 percen
t of the time it would be less costly to implement the vaccination program. A se
nsitivity analysis might also be conducted to determine the effect of changing s
ome or all of the parameters in the analysis. For example, if the risk of a flu
outbreak was 0.6 rather than 0.7 (and all other parameters stayed the same), the
optimal decision would be to not offer either vaccination program (Figure 6.15)
. For this example, dollars were used to represent costs (or benefits), but any
type of score could also be used. In the medical field, decision trees are often
used in deciding among a variety of treatment options using a measure or score
for the outcome of the treatment (Detsky et al. 1997; Ruland 2002).
M a k i n g
165
Initial Vaccination Program Number 1 2 3 X 7 P 1
No Initial Vaccination Program X 12 8 0 P 0.42 0.28 0.3
TABLE 6.4 Risk Analysis for HMO Vaccination Program Decision
X = Cost in millions of dollars; P = Probability
60.0% Flu outbreak Program 7 40.0% No flu outbreak 0 Program HMO vaccination deci
sion Vaccination program #1 7 7 Flu 7 0 60.0% A Costs 12.2 40.0% 60.0% Flu outbreak
0 Vaccination program #2 10.4 B 4 60.0% C Costs 10.4 40.0% D 40.0% No flu outbreak
0 8 12 6
FIGURE 6.15 Decision Analysis Sensitivity to Change in Risk of Flu Outbreak
No program 0
Flu 6.24
No program 0
Decision trees can be a powerful aid to evaluating and choosing the optimal cour
se of action. However, care must be taken when using them. Possible outcomes and
the probabilities and benefits associated with them are only estimates, and the
se estimates could differ greatly from the reality. Also, when using expected va
lue (or expected utility) to choose the optimum path, there is also an underlyin
g assumption that the decision will be made over and over. On average the expect
ed payout is received, but in each individual situation different amounts are re
ceived.
Implementation: Force Field Analysis
Derived from the work of Kurt Lewin (1951), force field analysis is a technique
for evaluating all of the various forces for and against a proposed
166
BOX 6.3 Common Forces to Consider in Force Field Analysis
Available resources Costs Vested interests Regulations Organizational structures
Present or past practices Institutional policies or norms Personal or group atti
tudes and needs Social or organizational norms and values
change. It can be used to decide if a proposed change can be successfully implem
ented. Alternatively, if a decision to change has already been made, force field
analysis can be used to develop strategies that will enable the change to be im
plemented successfully. In any situation, driving forces will help to achieve th
e change and restraining forces will work against the change. Force field analys
is identifies these forces and assigns relative scores to each. Box 6.3 lists ty
pical forces that should be considered. If the total score of the restraining fo
rces is greater than the total score of the driving forces, the change may be do
omed to failure. Force field analysis is typically used to determine how to stre
ngthen or add driving forces or weaken the restraining forces to enable successf
ul implementation of a change. Patients at VVH believed that they were insuffici
ently involved in and informed about their care. After analyzing this problem, h
ospital staff believed that the problem could be solved (or lessened) by moving
shift handover from the nurses station to the patients bedside. A force field anal
ysis was conducted and is illustrated in Figure 6.16. Although the restraining f
orces were greater than the driving forces in this example, a decision was made
to implement the change in handover procedures. To improve the projects chances f
or success, a protocol was developed for the actual procedure, making explicit t
he following: Develop and disseminate the protocol (new driving force +2). Confi
dential information will be exchanged at the nurses station, not at the bedside h
andover (decrease fear of disclosure by 2). Lateness was addressed and solutions
were developed and incorporated into the protocol (decrease problems associated
with late arrivals by 2). These changes increased the driving forces by 2, to 2
1, and decreased the restraining forces by 4, to 17. The change was successfully
implemented; more importantly, patients felt more involved in their care and co
mplaints were reduced.
M a k i n g
167
Driving Forces
Restraining Forces
FIGURE 6.16 Force Field Analysis
4
4
Critical incidents on the increase Plan: Change to bedside shift handover
Ritualism and tradition
4
Staff knowledgeable in change management
Fear that this may lead to more work
4
3
Increase in discharge against medical advice
Fear of increased accountability
3
5
Complaints from patients and doctors increasing
Problems associated with late arrivals
5
3 Care given is predominantly biomedical in orientation Total: 19
SOURCE: Adapted from Kassean and Jagoo (2005).
Possible disclosure of confidential information Total: 21
5
Conclusion
The tools and techniques outlined in this chapter are intended to help organizat
ions along the path of continuous improvement. The choice of tool and when to us
e that tool is dependent on the problem to be solved; in many situations, severa
l tools from this and other chapters should be used to ensure that the best poss
ible solution has been found.
1. Answer the following questions quickly for a fun illustration of some of the
ten decision traps: Can a person living in Milwaukee, Wisconsin, be buried west
of the Mississippi? If you had only one match and entered a room where there was
a lamp, an oil heater, and some kindling wood, which would you light first? How
many animals of each species did Moses take along on the ark?
168
If a doctor gave you three pills and said to take one every half hour, how long
would they last? If you have two U.S. coins totaling 55 cents and one of the coi
ns is not a nickel, what are the two coins? What decision traps did you fall int
o when answering these questions? 2. Discuss a problem your organization has sol
ved or a suboptimal decision the organization made because the frame was incorre
ct.
Chapter Exercises
1. In the HMO vaccination program example, analyze the situation if the probabil
ity of a flu outbreak is only 65 percent and the cost of the vaccination program
is $8 million. What would your decision be in this case? 2. In the DRG case mix
problem, VVH has determined that it could convert 15 of its routine beds to ICU
beds for a cost of $2,000. What should VVH do and why? 3. The high cost of medi
cal care and insurance is a growing societal problem. Develop a mind map of this
problem. (Advanced: Use Inspiration software.) 4. Individually or in teams, dev
elop a map of a healthcare process or system with which you are familiar. Make s
ure that your process map has a start and an endpoint, all inputs and outputs ar
e defined, and all key process steps are included. Explain your map to the rest
of the classthis may help you to determine if anything is missing. (Advanced: Use
Microsoft Visio.) 5. For this exercise, choose a service offered by a healthcar
e organization and create a service blueprint. You may have to imagine some of t
he systems and services that take place backstage if you are unfamiliar with the
m. 6. Think of a problem in your healthcare organization. Perform an RCA of the
identified problem using the five whys technique and a fishbone diagram. 7. Pick
one solution to the problem identified in exercise 4 and do a force field analy
sis. 8. Complete the TOC exercise found at: http://webcampus.stthomas.edu/ cob/j
mhays/ (Note: this is not a healthcare exercise, but it does help in understandi
ng TOC.)
References
Buzan, T. 1991. Use Both Sides of Your Brain. New York: Plume. Buzan, T., and B.
Buzan. 1994. The Mind Map Book: How to Use Radiant Thinking to Maximize Your Br
ains Untapped Potential. New York: Dutton. Detsky, A. S., G. M. Naglie, M. D. Kra
hn, D. M. Naimark, and D. A. Redelmeier. 1997. Primer on Medical Decision Analysi
s: Parts 1 to 5. Medical Decision Making 17 (2): 12359.
M a k i n g
169
Goldratt, E. M., and J. Cox. 1986. The Goal: A Process of Ongoing Improvement. N
ew York: North River Press. Institute for Healthcare Improvement. 2005. Failure M
ode and Effects Analysis Tool. [Online information; retrieved 1/1/06.] www.ihi.or
g/ihi/workspace/ tools/fmea/. Ishikawa, K. 1985. What Is Total Quality Control?
Translated by D. J. Lu. Englewood Cliffs, NJ: Prentice Hall Inc. Joint Commissio
n. 2005a. Sentinel Event Forms and Tools. [Online information; retrieved 1/1/06.]
www.jointcommission.org/SentinelEvents/Forms/. . 2005b. Sentinel Event Policy and Pr
ocedures. [Online information; retrieved 1/1/06.] www.jointcommission.org/Sentine
lEvents/Policyand Procedures/. . 2001. Patient Safety StandardsHospitals. [Online info
rmation; retrieved 1/1/06.] www.jcrinc.com/subscribers/perspectives.asp?durki=29
73 &site=10&return=2897. Kassean, H. K., and Z. B. Jagoo. 2005. Managing Change i
n the Nursing Handover from Traditional to Bedside HandoverA Case Study from Maur
itius. BMC Nursing 4: 1. Lewin, K. 1951. Field Theory in Social Science: Selected
Theoretical Papers, edited by D. Cartwright. New York: Harper. Reid, S. 2000. Us
ing the Theory of Constraints Methodology to Increase, Improve Services to Patient
s. ImpAct 6 (March/April). Ruland, C. 2002. Patient Preferences in Health Care Dec
ision Making. [Online information; retrieved 1/7/06.] www.dbmi.columbia.edu/homep
ages/ cmr7001/sdm/html/methods.htm. Russo, J. E., and P. J. H. Schoemaker. 1989.
Decision Traps: The Ten Barriers to Brilliant Decision Making and How to Overco
me Them. New York: Simon & Schuster. Shostack, G. L. 1984. Designing Services tha
t Deliver. Harvard Business Review 62 (1): 13339. Veterans Administration National
Center for Patient Safety. 2006. Healthcare Failure Mode and Effect Analysis (HF
MEA). [Online information; retrieved 1/7/06.] hwww.patientsafety.gov/SafetyTopics
.html#HFMEA. Wolstenholme, E. 1999. A Patient Flow Perspective of UK Health Servi
ces: Exploring the Care for New Intermediate Care Initiatives. System Dynamics Revi
ew 15 (3): 25371. Womack, D., and S. Flowers. 1999. Improving System Performance:
A Case Study in the Application of the Theory of Constraints. Journal of Healthca
re Management 44 (5): 397405. Wysocki, B., Jr. 2004. To Fix Health Care, Hospitals
Take Tips from Factory Floor. Wall Street Journal, April 9, A1A5. [Online informa
tion; retrieved 7/1/06.] http://prhi.org/pdfs/ACF28EB.pdf.
CHAPTER
7
USING DATA AND STATISTICAL TOOLS FOR OPERATIONS IMPROVEMENT
CHAPTER OUTLINE
Operations Management in Action Overview Data Collection Logic Selection or Samp
ling Census versus sample Nonrandom sampling methods Simple random sampling Stra
tified sampling Systematic sampling Cluster sampling Sample size Measurement Pre
cision Accuracy Reliability Validity Ethical and Legal Considerations Graphic To
ols Mapping Visual Representations of Data Histograms and Pareto Charts Dot Plot
s Turnip Graphs Normal Probability Plots Scatter Plots Mathematic Descriptions M
easures of Central Tendency Mean Median Mode Measures of Variability Range Mean
absolute deviation Variance Standard deviation Coefficient of variation Probabil
ity and Probability Distributions Determination of Probabilities Observed probab
ility Theoretical probability Opinion probability Properties of Probabilities Bo
unds on probability Multiplicative property Additive property Conditional probab
ility Probability Distributions Discrete Probability Distributions Binomial dist
ribution Poisson distribution Continuous Probability Distributions Uniform distr
ibution Normal distribution Triangular distribution Exponential distribution Chi
Square Test Confidence Intervals and Hypothesis Testing Central Limit Theorem H
ypothesis Testing Type I and Type II errors Equal variance t test Proportions Pr
actical versus statistical significance ANOVA/MANOVA One way ANOVA Two way and h
igher ANOVA MANOVA/ MANCOVA Simple Linear Regression Interpretation Coefficient
of Determination and Correlation Coefficient Problems with correlation coefficie
nts Statistical Measures of Model Fit F test t Test Assumptions of Linear Regres
sion Transformations Multiple Regression The General Linear Model Additional Tec
hniques Artificial Neural Networks
170
Design of Experiments Conclusion

Which Technique to Use Discussion Questions
Chapter Exercises References
additive property of probability error, see Type I error nlysis of vrince (A
NOVA) re smple, see cluster smple rtificil neurl network (ANN) Byes theor
em error, see Type II error inomil distriution ounds on proility census c
entrl limit theorem centrl tendency Cheyshevs rule chi-squre test (2) luster
sample oeffiient of determination (r2) oeffiient of variation (CV) ondition
al probability onfidene interval (CI) onveniene sample, see nonrandom sample
orrelation oeffiient (r) design of experiments (DOE) dot plot exponential di
stribution F-test general linear model (GLM) histogram hypothesis test logi of
a study mean absolute deviation (MAD) mean, median, mode multiollinearity multi
ple regression multipliative property of probability multivariate analysis of v
ariane (MANOVA) nonrandom sample normal distribution normal probability plot ne
ural network (NN) observed probability odds opinion probability Pareto hart Poi
sson distribution population mean (), see mean population standard deviation (),
ee tandard deviation population variance (2), ee variance population practical
ignificance preciion probability ditribution proportion range reliability am
ple ample ize catter plot equential ample, ee ytematic ample imple lin
ear regreion imple random ample tandard deviation tandard normal ditribut
ion, ee z-ditribution tatitical ignificance tratified ample ytematic a
mple t-ditribution theoretical probability tranformation tree diagram triangul
ar ditribution t-tet turnip graph Type I error Type II error uniform ditribut
ion validity variance Venn diagram z-ditribution
171
Performance Improvement Tool, Technique, and Program

Operation Management in Action
Who Counting: Flu Death, Iraqi Dead Number Skewed
(Reprinted by permiion of the author, John Allen Paulo)
Fear of an avian-flu pandemic i widepread and growing. It i, of coure, a ver
y eriou concernall the more reaon for report on avian flu to be a clear and
informative a poible. On the whole, thi tandard ha been reaonably approxi
mated, but one glaring exception ha been the reporting on the mortality rate fr
om H5N1 avian flu. New tory after new tory repeat the tatitic that out of
140 or o human cae of avian flu reported o far in Southeat Aia, more than
half have reulted in death. The reporter then intone that the mortality rate
for avian flu i more than 50 percent. Thi, of coure, i a terrifying figure.
But before examining it, let firt look for a bit of perpective. The tandard
ort of influenza viru, it believed, infect omewhere between 20 million and 60
million people in thi country annually. It kill an average of 35,000, and it
thu ha a mortality rate that i a minucule fraction of 1 percent. The wine f
lu in the 1970 killed a handful of people, more of whom may have died from the
vaccine for it than from the dieae itelf. And the Spanih flu of 1918 to 1919t
he deadliet pandemic in modern hitory and alo an avian flukilled 500,000 to 70
0,000 people here and an etimated 20 million to 50 million people worldwide. Mo
t aement of it mortality rate range between 2 percent and 5 percent. If t
he avian H5N1 viru mutated o that human-to-human tranmiion wa a eay a i
t i with the normal flu, and if the mortality rate of more than 50 percent held
, the U.S. alone would be facing ten of million of death from the avian flu.
Sample Bia There i one glaring problem with thi purported mortality rate of m
ore than 50 percent, however: It i baed on thoe cae that have been reported
, and thi lead to an almot textbook cae of ample bia. You wouldnt etimate
the percentage of alcoholic by focuing your reearch on bar patron, nor would
you etimate the percentage of port fan by hanging around port tadium. W
hy do omething analogou when etimating the avian-flu mortality percentage? Wh
o goe to hopital and emergency room but people who are very ick? Who doent
go to hopital and emergency room but people who are not o very ick? Given h
ow many people in Southeat Aia deal with chicken and other poultry in a hand
-on way, ome unknown, but likely quite large num172
C h a p t e r 7 : U i n g D a t a a n d St a t i t i c a l To o l f o r O
p e ra t i o n I m p rove m e n t
173
ber of them have contracted avian H5N1 flu over the pat everal year. It ymp
tom are many time inditinguihable from thoe of the normal flu. Some no doub
t have died of it, but the majority are likely to have recovered. The people who
have recovered from the avian flu dont make it into the mortality-rate calculati
on, which are, a a conequence, likely to be kewed ubtantially upward. Litt
le I Known About Change and Virulence The Center for Dieae Control and Prev
ention cautiouly acknowledge on it webite that it i quite poible that the
only cae currently being reported are thoe in the mot everely ill people. Ra
ndom amplethe lifeblood of tatiticare hard to obtain under the bet of circum
tance, and canvaing poor people in rural Southeat Aia certainly doent qual
ify. The fact i that not only i it almot impoible to etimate how long it m
ight take for the preent H5N1 viru to mutate into one eaily tranmitted betwe
en human, but it almot impoible a well to etimate the virulence of uch a
viru. Sample bia, variou epidemiological model, and the fact that extremely
virulent virue are le likely to preadbecaue they kill a high percentage of
their hotall ugget that a mortality rate of more than 50 percent i much, muc
h too high. Thi, of coure, doe not mean complacency i in order. It doe mean
that everyonehealth profeional, the media, and politicianhould try very hard
not to hype rik or minimize them. Potcript on Iraqi War Dead Another figure
in the new recently ha been the number of Iraqi killed in the war. Preident
Buh mentioned lat month that in addition to the more than 2,100 American old
ier killed o far in Iraq, that there were approximately 30,000 Iraqi killed.
He wa likely referring to the approximate figure put out by Iraq Body Count, a
group of primarily Britih reearcher who ue online Wetern media report to c
ompile an extenive lit of Iraqi civilian killed. The organization check the
name and aociated detail of thoe killed. It necearily mie all thoe wh
oe name dont make it into the report, and it make no attempt to etimate the
number it mie. The group lit contain almot 30,000 name at preent. A tud
y that appeared in the pretigiou Britih medical journal, The Lancet, in Octob
er 2004, ued tatitical ampling technique to etimate all Iraqi killed beca
ue of the war and it myriad direct and indirect effect. The figure reearcher
came up with at that time15 month agowa approximately 100,000 dead, albeit wit
h a large margin of error. The
174
Lancet tudy ued the ame technique that Le F. Robert, a reearcher at John
Hopkin and lead author of the tudy, ued to invetigate mortality caued by t
he war in Bonia, the Congo, and Rwanda. Although Robert work in thoe location
wa unquetioned and widely cited by many, including Britih Prime Miniter To
ny Blair and former Secretary of State Colin Powell, The Lancet etimate on Ira
q were unfortunately dimied or ignored in 2004. Thee lat 15 month have con
iderably raied the American death toll, the IBC number, and any update that m
ay be in the work for The Lancet taggering 100,000 figure. In fact, if the Lan
cet etimate roe at a rate proportional to the IBC number ince October 2004fr
om about 17,000 then to about 30,000the updated figure would be approximately 175
,000 Iraqi dead ince the war began.
Overview
The focu of thi chapter i on data collection and analyi. Knowledge-baed ma
nagement and improvement require that deciion be baed on fact rather than on
feeling or intuition. Collecting the right data and analyzing them correctly en
able fact-baed deciion making. Many of the philoophie and tool (e.g., qual
ity, Lean, imulation) dicued in later chapter of thi book require a baic
knowledge of data collection, probability, and tatitical analyi. For thoe r
eader with little or no background in tatitic or probability, thi chapter i
meant to provide an introduction to the baic concept and tool, many of whic
h are an integral part of the continuou improvement philoophy of quality and L
ean. For reader with greater undertanding of tatitic and probability, thi
chapter i intended to be a review of the important concept underlying the meth
odologie covered in later chapter. The major topic covered include: Data coll
ection; Graphic tool for data preentation and analyi; Probability and probab
ility ditribution; and Mathematic tool for data analyi including hypothei
teting, analyi of variance, and regreion.
More in-depth coverage of the topic in thi chapter can be found on ache.org/bo
ok/OpManagement. Thi chapter give reader a baic undertanding of probabili
ty and tatitic to enable them to:
175
Collect valid data; Analyze data uing the appropriate graphic or mathematic tec
hnique; and Ultimately, make deciion baed in fact.
Data Collection
To deign effective and efficient ytem or improve exiting ytem, knowledge
of the ytem, including both input to the ytem and the deired output, i n
eeded. The goal of data collection i to obtain valid data to better undertand
and improve the ytem being tudied. Deciion or olution baed on invalid da
ta are doomed to failure. Enuring that the data obtained are valid i the mot
important part of any tudy, and often the mot problematic. A valid tudy ha n
o logic, ampling, or meaurement error.
Logic
The firt tep in data collection i determining the quetion() that will be a
ked of the data. Why are the data needed, and what will they be ued for? Are th
e pattern of the pat going to be repeated in the future? If there i reaon to
believe that the future will look different than the pat, data from the pat w
ill not help to anwer the quetion, and other, nonquantitative method hould b
e ued. Thi i the logic phae of the data collection proce, where the focu
i on enuring that the right quetion i being aked and that it i poible to
anwer that quetion.
Selection or Sampling
Cenu veru ample Next, one need to determine who or which object are relev
ant to the quetion or ituation. Here, a determination i made a to whether a
ample or a cenu i appropriate. Typically, a cenu i ued if all of the dat
a are available in an eaily acceible format. A ample i ued if it would be
cot or time prohibitive to obtain all of the data in a uable format. If it i
determined that a ample i appropriate, how to ample from the population of in
teret mut be determined. Nonrandom ampling method Nonrandom, or convenience,
method are election procedure where there i no pecific probability of ele
cting any particular unit. They hould only be ued a a lat reort, if random
method are unavailable or would be cot prohibitive, becaue of the poibility
that they will not be repreentative of the targeted population of interet. No
nrandom method include elf-elected
176
group in which the member are volunteer and group in which the unit are mo
t acceible and convenient for the reearcher.
Simple random ampling In imple random ampling, every unit in the population h
a an equal chance of being choen. Stratified ampling In tratified ampling,
the population i firt divided into egment, and a random ample i taken from
each egment. Stratified ampling i ued to force repreentation by all egmen
t of the population. Thi type of ampling i typically ued where a imple ran
dom ample might mi ome egment of the population, uually becaue the popul
ation contain few of them. Sytematic ampling In ytematic, or equential, a
mpling every 10th or Xth unit i choen. Although thi type of ampling i omet
ime eaier to adminiter, caution mut be ued to enure that the population i
not work mirepreented. For example, if a deciion wa made to look at every 5
th work day and thi alway fall on Friday, the ample obtained might not be re
preentative of the whole week. Cluter ampling In cluter, or area, ampling,
unit are grouped or clutered, and a imple random or tratified ample i take
n from the grouped unit. Thi i typically done when it would be cot or time p
rohibitive to obtain the data from a imple random ample. For example, if a tu
dy wa baed on face-to-face interview of patient, and the population of inter
et wa every peron who viited an emergency department for heart-related probl
em in Minneota, generating a lit of all people who viited an emergency depar
tment for heart-related problem in Minneota and then randomly ampling from th
i lit would neceitate extenive travel to reach the ubject. Intead, the h
opital could be clutered by ize, a random ample of hopital within thee c
luter would be taken, and only patient at the choen hopital would be inter
viewed. Thi would greatly reduce the interviewer travel. Sample ize Sample iz
e i determined by how certain the concluion of a tudy are deired to be. Pop
ulation ize i irrelevant; larger population do not need larger ample ize.
However, variation in the population doe affect the needed ample ize: A vari
ation increae, larger ample ize are needed to obtain the ame level of cert
ainty. Larger ample (auming they are valid) enable greater certainty about t
he concluion.
177
Meaurement
In the meaurement phae of the data collection proce, one mut determine how
to meaure the characteritic of interet. Often, direct meaure of the charact
eritic of interet are difficult to obtain, and a proxy for the variable mut b
e identified. It i ueful to ditinguih between two broad type of variable:
categoric, or qualitative, and numeric, or quantitative. Each i broken down int
o two ubtype: qualitative data can be ordinal or nominal, and quantitative dat
a can be dicrete (often, integer) or continuou. Becaue qualitative data alway
have a limited number of alternative value, they are ometime alo decribed
a dicrete. All qualitative data are dicrete, wherea ome numeric data are d
icrete and ome are continuou. For tatitical analye, qualitative data can
often be converted into dicrete numeric data by counting the number of unit in
each category. Meaurement error arie from a lack of preciion, accuracy, or
reliability.
Preciion The preciion of a meaurement i it degree of finene. For example,
in a tudy of patient weight, weight could be meaured to the nearet ten pound
, one pound, one ounce, and o forth. The preciion of a meaurement for a tud
y hould be determined on the bai of the tudy requirement. Typically, greate
r preciion i more cotly or time conuming. Accuracy Accuracy i related to wh
ether the meaurement are on target. An accurate meaurement i not biaed in any
way. For example, in a tudy of patient compliance, the accuracy of the meaure
ment would be poor if the patient did not tell the truth. Reliability Reliabili
ty i related to repeatability. Will the meaurement be the ame if it i repeat
ed? If the tudy wa repeated on the ame (or a imilar) ample, would the outco
me be the ame? If a patient i teted for diabete, the tet would be reliable
if it could be repeated and the ame reult obtained. A pictorial comparion i
hown in Figure 7.1. Validity The importance of enuring that a tudy i valid
cannot be overemphaized. For a tudy to be valid, it mut have no logic, ampli
ng, or meaurement error. Each of thee threat to validity hould be aeed
and eliminated to enure the validity of the data and tudy. Concluion or deci
ion made baed on an invalid tudy are themelve invalid. Valuable time and e
ffort will
178
FIGURE 7.1 Reliability and Accuracy
Reliable, but not accurate
Reliable and accurate
Not reliable but, on average, accurate
be wated on data collection if thoe data do not enable better deciion. Figur
e 7.2 illutrate the iue and choice involved in data collection.
Ethical and Legal Conideration
Often, ethical and legal conideration are involved in collecting, analyzing, a
nd reporting data. For healthcare organization in particular, patient confident
iality and privacy may need to be addreed in the deign and execution of any
tudy. Ethical conideration and legal compliance hould be an integral part of
the deign of any tudy undertaken by an organization.
Graphic Tool
Mapping
Graphic tool uch a mind mapping can be employed to enable the collection of v
alid data. A mind map can be ued to help frame the problem or quetion in an at
tempt to avoid logic error (Figure 7.2 and Chapter 6).
Viual Repreentation of Data
Once valid data are collected, thoe data need to be analyzed to anwer the orig
inal quetion or make a deciion. The data need to be examined not only to deter
mine their general characteritic, but alo to look for intereting or unuual
pattern. Subequent ection of thi chapter outline numeric tool that can be
employed for thi purpoe. Viual repreentation of the data can be powerful in
both anwering quetion and convincing other of the anwer. Tufte (1983, 199
0, 1997) provide guidance for viual preentation of data. The following ectio
n preent graphic analyi tool. The human mind i powerful and ha the abilit
y to dicern pattern in data, which numeric method can then be ued to validat
e. The firt tep in data analyi i alway to graph the data.
179
Logic error Wrong characteritic/ wrong problem IDEA
Source external Accuracy low Preciion low Managerial Strategic Operational
Source internal/external Accuracy mid Preciion mid
FIGURE 7.2 Data Collection Map
WHY Creating critiquing DATA
Ue
Source internal Accuracy high Preciion high
Logic
Validity iue
Logic WHAT
Future not like pat
Stop! Delphi, etc.
Meaurement Selection/ ampling
Future will be like the pat
WHO
Population
HOW Quantitative
Continuou interval ratio
Sample
Size
Data type
Qualitative
Dicrete nominal ordinal
Stratified
Sequential Meaure Methodology Repeatable
Say = do
Accurate
Correct level Random Cluter Reliable Precie
Hitogram and Pareto Chart

Hitogram and Pareto chart are two of the even baic quality tool (Chapter 8
). A hitogram (Figure 7.3) i ued to ummarize dicrete or continuou data. Th
ee graph can be ueful for invetigating or illutrating important characteri
tic of the data uch a their overall hape, ymmetry, location, pread, outlie
r, cluter, and gap. However, for ome ditribution, a particular choice of
bin width can ditort the feature of a data et. (For an example of thi proble
m, ee the Old Faithful Hitogram applet linked from the companion web ite at a
che.org/book.OpManagement.)
180
FIGURE 7.3 Hitogram of Length of Hopital Stay
Frequency
14 12 10 8 6 4 2 0 12 34 56 78 910 1112 1314 1516 1718
Length of hopital tay (day)
To contruct a hitogram, the data are divided or grouped into clae. For each
group, a rectangle i contructed with it bae equal to the range of value in
the group and it area proportional to the number of obervation falling into
the group. If the range are the ame length, the height of the hitogram will a
lo be proportional to the number of obervation falling into that group. For c
ategoric data, a bar chart i typically ued, and the data are grouped by catego
ry. The height of each bar i proportional to the number of obervation. A Pare
to diagram (Figure 7.4) i imply a bar chart or hitogram orted by frequency o
f occurrence.
Dot Plot
A dot plot (Figure 7.5) i imilar to a hitogram, but the frequency of occurren
ce i repreented by a dot. Dot plot are ueful for diplaying mall data et
with poitive value becaue they are quick and eay to contruct by hand. FIGUR
E 7.4 Pareto Diagram of Diagnoi Category
12 10 Frequency 8 6 4 2 0
ne alig op na la nt m P yc ho e
liv er
H di ear e t a e
on
m
Pn
Diagnoi
M
Fr ac
De
ue
tu
re
y
ia
181
FIGURE 7.5 Dot Plot of Length of Hopital Stay
3
6
9 Day
12
15
18
Turnip Graph
A turnip graph i imilar to a dot plot, but it i centered. Interval are on th
e vertical axi of the plot, and ymbol (or dot) are placed horizontally on th
e interval to repreent the frequency of occurrence or number of obervation i
n that interval (Figure 7.6).
Normal Probability Plot
A normal probability plot (Figure 7.7), ometime called a quantile plot, i a v
iual diplay of the normality (or lack of normality) in a given et of data. Ma
ny tatitical tet and technique aume or require that the data be nor80.0 75.0 70.0 65.0 60.0 55.0 50.0 45.0
NOTE: Each dot repreent the core of a region on a quality meaure for diabete
. The numerator i the number of patient with diabete who received the medica
lly neceary care, an examination. The denominator i the number of patient wi
th diabete living in the region. SOURCE: Wennberg (2005). Data from the Dartmou
th Atla Project, figure copyrighted by the Trutee of Dartmouth College. Ued
with permiion.
Elmira Rocheter Syracue Eat Long Iland Albany White Plain Binghamton Buffal
o Manhattan Bronx
69.5 67.1 66.3 66.0 65.8 65.8 64.5 64.4 64.3 63.4
FIGURE 7.6 Turnip Graph: Medicare Enrollee with Diabete Receiving Eye Exam Am
ong 306 Hopital Referral Region, 2001
182
Expected cumulative probability
FIGURE 7.7 Normal Probability Plot of Length of Hopital Stay
1.00
0.75
0.50
0.25
0.00 0.00
0.25
0.50
0.75
1.00
Oberved cumulative probability
mally ditributed. The normal plot can help to determine thi. Although other gr
aphic diplay can help to determine if the ditribution i bell haped, the nor
mal plot i more ueful in determining ymmetry, lack of outlier, and weight of
the tail. The normal plot i a catter plot of the actual data value plotted
againt the ideal value from a normal ditribution. If the data are normally dit
ributed, the plotted point will lie cloe to a diagonal traight line. Sytemat
ic deviation from a line indicate a non-normal ditribution.
Scatter Plot
Scatter plot are another of the even baic quality tool (Chapter 8). A catte
r plot graphically diplay the relationhip between a pair of variable and can
give initial information on whether two variable are related, how trongly the
y are related, and the direction of the relationhip. For example, i there a re
lationhip between length of hopital tay and weight? Doe length of tay incre
ae (decreae) a weight increae? How trong i the relationhip between lengt
h of tay and weight? A catter plot can help to anwer thee quetion. Regre
ion, the tatitical tool related to catter plot that give more detailed, num
eric anwer to thee quetion, i dicued later in thi chapter. To contruc
t a catter plot related to the aforementioned quetion, data on length of hop
ital tay and weight from the population of interet would be collected. Typical
ly, the caue, or independent variable, i on the horizontal (x) axi and the effec
t, or dependent variable, i on the vertical (y) axi. Each pair of variable i
plotted on thi graph. Some typical catter plot and their interpretation are
hown in Figure 7.8.
183
Strong Negative Correlation Y
Strong Poitive Correlation Y
FIGURE 7.8 Typical Scatter Plot and their Interpretation
r = 0.86 Negative Correlation Y
X
r = 0.91 Poitive Correlation Y
X
No Correlation Y
r = 0.64
X
r = 0.70
X
r = 0.06
X
Mathematic Decription
When decribing or ummarizing data, three characteritic are uually of intere
t: central tendency, pread or variation, and probability ditribution. In thi
ection, the following imple data et i ued to illutrate ome of thee mea
ure: 3, 6, 8, 3, 5.
Meaure of Central Tendency
The three common meaure of central tendency are mean, median, and mode.
Mean The mean i the arithmetic average of the population:
x , N where x = individual value and N = number of value in the population. Pop
ulation mean = = The population mean can be etimated from a ample: Sample mean
= x = x , n where n = number of value in the ample. 3+ 6 +8+3+5 =5 For our im
ple data et, x = 5
184
Median The median i the middle value of the ample or population. If the data a
re arranged into an array (an ordered data et),
3, 3, 5, 6, 8, 5 would be the middle value or median.
Mode The mode i the mot frequently occurring value. In the previou example, t
he value 3 occur more often (two time) than any other value, o 3 would be the
mode.
Meaure of Variability
Several meaure are commonly ued to ummarize the variability of the data, inc
luding variance, tandard deviation, mean abolute deviation (MAD), and range.
Range A imple way to capture the variability or pread in the data i to take t
he range, the difference between the high and low value. All of the information
in the data i not being ued with thi meaure, but it i imple to calculate:
Range = x high x low = 8 3 = 5
Mean absolute deviation Another possible measure of the variability or spread in
the data is the average difference from the mean. However, for any data set thi
s will equal zero, because the values above the mean will balance the values bel
ow the mean. One way to eliminate this problem is to take the absolute value of
the differences from the mean. This measure is called the MAD and is commonly us
ed in forecasting to measure variability. However, absolute values are difficult
to work with mathematically:
MAD =
x x
n
=
2 +1+ 3+ 2 + 0 8 = = 1.6 5 5
Variance The average square difference from the meancalled the varianceprovides an
other measure of the variability in the data. For a population, variance is a
C h a p t e r 7 : U s i n g D a t a a n d St a t i s t i c a l To o l s f o r O
p e ra t i o n s I m p rove m e n t
185
good estimator. However, for a sample it can be proven that variance is a biased
estimator and needs to be adjusted; rather than dividing by n, the number of va
lues in the sample, it must be divided by n1: 4 + 1 + 9 + 4 + 0 18 = = 3.6 N 5 5 (
x x)2 = 4 + 1 + 9 + 4 + 0 = 18 = 4.5 2 Sample variance = s = n 1 51 4
2
Population variance = 2 =
(x )
=
Standard deviation Taking the square root of the variance will result in the uni
ts on this measure being the same as the units of the mean, median, and mode. Th
is measure, the standard deviation, is the most commonly used measure of variabi
lity.
Population standard deviation = 2 = =
(x )
N
2
4 + 4 + 0 + 1+ 9 18 = = 3.6 = 1.9 5 5
Sample standard deviation = s 2 = =
(x x)
n
2
18 4 + 4 + 0 + 1+ 9 = = 4.5 = 2.1 4 51
Coefficient of variation Given another data set (data B) with a standard deviati
on of 5 and the sample data set (data A) with a standard deviation of 1.9 and a
mean of 5, which data are more variable? This question cannot be answered withou
t knowing the mean of data B. If the mean of data B is 5, then data B is more va
riable. However, if the mean of data B is 25, data B is less variable. The coeff
icient of variation (CV) is a measure of the relative variation in the data.
Coefficient of variation = CV = or

1.9 = = 0.4 x 5
186
Probability and Probability Ditribution
Throughout hitory, people have tried to deal with the uncertaintie inherent in
the world. Myth, magic, and religion all played a role in explaining the variab
ility of the univere, and inexplicable event were attributed to a higher being
, fate, or luck. People have a need to feel in control, or at leat a deire to
predict how the world will behave. Two type of model exit to explain what i
een in the world: determinitic and probabilitic. In a determinitic model, th
e given input determine the output with certainty. For example, given a peron
date of birth and the current date, her age can be determined. The input determ
ine the output. Date of Birth Current Date Age Model Peron Age
In a probabilitic model, the given input provide only an etimate of the outpu
t. For example, given a peron age, her remaining life pan can only be etimate
d. Age Life Span Model Peron Remaining Life Span
Determination of Probabilitie
Probabilitie can be determined through obervation or experimentation; through
applying theory or reaon; or ubjectively, through opinion.
Oberved probability Oberved probability i a ummary of the obervation or ex
periment and i referred to a empirical probability or relative frequency. Ob
erved probability i the relative frequency of an eventthe number of time the ev
ent occurred divided by the total number of trial.
P(A) =
Number of time A occurred r = r Total number of obervation, trial, or experi
ment n
Drug or protocol effectivene i often determined in thi manner: Number of tim
e patient cured r = Total number of patient given the drug n h
P (Drug i effective) =
187
Theoretical probability The econd method of determining probability, the theore
tical relative frequency of an event, i baed on logicit i the theoretical numb
er of time an event will occur divided by the total number of poible outcome
.
P(A) =
Number of time A could occur r = m Total number of poible outcome n
Caino revenue are baed on thi theoretical determination of probability. If a
card i randomly elected from a common deck of 52, the probability that it wil
l be a pade i determined a follow: Number of pade in the deck 13 = = 0.25
Total number of card in the deck 52
P (Card i a pade) =
Opinion probability The final method for determining probability i baed on opi
nion. Opinion probability i a ubjective determination baed on an individual p
eronal judgment about how likely an event i to occur; it i not baed on a pre
cie computation but i often a reaonable aement by a knowledgeable peronit
i an opinion about the number of time an event will occur divided by the imag
inary total number of poible outcome or trial.
P(A) =
l Opinion of number of time an event will occur r = Theoretical total n
Handicapping hore race or porting event i an example of thi type of probab
ility determination:
P (Secretariat winning the Belmont Stake) = Opinion on the number of time Secr
etariat would win the Belmont r o = Imaginary total number of time the Belmont
would be run n
Odd are ometime ued to refer to thi type of probability. In the previou ex
ample, if the handicapper believe the odd of Secretariat winning the Belmont S
take are 20 to 3, thi mean that if the Belmont i run 23 time under conditio
n identical to the time in quetion, Secretariat would win 20 time and loe 3
time. No matter how they are determined, all probabilitie have certain propert
ie.
188
Propertie of Probabilitie
Bound on probability Probabilitie are bounded. The leat number of time an ev
ent could occur i zero; therefore, probabilitie mut alway be greater than or
equal to zero, and an event that cannot occur will have a probability of zero.
The larget number of time an event could occur i equal to the total poible
number of outcomet cannot be any larger; therefore, probabilitie mut alway be
le than or equal to one:
0 P(A) 1
The um of the probabilitie of all poible outcome i one. From thi property
, it follow that:
P (A) + P (A ) = 1 and 1 P (A ) = P (A )
where A i not A. Thi property can be ueful when determining probabilitie: O
ften, it i eaier to determine the probability of not A than to determine the pro
bability of A.
Multiplicative property Two event are independent if the outcome of one event d
oe not affect the outcome of the other event. For two independent event, the p
robability of both A and B occurring, or the interection () of A and B, i the p
robability of A occurring time the probability of B occurring:
P (A and B occurring) = P (A B) = P (A) P (B)
Combining a coin to with a die to, what i the probability of obtaining both
head and a three? 1 1 1 = 2 6 12
P (H 3) = P (H ) P (3) =
A tree diagram (Figure 7.9) or a Venn diagram (Figure 7.10) can be ued to illu
trate thi property. Note that deciion tree (Chapter 6) are different from the
tree diagram preented here. Deciion tree follow a time progreion and anal
ye of the choice that can be made at particular point in time.
189
Coin To
Die To 1 2 3 P(H 4 5 6
Probability 1/12 1/12 1/12 3) = 1/12 1/12 1/12 1/12 1/12 1/12 1/12 1/12 1/12 1/1
2
FIGURE 7.9 Tree Diagram Multiplicative Property
H
Start 1 2 3 T 4 5 6 P(H) = 1/2 P(3) = 1/6
To head
To 3
5
1
1
5
FIGURE 7.10 Venn Diagram Multiplicative Property
190
The multiplicative property provide a way to tet whether event are independen
t. If they are not independent,
P (A B) P (A) P (B).
Additive property For two event, the probability of A or B occurring, the union
() of A with B, i the probability of A occurring, plu the probability of B occ
urring, minu the probability of both A and B occurring:
P (A or B occurring) = P (A B) = P (A) + P (B) + P (A B)
Combining a coin to with a die to, what i the probability of obtaining head
or a 3, but not both?
P (H 3) = P (H ) + P (3) P (H 3) =
7 1 1 1 6 2 1 = + = + 2 6 12 12 12 12 12
A tree diagram (Figure 7.11) or Venn diagram can be ued to illutrate the addit
ive property.
Conditional probability In ome cae, it i poible to revie the etimate of
the probability of an event occurring if more information i obtained. For examp
le, uppoe a patient uually wait in the emergency department for fewer than 3
0 minute before being moved into an exam room. However, on Friday night, when
the department i buy, the wait i longer. Therefore, if it i a Friday night,
the probability of waiting for 30 minute or le i lower. Thi i the conditio
nal probability of waiting le than 30 minute given that it i a Friday night.
The conditional probability that A will occur given that B occurred i
P (A B) = P (A B) P (B)
Suppoe a tudy wa done of 100 emergency department patient; 50 patient were
oberved on a Friday night and 50 patient were oberved at other time. On Frid
ay night, 20 people waited fewer than 30 minute, but 30 people waited more than
30 minute. At other time, 40 people waited fewer than 30 minute, and only 10
people waited more than 30 minute. A contingency table (Table 7.1) i ued to
ummarize thi information. Contingency table are ued to examine the relation
hip between qualitative or categoric variable. A contingency table how the f
requency of one vari-
191
Coin To
Die To 1 2 3 P(H 4 5 6
Probability 1/12 1/12 1/12 3) = 1/12 1/12 1/12 1/12 1/12 1/12 1/12 1/12 1/12 1/1
2 7/12
FIGURE 7.11 Tree Diagram Additive Property
H
Start 1 2 3 T 4 5 6 P(H) = 1/2 P(3) = 1/6 P(H
3) = 7/12
30-Minute Wait Friday night Other time Total 20 40 60
> 30-Minute Wait 30 10 40 50 50 100
TABLE 7.1 Contingency Table for Emergency Department Wait Time
able a a function of another variable. The column an obervation i in ( or > 30
minute) i contingent upon (depend on) the row the ubject i in (time of day
). For all patient, the probability of waiting longer than 30 minute i:
P (Wait > 30 minute) =
Number of patient who wait > 30 Minute Total number of patient 40 = 100 = 0.4
0
192
Furthermore, the (conditional) probability of waiting more than 30 minute given
that it i Friday night i: P (Wait > 30 Minute and Friday night) P (Friday ni
ght) Number of Patient Who Wait > 30 Minute on Friday night Number of Patient
on Friday night 30 = 50 = 0.60 P (A Friday night) = A tree diagram for thi exa
mple i hown in Figure 7.12. Note that P (A B) = P (A B) P (B) = P (B A) P (A )
, and if one event ha no effect on the other event (the event are independent)
, then P (A B) = P (A) and P (A B) = P (A) P (B) . In the coin and dice example,
the coin to and die to are independent event, o the probability of toin
g a 6 i the ame no matter the outcome of the coin to. For the emergency depa
rtment wait time example, if night and wait time were independent, then the prob
ability of waiting fewer than 30 minute on a Friday night would be 0.5 0.6 = 0.
30. But thi i not true, wait time and night are not independentthey are related
. From thi imple tudy, it could not be concluded that Friday night caue wai
t time. Baye theorem allow the ue of new information to update the conditional
probability of an event. It i tated a follow:
P (A B) =
P (B A) P (A) P (A B) P (B A) P (A) = = P (B) P (B) P (B A) P (A) + P (B A ) P
(A )
Baye theorem i often ued to evaluate the probability of a falepoitive tet r
eult. If a tet for a particular dieae i performed on a patient, FIGURE 7.12
Tree Diagram Emergency Department Wait Time
Start 30 minute Other 30 minute 0.1 ^ 0.4
Night
Wait Time 30 minute
Probability 0.2 0.3
Conditional Probability 0.3/0.5 = 0.6
Friday 30 minute ^
193
there i a chance that the tet will return a poitive reult even if the patien
t doe not have the dieae. Baye theorem allow the determination of the probab
ility that a peron who tet poitive for a dieae actually ha the dieae. F
or example, if a teted patient ha the dieae, the tet report that with 99 p
ercent accuracy, and if the patient doe not have the dieae, the tet report
that with 95 percent accuracy; alo, uppoe that the incidence of the dieae i
rareonly 0.1 percent of the population ha the dieae:
P ( No dieae | Tet poitive ) =
P ( Tet poitive | No dieae ) P ( No dieae ) + P ( Tet poitive | Dieae
) P ( Dieae ) 0.050 0.999 = 0.981 0.050 0.999 + 0.990 0.001
P ( Tet poitive | No dieae ) P (No dieae)
A tree diagram (Figure 7.13) help to illutrate thi problem. The tet reult
are poitive 0.00099 + 0.04995 = 0.05094 of the time; 0.049955 of that time, the
peron doe not have the dieae. Therefore, the probability that a peron doe
not have the dieae, although the tet for the dieae wa poitive, i: 0.049
955 = 0.981, or 98.1 percent 0.05094
Probability Ditribution
A probability ditribution i a tabulation or function ued to ummarize knowled
ge about the random variable in quetion. It repreent the probability of an ev
ent aociated with all poible outcome. Probability ditribution can be ued
to model or imulate a ituation and can help to olve
Patient
Tet Reult Probability Poitive 0.990 Negative 0.010 Negative 0.950 Poitive 0.
050 0.00099 0.00001 0.94905 0.04995
Ha dieae 0.001 Start No dieae 0.999
FIGURE 7.13 Tree Diagram Baye Theorem Example
194
problem and anwer quetion. See Chapter 10 and 12 for example of the ue o
f probability ditribution.
Dicrete Probability Ditribution
Dicrete probability ditribution are aociated with dicrete random variable
. The variable can be decribed by an exact (only one) number or decription.
Binomial ditribution The binomial ditribution decribe the number of time th
at an event will occur in a equence of event. The event i binaryit will either
occur or it will not. The binomial ditribution i ued to determine whether an
event will occur, not the magnitude of the event. For example, in a clinical tr
ial the variable of interet might be whether the patient urvive. The binomial
ditribution decribe the probability of urvival, not how long the patient u
rvive. The binomial ditribution i pecified by the number of obervation, n,
and the probability of occurrence, p. An example often ued to illutrate the b
inomial ditribution i the toing of a coin. If a coin i toed three time,
the poible outcome are zero, one, two, or, three head. The probability of ob
taining thi reult i, repectively, 1/8, 3/8, 3/8, and 1/8. Figure 7.14 illut
rate the tree diagram aociated with thi example, and Figure 7.15 how the r
eulting probability ditribution. The binomial ditribution function i:
P(x) = n! p x ( 1 p) n x x!(n x)!
FIGURE 7.14 Tree Diagram Binomial Distribution
Toss 1
Toss 2 H
Toss 3 H T
Number Probability heads 0.125 0.125 0.125 0.125 0.125 0.125 0.125 0.125 3 2 2 1
2 1 1 0
H H T T Start H H T T H T T
C h a p t e r 7 : U s i n g D a t a a n d St a t i s t i c a l To o l s f o r O
195
0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0 0 1 2 3 Number of heads in three tosse
s
FIGURE 7.15 Binomial Distribution
where ! = factorial. Therefore, the probability of obtaining two heads in three
tosses can be calculated as: P (2 heads) = 3! 0.52 (1 0.5)32 2!(3 2) 3 21 = 0.5 0.5
(0.5)1 (2 1) (1)
Probability
= 3 0.53 = 0.375
Poisson distribution The Poisson distribution also takes on integer values: x =
0, 1, 2, 3, etc. It is used to model the number of events (e.g., number of telep
hone calls to a triage center, number of patients arriving at an emergency depar
tment) in a specific time period. The Poisson distribution is determined by one
parameter, , the average number of events in the given interva. The Poisson dist
ribution function is:
ex x!
P (x ) =
where e = the base of the natura ogarithm, 2.72. If patient arrivas at the em
ergency department foow a Poisson distribution, and the average number of pati
ents arriving at the department every hour is 3 (mean arriva rate = 3 patients/
hour), then the probabiity of 5 patients arriving woud be:
P (5) =
e x e 3 35 0.0498 243 = = = 0.10 x! 5 4 3 21 5!
196
Performance Improvement Toos, Techniques, and Programs
FIGURE 7.16 Poisson Distribution
Probabiity
0.25 0.20 0.15 0.10 0.05 0 1 2 3 4 5 6 7 8 9 10 11 Number of patient arrivas in
1 hour
Figure 7.16 iustrates the Poisson distribution when the mean arriva rate is 3
patients/hour.
Continuous Probabiity Distributions
Continuous probabiity distributions are associated with continuous random varia
bes, which can theoreticay take on an infinite number of possibe vaues. Con
tinuous random variabes are usuay measurements of some attribute (height, wei
ght, ength, etc.). For exampe, the height of a person coud be measured to the
nearest meter, centimeter, miimeter, and so on; a persons height coud be 2 me
ters, 2.10 meters, 2.105 meters, etc. Height coud be measured to an infinite nu
mber of possibe vaues. Continuous probabiity distributions are modeed by a s
mooth curve, where the tota area under the curve is one (P(x) 1), and the area
under the curve between two distinct points defines the probabiity for that int
erva. In a continuous probabiity distribution, the probabiity at a singe poi
nt is aways zero.
Uniform distribution The uniform distribution is the simpest continuous distrib
ution. The probabiity of occurrence is the same for a outcomes. This distribu
tion is usefu for modeing. The uniform distribution function is:
P (x ) =
1 for a x b b a (b a)2 b a = median, and = . 2 12
For the uniform ditribution =
Normal ditribution When people firt began oberving the occurrence of event a
nd contructing frequency diagram, a bell-haped ditribution wa often oberve
d. Thi
197
hape occurred o frequently that people began to expect it and it came to be ca

lled the normal ditribution. The normal ditribution i ymmetric and unimodal;
the mean i the ame a the mode and median. The normal ditribution wa common
ly oberved becaue many variable are ditributed approximately normally. Event
that are the aggregation of many maller but independent and poibly unoberv
able random event will follow a normal ditribution. For example, a peron weig
ht i the reult of many random event including uch thing a genetic, calori
c intake, and exercie. The central limit theorem (dicued later in thi chapt
er) i the baic principle underlying the occurrence of the normal ditribution.
Statitical proce control i baed in the central limit theorem and normal di
tribution (Chapter 8). The normal ditribution provide a reaonable approximat
ion of the ditribution of many random variable; it i well-behaved and mathemati
cally tractable. Additionally, many tatitical technique require that the indi
vidual data follow a normal ditribution. Fortunately, many of thee tet work
well even if the ditribution i only approximately normal, and ome tet work
even with ignificant deviation from normality. Finally, the normal ditributio
n can be decribed with two parameter: the mean () and the tandard deviation ().
The normal ditribution function i:
P(x) =
1 2
2
e (x ) 2 / 2
2
where = the population proportion. The normal ditribution i indicated by x ~ N
(, 2). The tandard normal ditribution, or z ditribution, i the normal ditribu
tion with a mean of zero and a tandard deviation of one. Any normal ditributio
n can be tranformed to a tandard normal ditribution by: z = ( x ) . Thi i c
alled the z core and repreent the number of tandard deviation away from the
mean for a particular value. For intance, if a peron cored 60 on a tet with
a mean of 50 and a tandard deviation of 10, hi z-core would be 1, becaue hi
core i 1 tandard deviation above the mean. Figure 7.17 illutrate ome nor
mal ditribution. Probabilitie aociated with z-core are widely available i
n table or tatitical oftware. Some often-ued z-core and their aociated
probabilitie are hown in Table 7.2. Even if the original ditribution i not n
ormal, omething i known about the probabilitie baed on the number of tandar
d deviation from the mean. Chebyhev rule tate that for any ditribution, the
proportion of value between k tandard deviation i at leat (1 1/k 2).
198
FIGURE 7.17 Normal Probability Ditribution
1.0 0.9 0.8 0.7 0.6 P(X) 0.5 0.4 0.3 0.2 0.1 0 5 4 3 2 1 0 X 1 2 3 4 5 = 0, = 0.5 = 0
, = 1.0 = 0, = 2.5 = 2, = 0.7
TABLE 7.2 z-Score Probabilitie
z-Score Limit 1z 2z 3z
Proportion Within the Limit if Ditribution I Normal 0.680 0.950 0.997
Proportion Within the Limit Even if Ditribution I Not Normal
0.750 0.890
Triangular ditribution The triangular ditribution i often ued in modeling an

d imulation becaue the needed parameter are relatively eay to etimate. The
triangular ditribution i decribed by the mode, minimum, and maximum value. I
t can be difficult for people to etimate the tandard deviation of a random var
iable but relatively eay for them to etimate the minimum and maximum expected
value. The triangular ditribution function i:
2(x a) (b a)(c a) for a x c P(x) = 2(b x) for c x b (b a)( b c)
C h a p t e r 7 : U s i n g D a t a a n d St a t i s t i c a To o s f o r O
199
2.5
FIGURE 7.18 Trianguar Distribution
Min = 0.0, Mode = 1.0, Max = 2.0 Min = 0.0, Mode = 0.5, Max = 2.0 Min = 0.0, Mod
e = 0.5, Max = 1.0
2
1.5 P(X) 1 0.5 0 0 0.5 1 X 1.5 2
As noted, many statistica tests require that the data be normay distributed.
The trianguar distribution (Figure 7.18) has properties simiar to the norma d
istribution and, therefore, does not vioate assumptions of normaity to a arge
degree. For the trianguar distribution, where b c = c a (symmetric), (b c )2 m
edian = mode and = c = median = mode and = . 6
Exponential ditribution The exponential ditribution i pecified by only one p
arameter and i memoryle. If an event follow an exponential ditribution with p
arameter l and ha not occurred prior to ome time, the ditribution of the time
until it doe occur alo follow an exponential ditribution with the ame para
meter, l. The exponential ditribution i ued in imulation to model waiting ti
me. For example, the time until the next patient arrival can be modeled with the
exponential ditribution. The exponential ditribution function i:
P (x ) = e x for x 0
where is the arriva rate in number/unit of time or distance and 1/ is the mean t
ime between arrivas. For the exponentia distribution = mean = 1/, median = n(2
)/, mode = 0, and = 1/. Figure 7.19 iustrates this distribution.
200
FIGURE 7.19 Exponentia Distribution
2.0 1.8 1.6 1.4 1.2 P(X) 1.0 0.8 0.6 0.4 0.2 0 0 1 2 3 X 4 5 6 = 0.5 = 1.0 = 2.0
Chi Square Test
The chi square test of independence is used to test the reationship between qua
itative or categoric variabes. For exampe, to compare staffing eves at vari
ous cinics, frequency data woud be coected and tabuated (Tabe 7.3). First,
the expected frequency for each ce is computed under the assumption that no d
ifference in staffing exists. The tabe shows that of the 40 empoyees, 20 work
at Cinic A. If there were no difference in staffing, the expected frequency for
the nurseCinic A ce woud equa the product of the tota number of nurses and
the proportion of staff at Cinic A, or 22 (20/40) = 11. The expected frequenci
es are shown in brackets in Tabe 7.3. The chi square test for independence is:
2 =
(E O )2 E (11 12)2 (9 8)2 (11 10)2 (9 10)2 = + + + 11 9 11 9 = 0.40
where E = the expected frequency and O = the observed frequency Higher vaues of
2 provide evidene that differenes in staffing exist aross linis. In this a
se, the hi-square value does not support the ontention that there are differen
es in staffing.
C h a p t e r 7 : U s i n g D a t a a n d St a t i s t i a l To o l s f o r O
201
Staff Type Clini A B Nurse 12 [(20 22)/40 = 11] 10 [(20 22)/40 = 11] 22 Dotor
8 [(20 18)/40 = 9] 10 [(20 18)/40 = 9] 18 Total 20 20 40
TABLE 7.3 Contingeny Table for Staffing Level and Clini
In a table with two rows and two olumns, the hi-square test of independene is
equivalent to a test of the differene between two sample proportions. In this
example, the question is whether the proportion of nurses differs as a funtion
of lini. Note that the test of the differene between proportions for these da
ta (see Hypothesis Testing, Proportions, later in this hapter) results in a z o
f 0.63, whih, when squared, equals 0.40, the hi-square value. A hi-square tes
t an also be performed when there are more than two rows or more than two olum
ns (greater number of ategories).
Confidene Intervals and Hypothesis Testing
Central Limit Theorem
The entral limit theorem states that as the sample size beomes large, the samp
ling distribution of the mean approahes normality, no matter the distribution o
f the original variable. Additionally, the mean of the sampling distribution is
equal to the mean of the population and the standard deviation of the sampling d
istribution of the mean approahes n , where i the tandard deviation of the po
pulation and n i the ample ize. Thi mean that if a ample i taken from any
ditribution, the mean of the ample will follow a normal ditribution with mea
n = and tandard deviation = n , commonly called the tandard error of the mean
( x or SE). The central limit theorem can be ued to determine a confidence inte
rval (CI) for the true mean of the population. If the tandard deviation of the
population i known, a CI for the mean would be:
x z /2 x x + z /2 x x z /2 n
x + z /2
n
202
where za/2 i the z value aociated with an upper- or lower-tail probability of
. In other words, to otin 95 percent CI, the upper- nd lower-til proili
ties would e 0.025 (2.5 percent in the upper til nd 2.5 percent in the lower
til, leving 95 percent in the middle), nd the ssocited z-vlue is 1.95 (2 i
s commonly used). Note tht incresing the smple size will tighten the confidence
limits. If the popultion stndrd devition () i unknown, the ample tandard
deviation () i ued to etimate the tandard error of the mean: x z / 2 x x +
/ 2 x x z / 2 s n x + z / 2 s n
Smll smples (generlly, n < 30) do not follow z distriution; they follow
t distriution. The t distriution hs greter proility in the tils of the d
istriution thn z distriution nd vries ccording to the degrees of freedom
, n 1, where n is the smple size. Therefore, for smll smples, the following e
qution is used: x t /2 s n x + t /2 s n
For exmple, if the witing time for rndom smple of 16 ptients ws mesured
nd their men wit time ws found to e 10 minutes with stndrd devition o
f 2 minutes, 95 percent CI for the true vlue of wit time would e: x t /2 10
2.13 s x + t /2 s n 2 16
n 2
16 10 1.06 10 + 1.06 8.94 11.06
10 + 2.13
This means that 19 out of 20 times, if a simiar sampe is taken, the CI obtaine
d woud incude the true vaue of the mean wait time. If a arger sampe of 49 p
atients had been taken and their mean wait time was found to be 10 minutes, with
a standard deviation of 1 minute, a 95 percent CI for the true vaue of the mea
n woud be:
C h a p t e r 7 : U s i n g D a t a a n d St a t i s t i c a To o s f o r O
203
x z / 2 10 2 1
s n
x + z / 2 1
s n
10 + 2 49 49 10 0.3 10 + 0.3 9.7 10.3
Hypothesis Testing
In the previous section, a range of ikey vaues for the popuation parameter o
f interest coud be obtained by computing a CI. This interva coud be used to d
etermine whether caims about the vaue were correct by determining whether the
CI captured the caimed vaue. In the wait time exampe, if it was caimed that
wait time for most patients was eight minutes, this caim woud be rejected base
d on the information obtained. However, if it was caimed that the mean wait tim
e was ten minutes, the study woud support this caim. Hypothesis testing is a f
orma way of testing such caims and is cosey reated to CIs. In hypothesis te
sting there is a beief, caed the nu hypothesis; a competing beief, caed
the aternate hypothesis; and a decision rue for evauating the beiefs. In the
wait time exampe, these woud be: Beief, or Ho: = 8 minutes Aternative beie
f, or Ha: 8 minutes Decision rue: If t t*, reject the nu hypothesis, where t
= ( x ) x , the number of tandard error away from the mean, and t* i the tet
tatitic baed on the deired confidence level and the degree of freedom. If
t i greater than t*, it would be unlikely to find a ample mean that i differe
nt from the true value of the mean; therefore, the belief about the true value o
f the mean (Ho) would be rejected. In the wait time example, t* for a 95 percent
CI with 15 degree of freedom (ample ize of 16) i 2.13. Therefore, t = ( x )
x = (10 8) 0.5 , and t t*. The beief (Ho) woud be rejected. More typicay, h
ypothesis testing is used to determine whether an effect exists. For instance, a
pharmaceutica company wants to evauate a new headache remedy by administering
it to a coection of subjects. If the new headache remedy appears to reieve h
eadaches, it is important to be abe to state with confidence that the effect wa
s reay due to the new remedy, not just chance. Most headaches eventuay go aw
ay on their own, and some headaches (or some peopes headaches) are difficut to
reieve, so the company can make two kinds of mistake: incorrecty concuding th
at the remedy works when in fact it does not, and faiing to notice that an effe
ctive remedy
204
works. The nu hypothesis (Ho) is that the remedy does not reieve headaches; t
he aternative hypothesis (Ha) is that it does.
Type I and Type II errors A Type I, or , error occurs if the compny concludes th
t the remedy works when in fct it does not. A Type II, or , error occurs if the
remedy is effective ut the compny concludes tht it is not. Hypothesis testin
g is similr to determining of guilt within the U.S. criminl court system. In
tril, it is ssumed the defendnt is innocent (the null hypothesis) until prov
en guilty (the lterntive hypothesis); evidence is presented (dt), nd the ju
ry decides whether the defendnt is guilty or not guilty sed on proof (decisio
n rule) tht hs to e eyond resonle dout (level of confidence). A jury d
oes not declre the defendnt innocent, just not guilty. If the defendnt is re
lly innocent ut the jury decides tht she is guilty, then it hs sent n innoce
nt person to jil (Type I error). If defendnt is relly guilty, ut the jury
finds him not guilty, criminl is free (Type II error). In the U.S. criminl c
ourt system, Type I error is considered more importnt thn Type II error, s
o Type I error is protected ginst to the detriment of Type II error. This
is nlogous to hypothesis testing (Kenney 1988), s illustrted in Tle 7.4. U
sully, the null hypothesis is tht something is not present, tht tretment h
s no effect, or tht no difference exists etween the effects of different tre
tments. The lterntive hypothesis is tht some effect is present, tht tretm
ent hs n effect, or tht there exists difference in the effects of different
tretments. Assuming the null hypothesis is true should mke it possile to com
pute the proility tht the test rejects the null hypothesis, given tht it is
true (Type I error.) The decision rule is sed on the proility of otining
smple men (or other sttistic) given the hypothesized men (or other sttis
tic). A comprison of wit time t two different clinics, on two different dys,
or during two different periods would use the following hypothesis test:
Ho: 1 = 2 H: 1 2 Decision rule: If t t*, reject Ho
(Note tht t* is usully determined with sttisticl softwre using the Stherw
ite pproximtion ecuse the two-smple test sttistic does not exctly follow
t-distriution.) Tle 7.5 illustrtes the errors ssocited with this exmple
.
C h p t e r 7 : U s i n g D t n d St t i s t i c l To o l s f o r O
p e r t i o n s I m p rove m e n t
205
Relity Assessment or guess Innocent Guilty Innocent Type I error Guilty Type II
error
TABLE 7.4 Type I nd Type II ErrorCourt System Exmple
Relity Wit times t the two clinics re the sme (1 = 2) Wit times t the two c
linics re not the sme (1 2) Type II error
Assessment or guess Wit times t the two clinics re the sme (1 = 2) Wit times
t the two clinics re not the sme (1 2)
TABLE 7.5 Type I nd Type II ErrorClinic Wit Time Exmple
Type I error
Equl vrince t-test If the wit time t two different clinics were of interest
, wit time for rndom smple of ptients from ech clinic might e mesured.
If wit time for smple of 10 ptients (for explntory purposes only) from e
ch clinic were mesured nd it ws determined tht Clinic A hd men wit time
of 12 minutes, Clinic B hd men wit time of 10 minutes, nd oth hd stn
drd devition of 1.5 minutes, the stndrd devitions could e pooled nd the d
istriution would follow t-distriution with (n1 + n2 2) degrees of freedom. A
t 95 percent confidence level,
(x 1 x 2 ) ( 1 2 ) sp 1 1 + n1 n 2 =
2 2 (n1 1)s 1 + (n 2 1)s 2 = 1.5 n1 + n 2 2
t=
where s p =
(12 10) (0) 1.5 1 1 + 10 10
2 = 2.99 t * = 2.10 0.67
Therefore, this test woud reject Ho, the beief that the mean wait time at the
two cinics is the same.
206
Aternativey, a 95 percent CI for the difference in the two means coud be foun
d: 1 1 + n1 n2
(x 1 x 2 ) t * s p 0.6 1 2 3.4
2 (2.10 0.67) 1 2 2 + (2.10 0.67)
Because the interva does not contain zero, the wait time for the two cinics is
not the same. Statistica software provides the p vaue of this test. The p va
ue of a statistica significance test represents the probabiity of obtaining va
ues of the test statistic that are equa to or greater than the observed test s
tatistic. For the wait time exampe, the p vaue is 0.015, meaning that Ho woud
be rejected with a confidence eve of up to 98.5 percent, or that zero woud n
ot be contained in a 98.5 percent CI for the mean. Smaer p vaues cause reject
ion of the nu hypothesis. t Tests can aso be used to examine the mean differe
nce between paired sampes and can be performed when the standard deviations of
the means differ. (See the companion website for more information on these types
of t tests.)
Proportions Reca the exampe in which staffing eves at two cinics were comp
ared.
Ho: 1 = 2 Ha: 1 2 Decision rule: If z z*, reject Ho The roortion of nurses at Clin
ic A was 12/20 = 0.60, and the roortion of nurses at Clinic B was 10/20 = 0.50
. The standard error of the difference in samle roortions is
(1 ) (1 ) + n1 n2
where =
n1 1 + n2 2 20(0.6) + (20)0.5 = = 0.55 n1 + n2 40
At a 95 ercent confidence level,
C h a t e r 7 : U s i n g D a t a a n d St a t i s t i c a l To o l s f o r O
e ra t i o n s I m rove m e n t
207
z=
(1 2 ) ( 1 2 ) (1 ) (1 ) + n1 n2
=
(0.60 0.50) (0) (0.55)(0.45) (0.55)(0.45) + 20 20
=
0.10 = 0.64 < t * = 1.96 0.157
Therefore, Ho could not be rejected and there could be no difference in the ro
ortion of nurses at each clinic. A CI for a roortion can be found from the fol
lowing: (1 ) n
z / 2 p + z / 2 p where p = p z / 2 p (1 ) + z / 2
p (1 ) n
A 95 ercent CI for the difference in the two roortions of nurses is (1 )
(1 ) + n1 n2
( 1 2 ) z
0.10 (1.96 0.157) 1 2 0.10 + (1.96 0.157) 0.2 1 2 0.41 Because the
ins 0, the roortion of nurses at the two clinics may not be different. The v
alue for this test is 0.53; therefore, Ho would not be rejected.
Practical versus statistical significance It is imortant to distinguish between
statistical and ractical significance. Statistical significance is related to
the ability of the test to reject the null hyothesis, whereas ractical signifi
cance looks at whether the difference is large enough to be of value in a racti
cal sense. If the samle size is large enough, statistical significance can be f
ound for small differences when there is limited or no ractical imortance asso
ciated with the finding. For instance, in the clinic wait time examle, if the m
ean wait time at Clinic A was 10.1 minutes, the mean wait time at Clinic B was 1
0.0 minutes, and the standard deviation for both was 1 minute, the difference wo
uld not be significant if the samle size at both clinics was 10; if, however, t
he samle size was 1,000, the difference would be statistically significant. The
Minitab
208
Performance Imrovement Tools, Techniques, and Programs
outut for this examle is shown in Box 7.1. Tests for statistical significance
should not be alied blindlythe ractical significance of a difference of 0.1 mi
nute is a judgment call. BOX 7.1 Statistical Significance of Differences Minitab
Outut for Clinic Wait Time Examle
Two Samle T Test and CI
Samle 1 2 N 10 10 Mean 10.10 10.00 SD 1.00 1.00 SEM 0.32 0.32
Difference = 1 2 Estimate for difference: 0.100000 95% CI for difference: (0.839561
, 1.039561) T test of difference = 0 (vs not =): T value = 0.22 P value = 0.826
df = 18 Both use Pooled SD = 1.0000
Two Samle T Test and CI
Samle 1 2 N 1000 1000 Mean 10.10 10.00 SD 1.00 1.00 SEM 0.032 0.032
Difference = 1 2 Estimate for difference: 0.100000 95% CI for difference: (0.01229
5, 0.187705) T test of difference = 0 (vs not =): T value = 2.24 P value = 0.025
df = 1998 Both use Pooled SD = 1.0000
ANOVA/MANOVA
One way ANOVA Analysis of variance (ANOVA) is the most commonly used statistical
method for testing hyotheses about three or more means. In the clinic waiting
time examle, if there were three clinics, days, or time eriods of interest (1,
2, and 3), 3 searate t tests (comaring 1 with 2, 1 with 3, and 2 with 3) woul
d be needed. If there were seven clinics, 21 searate t tests would be needed. T
his would be time consuming but, more imortant, flawed. In each t test there is
a 5 ercent chance of the conclusion being wrong (if the confidence level is 95
ercent)of the 21 tests, one would be exected to give a false result.
C h a t e r 7 : U s i n g D a t a a n d St a t i s t i c a l To o l s f o r O
e ra t i o n s I m rove m e n t
209
ANOVA overcomes this roblem by enabling the detection of significant difference
s over more than two treatments. A single test is used to determine whether differ
ences exist between the means at the chosen significance level. An ANOVA of the
means of two grous will give the same results as a t test. The term analysis of
variance is often a source of confusion. Desite the name, ANOVA is concerned wit
h differences between means of grous, not differences between variances. Basica
lly, ANOVA determines the variation within the grous and comares that to the d
ifferences between the grous, taking into account how many observations are in
each grou. If the observed differences are much larger than would be exected b
y chance, statistical significance exists.
Two way and higher ANOVA If the effect of more than one treatment or grouing va
riable is of interest, the ANOVA technique can be extended. For examle, higher
level ANOVAs could be used to investigate differences in waiting times on variou
s days of the week, at various time intervals throughout those days, or at sever
al different clinics. MANOVA/MANCOVA Multivariate analysis of variance (MANOVA)
is simly an ANOVA with several deendent variables. If the outcomes of interest
included atient satisfaction and service time in addition to wait time at diff
erent clinics, a MANOVA could be used to evaluate the differences. The MANOVA co
uld be used to test the hyothesis that some of these outcomes are different at
different clinics. Instead of a univariate test, a multivariate test would be us
ed. If the overall multivariate test is significant, the conclusion is that the
resective effect (e.g., clinic) is significant. The next question would of cour
se be whether just wait time is different between clinics, just atient satisfac
tion is different, or both are different. After obtaining a significant multivar
iate test for a articular main effect or interaction, one would examine the uni
variate tests for each variable to interret the resective effect. In other wor
ds, one would identify the secific deendent variables that contributed to the
significant overall effect. MANOVA has several advantages over ANOVA when there
is interest in several different deendent variables. First, measuring several d
eendent variables in a single exeriment allows a better chance of discovering
which factor is truly imortant. Second, MANOVA can rotect against Tye I error
s that might occur if multile ANOVAs were conducted indeendently. Third, it ca
n reveal differences not discovered by ANOVA tests. Analyis of covariance (ANCOV
A) and multivariate analysis of covariance (MANCOVA) are extensions of ANOVA and
MANOVA. They are simly
210
ANOVA or MANOVA in which the categoric deendent variables are adjusted for differ
ences in one or more covariates (metric variables). This can reduce error noise wh
en error associated with the covariate is removed.
Simle Linear Regression
Regression is a statistical tool used to model the association of a variable wit
h one or more exlanatory variables. The variables are tyically metric, althoug
h there are ways to analyze categoric variables using regression. The relationsh
i(s) can be described using an equation. Simle linear regression is the simle
st tye of regression. The equation reresenting the relationshi between two va
riables is Y = X+ + . Most v ryon will r m mb r Y = mX + b from l m ntary schoo
l. In statistics, is used for the intercept (the from elementry school), sign
ifies the slope (the m from elementry school; in sttistics m or represents the
men, so different vrile nme is used), nd e is the error. A simple exmp
le will help to illustrte the concept of regression. Assume tht the reltionsh
ip etween numer of dependents nd yerly helthcre expense ws of interest n
d the dt in Tle 7.6 were collected (for explntory purposes only, s lrg
er dt set would e needed for true regression nlysis). First, to visully
exmine the nture of the reltionship etween the vriles, sctter plot of
the dt (Figure 7.20) would e produced. From the sctter plot, it ppers tht
liner reltionship exists line could e drwn tht est represents the relt
ionship etween the two vriles. However, mny different lines could e drwn
to represent the reltionship (Figure 7.21). Wht is the est line? It would e c
curte (hve no is) nd hve the lest mount of error. Tle 7.7 contins the
predicted Y (Y ) nd the
TABLE 7.6 Dt for Regression Exmple: Reltionship Between Numer of Dependents
nd Yerly Helthcre Expense
Numer of Dependents 0 1 2 3 4
Annul Helthcre Expense ($1,000) 3 2 6 7 7
211
Annul helthcre cost ($1,000)
8 7 6 5 4 3 2 1 0 0 1 2 3 Numer of dependents 4 5
FIGURE 7.20 Sctter Plot Numer of Dependents vs. Annul Helthcre Costs
error (e = YPredicted YActul = Y Y ) for ech of the hypothesized lines shown i
n Figure 7.21. The line Y = 1.2(X) + 2 is isedthe estimtes re high nd the su
m () of rrors is gr at r than 0. In a good mod l, th rrors will sum to 0, m an
ing th r is no biasth stimat s ar n ith r high nor low on av rag . Although t
hr of th four mod ls rrors sum to z ro, which of thos is b st? Not that th
rrors sum to z ro b caus som ar positiv and som ar n gativ . Th b st mo
d l would hav th small st total absolut rror. B caus absolut valu s ar ma
th matically intractabl , th rrors ar squar d inst ad. Th b st-fitting lin
has a minimum sum of squar d rror.
10 9 Annual h althcar cost ($1,000) 8 7 6 5 4 3 2 1 0 0 1 2 3 4 5 6 Numb r of d
p nd nts Y = 1.2X + 2 Y=X+3 Y=5 Y = 1.3X + 2.4
FIGURE 7.21
catt r Plot with Possibl R lationship Lin s
212
P rformanc Improv m nt Tools, T chniqu s, and Programs
TABLE 7.7 Errors Associat d with th Various Lin ar Mod ls
X Y Y = 1(X) + 3 Y = 1.2(X) + 2 0 1 2 3 4 3 2 6 7 7 3 4 5 6 7 0 2 1 1 0 0 2.0 3.
2 4.4 5.6 6.8
1.0 1.2 1.6 1.4 0.2 3.0

Y = 1.3(X) + 2 2.4 3.7 5.0 6.3 7.6
0.6 1.7 1.0 0.7 0.6 0

Y = 0(X) + 2 5 5 5 5 5 2 3 1 2 2 0
TABLE 7.8
um of
quar d Errors Associat d with th Various Lin ar Mod ls
X
0 1 2 3 4
Y
3 2 6 7 7
Y = 1(X) + 3 3 4 5 6 7
2
0 4 1 1 0 6
Y = 1.3(X) + 3 2.4 3.7 5.0 6.3 7.6
2
0.36 2.89 1.00 0.49 0.36 5.10
Y = 0(X) + 5 5 5 5 5 5
2
4 9 1 4 4 22
Th lin Y = 1.3(X) + 2.4 has th low st squar d rror of all th mod ls (Tabl
7.8). To d t rmin th b st-fitting lin , or l ast squar d rror lin , Gauss pro
v d that if b = (Y Y)(X X) (X X) 2 and a = Y bX , then = 0 and 2 is a minimu
Int rpr tation
Th arli r lin ar mod l is int rpr t d as follows. Th slop of th lin indica
t s that with ach additional d p nd nt th annual cost of h althcar ris s by $
1,300; according to th int rc pt, wh n th r ar no d p nd nts th annual cost
of h althcar is $2,400. If th r w r no information, wh r X = 0 (no data), th
int rc pt would only hav math matic significanc .
Co ffici nt of D t rmination and Corr lation Co ffici nt
Th n xt qu stion is, How good is th mod l? Is th stimat b tt r wh n informati
on (X = numb r of d p nd nts) is us d to cr at knowl dg Y = 1.3(X) + 2.1? If n
o information w r availabl , th m an annual h althcar cost could b us d as a
n stimat of v ry individuals annual h althcar cost. Not that this is Y = 0(X
) + 6, an unbias d stimat (Tabl 7.9); and th rror wh n using th mod l coul
d b compar d to th rror using only th m an.
C h a p t r 7 : U s i n g D a t a a n d
t a t i s t i c a l To o l s f o r O
p ra t i o n s I m p rov m n t
213
% Improv m nt in rror =
M an rror Model error 22 5.1 = = 77% Mean error 22
There is a 77 ercent imrovement in error level. Another way to look at this wo
uld be to determine whether the error is exlained by the model (Figure 7.22): E
xlained error 16.9 = = 0.77 Total error 22
X 0 1 2 3 4 S
Y 3 2 6 7 7 25
Y 2.4 3.7 5.0 6.3 7.6 30.0
e 0.6 1.7 1.0 0.7 0.6 0.0
2 0.36 2.89 1.00 0.49 0.36 5.10
E
Y 5 5 5 5 5 25
Y Y 2 3 1 2 2 0
( Y Y )2 4 9 1 4 4 22
TO
Y Y 2.6 1.3 0.0 1.3 2.6 0.0
( Y Y )2 6.76 1.69 0.00 1.69 6.76 16.90
R
TABLE 7.9 R gr ssion
ums of
quar s
E =
um of squar s rror
TO =
um of squar s total
R =
um of squar s r gr
ssion
8 Annual h althcar cost ($1,000) 7 6 5 4 3 2 1 0 0 1 2 3 Numb r of d p nd nts 4
5 Explain d Un xplain d
FIGURE 7.22 Explain d and Un xplain d Error
214
P rformanc Improv m nt Tools, T chniqu s, and Programs
This m asur of how w ll th mod l fits th data is call d th co ffici nt of d
t rmination (r 2). Not that this is not a statistical t st, but rath r a m asur
of th p rc ntag of rror xplain d by th mod l. Th squar root of this num
b r is call d th corr lation co ffici nt (r). A n gativ corr lation co ffici n
t indicat s a n gativ slop , and a positiv corr lation co ffici nt indicat s a
positiv slop . Th corr lation co ffici nt is a m asur of th lin ar r lation
ship b tw n two variabl s, with a valu of on indicating p rf ct corr lation a
nd a valu of z ro indicating no r lationship. (R f r to Figur 7.8 for som sca
tt r plots and th ir corr lation co ffici nts.)
Probl ms with corr lation co ffici nts Th co ffici nt of d t rmination and th
corr lation co ffici nt ar both m asur s of th lin ar r lationship b tw n two
variabl s. A scatt r plot of th two variabl s should always b xamin d wh n i
nitially valuating th appropriat n ss of a mod l.
tatistical t chniqu s for j
udging th appropriat n ss of th mod l ar discuss d lat r in this chapt r. Do
s a low r 2 m an th r is no r lationship b tw n two variabl s? No. Figur 7.23
illustrat s two cas s (1 and 2) in which r 2 and r ar both z ro. In cas 1, th
r is no r lationship; in cas 2, th r is a r lationship, just not a lin ar r
lationship. Th r lationship can b p rf ctly captur d with th quation Y = + 1X
+ 2X 2, curve or qudrtic reltionship. Lter in the chpter, curve-type rel
tionships re discussed. A low r 2 my lso men tht other vriles needed to
explin the outcome vrile re missing from the model. Multiple regression might
fix this prolem. Does resonle or high r 2 men the model is good fit to
the dt? No. Figure 7.23 illustrtes severl cses in which the model is not
good fit to the dt. The r 2 nd r cn e hevily influenced y outliers, s i
n cses 4 nd 6. In cse 5, etter model would e curve. Alwys look t the
sctter plot of the dt. Does high r 2 men tht useful predictions will e o
tined with the model? No. Rememer the previous discussion of prcticl nd st
tisticl significnce. Finlly, does high r 2 men tht there is cusl rel
tionship etween the vriles? Nocorreltion is not custion. The oserved cor
reltion etween two vriles might e due to the ction of third, unoserved
vrile. For exmple, Yule (1926) found high positive correltion etween ye
rly suicides nd memership in the Church of Englnd. However, memership in th
e Church of Englnd did not cuse suicides.
Sttisticl Mesures of Model Fit
If there is no liner reltionship etween the two vriles, the slope of the
est-fitting line will e zero. This ide underlies the sttisticl tests for the
goodness of fit of the model.
215
Y
(1)
Y
(2)
Y
(3)
FIGURE 7.23 Low nd High r nd r 2 Exmple Plots
r = 0.91 X
r = 0.05 Y (4)
X
r = 0.00 Y (5)
X
Y
(6)
r = 0.75
X
r = 0.79
X
r = 0.56
X
F-test The F-test is hypothesis test of whether ll vlues in the model Y =
+ X + ar qual to z ro. In th cas of simpl lin ar r gr ssion, th r is only o
n b and this is a t st of wh th r is zero.
Ho: ll vlues = 0 H: ll vlues 0 Decision rule: If F* F(1;1;n2), reject Ho M
en squre regression MSR SSR 1 F* = = = Men squre error MSE SSE n 2 If the tw
o variables are related, the regression line will exlain most of the variance,
and SSR will be large in comarison to SSE. Therefore, large values of F* imly
that there is a relationshi and the sloe of the line is not equal to zero.
t Test For simle linear regression, the t test will give the same answer as the
F test; this is not the case for multile regression. The t test is a hyothesi
s test of whether a articular b = 0.
Ho: b = 0 Ha: b 0 Decision rule: If t* > t(1; n2), reject Ho t* = /s Alterntivel
y, CI for would e:
216
Performnce Improvement Tools, Techniques, nd Progrms
t(1
; n2)
s t(1 ; n2)s
If the intervl contins zero, Ho cn e rejected. Sttisticl softwre will pro
vide these tests for liner regression s well s r nd R2. Figure 7.24 shows th
e Microsoft Excel output for the helthcre expense exmple. FIGURE 7.24 Regress
ion Output for Helthcre Expense Exmple
SUMMARY OUTPUT Regression Sttistics
Multiple R R Squre Adjusted R Squre Stndrd Error Oservtions 0.8765 0.7682
0.6909 0.8790 5
ANOVA
df
Regression Residul Totl 1 2 3
SS
7.6818 2.3182 10.0000
MS
7.6818 0.7727
F
9.9412
Significnce F
0.0511
Coefficients
Intercept Y$1000 Annul Helth cre Expense 0.9545
Stndrd Error
1.0162
t Stt
0.9393
P-vlue
0.4169
Lower 95%
4.1885
Upper 95%
2.2794
Lower 90% Upper 90%
3.3460 1.4369
0.5909
0.1874
3.1530
0.0511
0.0055
1.1873
0.1499
1.0320
RESIDUAL OUTPUT
Predicted X Numer of Oservtion Dependents Residuls
1 2 3 4 5 0.8182 0.2273 2.5909 3.1818 3.1818 0.8182 0.7727 0.5909 0.1818 0.8182
PROBABILITY OUTPUT
Stndrd Residuls
1.0747 1.0150 0.7762 0.2388 1.0747
Percentile
10 30 50 70 90
XNumer of Dependents
0 1 2 3 4
Y$1000 Annul Helth Cre Expense Residul Plot 0.5000 0.0000 0.5000 1.0000 Y$1000 A
nnul helth cre expense 0 2 4 6 8
Norml Proility Plot 5 4 3 2 1 0 0 20 40 60 80 100 Smple percentile
XNumer of dependents
XNumer of Dependents Predicted X Numer of Dependents
Y$1000 Annul Helth Cre Expense Line Fit Plot 6 4 2 0 0 2 4 6 8 Y$1000 Annul he
lth cre expense
XNumer of dependents
1.0000
Residuls
217
Assumptions of Liner Regression
Liner regression is sed on severl principl ssumptions: The dependent nd i
ndependent vriles re linerly relted. The errors ssocited with the model
re not serilly correlted. The errors re normlly distriuted nd hve const
nt vrince. If these ssumptions re violted, the resulting model will e misl
eding. Vrious plots (nd sttisticl tests) cn e used to detect such prolem
s. These plots re usully provided y the softwre nd should e exmined for e
vidence of violtions of the ssumptions of regression. A sctter plot of the o
served versus predicted vlue should e symmetriclly distriuted round digo
nl line. A sctter plot of residuls versus predicted vlue should e symmetric
lly distriuted round horizontl line. A norml proility plot of the resi
duls should fll closely round digonl line. If there is evidence tht the
ssumptions of liner regression re eing violted, trnsformtion of the dep
endent or independent vriles my fix the prolem. Alterntively, one or two e
xtreme vlues my e cusing the prolems. Such vlues should e scrutinized clo
sely: Are they genuine (i.e., not the result of dt entry errors), re they exp
linle, re similr events likely to occur gin in the future, nd how influe
ntil re they in the modelfitting results? If the vlues re merely errors, or
if they cn e explined s unique events not likely to e repeted, there my
e cuse to remove them. In some cses, however, the extreme vlues in the dt m
y provide the most useful informtion out vlues of some of the coefficients
or provide the most relistic guide to the mgnitudes of prediction errors.
Trnsformtions
If the vriles re not linerly relted or the ssumptions of regression re v
iolted, the vriles cn e trnsformed to possily produce etter model. Tr
nsformtions re pplied to ensure tht the model is ccurte nd relile. If
person jogged to her doctors ppointment, she would need to wit efore hving
her lood pressure mesured if high reding would result in dignosis of hyp
ertension. Blood pressure vlues otined immeditely fter exercising re unsui
tle for detecting hypertension; the reson for witing is not to void the di
gnosis of hypertension ut to ensure tht high reding cn e elieved. It is
the sme with trnsformtions. Deciding which trnsformtion is est is often n
exercise in tril nd error in which severl trnsformtions re tried to see w
hich one provides the est model. Possile trnsformtions include squre root,
squre, cue, log, nd inverse. If the dt re trnsformed, this needs to e c
counted for when interpreting the findings. For exmple, imgine tht the origin
l vrile ws mesured in dys ut, to improve the model, n inverse trnsform
tion ws
218
pplied. Here,
is trnsformed
(dys). If the
re violted.
this cse.
it would e importnt to rememer tht the lower the vlue for th

vrile (1/dys), the higher the vlue of the originl vrile
dependent vrile is inry (0/1), the ssumptions of regression
The logit trnsformtion of the vrile, ln[p/(1 p)], is used in
Multiple Regression
In multiple regression, multiple independent vriles re used to predict sin
gle dependent vrile. A person might e descried s hving rown hir. If som
eone wnted to identify this person, he would hve difficulty ecuse this is no
t very good description. However, expnding the description to include more ch
rcteristics (femle, 57" tll, 140 pounds, lue eyes, wering red shirt) wo
uld mke it much esier to identify this unknown person. The sme is true of mul
tiple regression: More informtion cn improve the model nd llow for etter
description of the dependent vrile. The eqution for multiple regression is:
Y = + 1X1 + 2X2 + + kXk + wh r th r ar k ind p nd nt variabl s. In th h a
lthcar xp ns xampl , th r could b many oth r possibl variabl s r lat d to
h althcar costs in addition to numb r of d p nd nts; xampl s includ ag of d
p nd nts, ag of insur d, and amount of d ductibl . If th s variabl s w r add
d to th mod l, a b tt r mod l might b d v lop d. Th sam valuations of mod
l fit us d for simpl lin ar r gr ssion ar us d for multipl r gr ssion. H r ,
th F-t st is a t st of wh th r all vlues re zero, nd individul t-tests cn
e used to determine whether n individul independent vrile in the model is
significnt. If the overll model is significnt ut some individul 10.5 pt vl
ues re not, these predictors cn e removed from the model. A good model will
e significnt overll, nd ech individul predictor will lso e significnt. O
ne dditionl issue to e wre of in multiple regression is multicollinerity o
f the predictor vriles. If two predictor vriles re highly correlted to e
ch other, the explntory vlue of the model will e compromised. A simple exm
ple is shown in Tle 7.10. TABLE 7.10 Multicollinerity Exmple
Y 11 22
X1 1 2
X2 10 20
219
Here, Y = X1 + X2. But X1 nd X2 re perfectly correlted; X2 = 10 X1, so Y is
lso = 10X1 + 0.10X2 nd 5X1 + 0.6X2. In the first cse, X1 nd X2 hve the sme
level of effect on Y. In the other cses, the effects pper different. Multicol
linerity will cuse this prolem. In multiple regression, predictors re dded
to or removed from the model until the resercher is stisfied tht the model is
s good s it cn possily e with the given dt. This process cn e s much
of n rt s science nd usully requires expertise nd experience.
The Generl Liner Model
All of the techniques discussed thus fr re forms of the generl liner model (
GLM). It is the most generl of ll the models nd llows for multiple predictor
vriles, which my e metric, ctegoric, or oth, nd for multiple dependent
vriles, which my lso e metric, ctegoric, or oth. A complete discussion o
f GLM is outside the scope of this text, ut it cn e used to uild complex mod
els.
Additionl Techniques
Artificil Neurl Networks
Artificil neurl networks (ANN) ttempt to mimic the processes found in iologi
cl systems using the power of the computer. Although complete tretment of th
e theory underlying ANNs is eyond the scope of this ook, simple explntion
follows. ANN softwre tkes lrge mounts of dt, trnsforms nd weights the in
put vriles to lern from those dt, nd then predicts n outcome. Trditionl
models like regression nlysis cn use limited mounts of dt to investigte t
he nture of reltionships in the dt. In contrst, ANNs usully require signif
icnt mounts of dt nd do not offer n explntion s to how or why the indep
endent vriles ffect the dependent vriles; they simply predict the outcome
given wht hs hppened in the pst. ANNs hve een found to offer solutions to
prolems where trditionl models would e extremely complicted nd time consu
ming to uild. Tody, ANNs re eing successfully used in res such s imge re
cognition (Miller, Blott, nd Hmes 1992), dignosis (Dyowski nd Gnt 2001),
nd medicl decision mking (Bxt 1995; Weinstein et l. 1992).
Design of Experiments
Design of experiments (DOE) is structured pproch to the investigtion of s
ystem or process. DOE strts with the identifiction of input vriles elieved
to hve n effect on the output vrile of interest. Levels of ech input vri
le re chosen such tht the relevnt rnge of ech input vrile
220
is covered. An experimentl pln is developed tht determines the level of ech
input vrile for ech tril nd does not require ll possile levels of every
input vrile. This pln mximizes the informtion gined while minimizing the
numer of trils needed to otin it. DOE enles the investigtion of numer
of input vriles t the sme time nd llows the investigtor to move from ch
nging one vrile t time nd possily missing or misstting ny effects of
comintion of input vriles. DOE is typiclly used to find nswers to questi
ons of wht is cusing the prolem or how it cn e minimized. DOE is employed w
hen the investigtors hve resonly good ide of oth the importnt fctors
nd their relevnt rnges, typiclly fter initil investigtions hve een cond
ucted using the tools outlined previously in this chpter.
Conclusion
An outline for nlysis is shown in Tle 7.11, with its reltionships to the pl
n-do-check-ct (PDCA) process for continuous improvement (Chpters 8 nd 9), th
e define-mesure-nlyze-improve (DMAIC) process of Six Sigm (Chpter 8), nd t
he key elements of decision mking (Chpter 6).
TABLE 7.11 Outline for Anlysis
1. Define the prolem/question. 2. Determine wht dt will e needed to ddress
the prolem/question. 3. Collect the dt. 4. Grph the dt. 5. Anlyze the d
t using the pproprite tool. 6. Fix the prolem. 7. Evlute the effectiveness
of the solution. 8. Strt gin.
PDCA Pln Pln Do Do Do Do Check Pln
DMAIC Define Define Mesure Anlyze Anlyze Improve Control Define
Key Element Frme Frme Gther Gther Conclude Conclude Lern Frme
221
Which Technique to Use
The sttisticl tool or technique chosen to nlyze the dt depends on the type
of dt collected. Tle 7.12 provides guide for choosing the pproprite tec
hnique sed on dt type.
Independent Vrile Ctegoric One
Dependent Vrile Ctegoric Metric Mny Ctegoric Metric Both Ctegoric Metric
Both Mny Ctegoric Metric Both One Mny One Mny One Mny One Mny One Mny One
Mny One Mny One Mny
Mthemtic
Grphic
Metric
One
Both
Ctegoric Metric Both
One Mny One Mny
c2 c 2 (lyered) t-Test MANOVA c2 c 2 (lyered) ANOVA MANOVA GLM Logit GLM Simpl
e regression GLM MANCOVA Logit GLM Multiple regression GLM GLM ANN ANCOVA MANCOV
A Simple regression Multiple regression GLM ANN
TABLE 7.12 Vrile Type nd Choice of Technique
Histogrm type Box plot
Box plot
Sctter plot
222
1. The U.S. presidentil elections of 1936, 1948 (Hrris Poll Online 2006), nd
2000 (Wikipedi 2006) re good exmples of flwed studies tht led to flwed con
clusions. Wht ws the flw in ech of these studies? Discuss sitution from y
our personl experience in which study tht ws not vlid (logic, smpling, or
mesurement errors) led to flwed decision or conclusion. 2. John Allen Pulo
s (http://cnews.go.com/Technology/WhosCounting/) nd Jordn Ellenerg (www.sl
te.com/?id=3944&QueryText=Do+the+ Mth&Action=DeprtmentSrch&GroupBy=Deprtment)
oth write on numers, sttistics, nd proility. Red n rticle of interest
to you nd discuss. 3. Discuss how you would redesign report you receive t w
ork to mke it more useful for you. Would visul presenttion of the dt e h
elpful? How would you present the dt? 4. Discuss the difference etween correl
tion nd custion. 5. Discuss the difference etween sttisticl nd prcticl
significnce. 6. The lnced scorecrd, Six Sigm, Len, nd simultion employ
mny of the tools, techniques, nd tests found in this chpter. Discuss how, wh
ere, nd why prticulr tool would e used for ech.
Chpter Exercises
The following prolems use dt from three dt sets ville on the compnion
we site. Ech dt set contins the rw dt s well s reduced or reorgnized
dt for ese of nlysis. 1. Think of question, prolem, or issue in your org
niztion. Design vlid study to ddress the issue. Be sure to discuss how you
would ddress ll spects of vlidity, including logic, smpling, nd mesureme
nt. Assuming tht you cn collect vlid dt, how would you nlyze them? How wo
uld you present them? 2. Using the dt in Helth Insurnce Coverge.xls, compr
e insurnce coverge in Minnesot to coverge in Texs. . Anlyze the vlidity
of the dt. . Produce histogrm to compre the two sttes. Do Minnesot nd
Texs pper to hve similr coverge types? c. Produce Preto chrt for the t
wo sttes. Wht does this chrt indicte? d. Wht is the proility tht resi
dent of Minnesot or Texs will e uninsured? Insured? Insured y Medicre or Me
dicid? e. Wht is the 95 percent CI for the proportion of uninsured in Texs? I
n Minnesot? Wht is the 99 percent CI?
223
f. Wht is the 99 percent CI for the difference in the two proportions? g. Set u
p nd perform hypothesis test to determine if the proportion of uninsured diff
ers t 95 percent confidence level etween the two sttes. h. Perform chi-sq
ure test to determine if there re differences in proportions of types of helt
hcre coverge etween Minnesot nd Texs. i. Comment on the sttement, Living i
n Texs cuses more people to e uninsured. Wht other informtion might e helpf
ul here? 3. Use the dt in the World Helth.xls to nlyze worldwide life expec
tncy. Answer the following questions. . Discuss the vlidity of the dt. . C
onstruct histogrm, stem plot, dot plot, nd norml proility plot of CIA li
fe expectncy t irth (yers), totl for 2006. Wht do these grphs indicte? I
s this rndom vrile normlly distriuted? c. Construct ox plots of CIA life
expectncytotl, mle, nd femle. Wht do these grphs show? d. Determine the me
n, medin, mode, rnge, vrince, nd stndrd devition for CIA life expectnc
ytotl, mle, nd femle. Wht do these numers show? e. Wht is the 95 percent C
I for men life expectncy of mles nd femles? The 99 percent CI? f. Wht is t
he 99 percent CI for the difference in the two mens? g. Set up nd perform hy
pothesis test to determine if life expectncy for mles nd femles differs t
95 percent confidence level. h. Perform n ANOVA to determine if life expectnc
ies for mles, femles, nd ll re different. i. Construct sctter plots of CIA
gross domestic product, telephones, televisions, physicins, nurses, phrmcist
s, hospitl eds, nd per cpit helth expenditures versus CIA life expectncy
totl. Wht do these grphs indicte? j. Construct correltion mtrix for ll
the dt. Wht does this indicte? k. Construct histogrm nd norml proility
plots for CIA gross domestic product, televisions, nd hospitl eds. (Bonus: U
se BestFit to determine the est-fitting proility distriution for ech of th
ese nd totl life expectncy.) Are these rndom vriles normlly distriuted?
l. Perform three seprte simple liner regression nlyses for CIA gross domes
tic product, televisions, nd hospitl eds with CIA life expectncy totl. Inte
rpret your results. (Note: Excel will not perform regression nlysis when dt
re missing. The workook World Helth Regression hs eliminted countries for w
hich there re no dt on life expectncy. You my need to sort the dt nd run
the nlysis only on complete dt.)
224
m. Discuss the following sttement: World life expectncy could e incresed if e
veryone in the world owned television. n. Look t x-y sctter plots for ech p
ir of vriles in l. Do the reltionships pper to e liner? Would trnsfor
mtion of the x vrile improve the regression? o. Bonus: Build multiple regr
ession model with the ville dt to predict life expectncy in countries thr
oughout the world. Be sure to eliminte ny outliers nd multicollinerity. Trn
sform the dt s you deem est.
References
Bxt, W. G. 1995. Appliction of Artificil Neurl Networks to Clinicl Medicine.
Lncet 346: 113538. Dyowski, R., nd V. Gnt. 2001. Clinicl Applictions of Art
ificil Neurl Networks. Cmridge, U.K.: Cmridge University Press. Gould, S.
J. 1985. The Medin Isnt the Messge. Discover 6: 4042. Hrris Poll Online. 2006. Two
Gigntic Blunders in the History of Election Polling. [Online informtion; retri
eved 8/8/06.] www.hrrispollonline.com/uk/ history.sp#lunders. Kenney, J. M. 1
988. Hypothesis Testing: Guilty or Innocent. Qulity Progress 21 (1): 5557. Miller,
A., B. Blott, nd T. Hmes. 1992. Review of Neurl Network Applictions in Medic
l Imging nd Signl Processing. Medicl nd Biologicl Engineering nd Computin
g 30 (5): 44964. Pulos, J. A. 2006. Whos Counting: Flu Deths, Irqi Ded Numers
Skewed. ABCNews.com, Jnury 1. [Online informtion; retrieved 11/5/07.] http://
cnews.go.com/Technology/WhosCounting/story?id=1432589. Tufte, E. R. 1997. Visu
l Explntions: Imges nd Quntities, Evidence nd Nrrtive. Cheshire, CT: Gr
phics Press. . 1990. Envisioning Informtion. Cheshire, CT: Grphics Press. . 1983. Th
e Visul Disply of Quntittive Informtion. Cheshire, CT: Grphics Press. Wein
stein, J. N., K. W. Kohn, M. R. Grever, V. N. Viswndhn, L. V. Ruinstein, A.
P. Monks, D. A. Scudiero, L. Welch, A. D. Koutsoukos, A. J. Chius, K. D. Pull
. 1992. Neurl Computing in Cncer Drug Development: Predicting Mechnism of Acti
on. Science 258: 44751. Wennerg, J. E. 2005. Vrition in Use of Medicre Services
Among Regions nd Selected Acdemic Medicl Centers: Is More Better? [Online inf
ormtion; retrieved 6/20/06.] www.ihi.org/IHI/Topics/Improvement/Improvement Met
hods/Literture/VritioninuseofMedicreservicesmongregionsnd selectedcdemic
mediclcentersIsmoreetter.htm.
225
Wikipedi. 2006. United Sttes Presidentil Election, 2000. [Online informtion; r
etrieved 8/8/06.] http://en.wikipedi.org/wiki/U.S._presidentil_ election,_2000
. Yule, G. U. 1926. Why Do We Sometimes Get Nonsense-Correltions Between Time-Se
ries?A Study in Smpling nd the Nture of Time-Series. Journl of the Royl Stti
sticl Society 89 (1): 163.
CHAPTER
8
QUALITY MANAGEMENT: FOCUS ON SIX SIGMA
CHAPTER OUTLINE
Opertions Mngement in Action Overview Defining Qulity Cost of Qulity Qulit
y Progrms Totl Qulity Mngement ISO 9000 Bldrige Criteri Six Sigm Culture
Ledership Trining Define-Mesure-Anlyze-ImproveControl Define Mesure Anlyz
e Improve Control Seven Bsic Qulity Tools Sttisticl Process Control Rivervie
w Clinic SPC Process Cpility nd Six Sigm Qulity Riverview Clinic Process C
pility Rolled Throughput Yield Additionl Qulity Tools Qulity Function Depl
oyment Riverview Clinic QFD Tguchi Methods Benchmrking Pok-Yoke Riverview Cli
nic Six Sigm Generic Drug Project Define Mesure Anlyze Improve Control Conclu
sion Discussion Questions Chpter Exercises References
226

Bldrige criteri enchmrking cuse-nd-effect digrm, see fishone digrm ce
ntrl limit theorem (CLT) check sheet continuous qulity improvement (CQI) contr
ol limits cost of qulity count chrt (C-chrt) criticl to qulity (CTQ) defect
s per million opportunities (DPMO) define-mesure-nlyze-improvecontrol (DMAIC)
Design for Six Sigm (DFSS) filure mode nd effects nlysis (FMEA) fishone d
igrm flow chrt green elt, lck elt, mster lck elt histogrm house of q
ulity, see qulity function deployment Institute of Medicine (IOM) Interntion
l Orgniztion for Stndrdiztion (ISO) Ishikw digrm, see fishone digrm
ISO 9000 series Joint Commission key process input vriles (KPIV) key process
output vriles (KPOV) Mlcolm Bldrige Ntionl Qulity Awrd (MBNQA) mistke
proofing, see pok-yoke Preto chrt pln-do-check-ct (PDCA) pok-yoke process
cpility (Cp, Cpk) process mp or digrm proportion chrt (P-chrt) Qulity A
ssurnce Project (QAP) qulity function deployment (QFD) rdio frequency identif
iction (RFID) rnge chrt (R-chrt) rolled throughput yield (RTY) root-cuse n
lysis (RCA) run chrt smple men ( X , X-r) sctter digrm seven sic qul
ity tools sigm, stndrd devition Six Sigm (6) pecification limit tatitica
l proce control (SPC) tatitical quality control (SQC), ee SPC total quality
management (TQM) voice of the cutomer (VOC)
227

In 2001, Heritage Valley Health Sytem (HVHS) began to look beyond traditional h
ealthcare proce improvement technique to offet the iue facing it and many
other healthcare organization. Some of thee iue were patient care and ati
faction, finance, taff retention and recruitment, and collaboration with phyi
cian on clinical initiative. The HVHS operation team ultimately decided to de
velop a Six Sigma program. The mot critical quetion facing the organization wa
, How can an improvement methodology traditionally linked to manufacturing apply
to healthcare? A related and frequently voiced concern wa, How can we learn from
Six Sigma expert who might not undertand patient care? Becaue Six Sigma i ge
nerally aociated with manufacturing, there wa concern that the deployment and
implementation of the improvement methodology in a healthcare environment would
nt work. To addre thi concern, HVHS educated it executive leaderhip team on
Six Sigma and piloted a few project prior to rolling it out acro the ytem.
One black belt team wa charged with evaluating the patient admiion proce to
obervation or inpatient tatu. Within one month, the admiion proce impro
vement team identified multiple ource of lot revenue baed on how patient we
re being claified upon admiion. More than $1 million in revenue wa captured
, and many internal procee were improved a a reult of thi firt project. T
he effort, from the initial project meeting to final taff education and proce
change, took even month. The econd project involved a tudy of operating roo
m utilization. A team of four, meeting once a week for everal month, identifie
d that ue of the organization ambulatory urgery center would improve by tranf
erring typical outpatient urgerie (e.g., for ear, noe, throat; minor orthoped
ic problem; and cataract) from HVHS operating room. The extra capacity generat
ed for the hopital operating room wa then filled with inpatient-type urgerie
, where the revenue covered overhead cot. An additional outcome from thi proje
ct included a redeigned urology room, which helped the hopital avoid tandem c
heduling of an additional room in cae more extenive intervention wa required.
The poitive reult of thee hort-term project provided the proof the organi
zation needed to undertand that Six Sigma could indeed be applied to healthcare
ituation.
SOURCE: Beaver, R. 2004. Six Sigma Succe in Health Care. Quality Diget March: 3
134. Reprinted with permiion.
228
C h a p t e r 8 : Q u a l i t y M a n a g e m e n t : Fo c u o n S i x S i g m
a
229
Overview
Quality management became imperative for the manufacturing ector in the 1970 a
nd 1980, for ervice organization in the 1980 and 1990 and, finally, for the
healthcare indutry in the 1990, culminating with a 1999 Intitute of Medicine
(IOM) report, To Err I Human, that detailed alarming tatitic on the number
of people actually harmed by the healthcare ytem and recommended major improve
ment in quality a related to patient afety. The report recognized the need fo
r ytemic change and called for innovative olution to enure improvement in
the quality of healthcare. The healthcare indutry i facing increaing preure
to not only increae quality but alo to reduce cot. Thi chapter provide an
introduction to quality management tool and technique that are being uccef
ully ued by healthcare organization. The major topic covered include: Definin
g quality; The cot of quality; Quality program, including total quality manag
ement (TQM)/continuou quality improvement (CQI), ISO 9000, the Baldrige criteri
a and Six Sigma (note that Six Sigma program are different from Six Sigma tool
); Six Sigma tool and technique including the define-meaure-analyzeimprove-co
ntrol (DMAIC) proce, even baic quality tool, tatitical proce control (S
PC), and proce capability; and Other quality tool and technique, including q
uality function deployment (QFD), Taguchi method, and poka-yoke. After completi
ng thi chapter, reader hould have a baic undertanding of quality, quality p
rogram, and quality tool, enabling application of the tool and technique to
begin improving quality in their organization.
Defining Quality
Although mot people would agree that enuring quality in healthcare i of the u
tmot importance, many would diagree on exactly what quality mean. The upplying
organization perpective include performance (or deign) quality and conforman
ce quality. Performance quality include the feature and attribute deigned in
to the product or ervice. Conformance quality i concerned with how well the pr
oduct or ervice conform to deired goal or pecification.
230
From the cutomer perpective, Garvin (1987) define eight dimenion of product
quality: 1. 2. 3. 4. 5. 6. 7. 8. Performance: operating characteritic Feature
: upplement to the baic characteritic of the product Reliability: the prob
ability that the product will work over time Conformance: product conformance to
etablihed tandard Durability: length of time that the product will continue
to operate Serviceability: eae of repair Ethetic: beauty related to the look
or feel of the product Perceived value: perception of the product
Parauraman, Zeithaml, and Berry (1988) define five dimenion of ervice qualit
y: 1. 2. 3. 4. Tangible: phyical facilitie, equipment, and appearance of per
onnel Reliability: ability to perform promied ervice dependably and accurately
Reponivene: willingne to help cutomer and provide prompt ervice Aura
nce: knowledge and courtey of employee and their ability to inpire trut and
confidence 5. Empathy: care and individualized attention The 2001 IOM report Cro
ing the Quality Cham outline ix dimenion of quality in healthcare: afe,
effective, patient-centered, timely, efficient, and equitable. Finally, the Qual
ity Aurance Project (2006) outline nine dimenion of quality in healthcare:
technical performance, acce to ervice, effectivene of care, efficiency of
ervice delivery, interperonal relation, continuity of ervice, afety, phyi
cal infratructure and comfort, and choice. Obviouly, quality and it variou d
imenion may be viewed in many way. The dimenion relevant to one organizatio
n or project may not be relevant to a different organization. The particular qua
lity dimenion on which to focu, and the relative weight of thoe dimenion,
hould be determined taking into account both context and project.
Cot of Quality
The cot of qualityor the cot of poor quality, according to Juran and Godfrey
(1998)are the cot aociated with providing a poor-quality product or ervice.
Croby (1979) ay that the cot of quality i the expene of noncomformancethe co
t of doing thing wrong. Quality improvement initiative and project cannot be
jutified imply becaue everyone i doing it, but rather mut be jutified on the
bai of financial or ocietal benefit. However, Cleverley and Harvey (1992) f
ound that lower-quality hopital have lower level of profitability, and more
a
231
recent reearch (Midwet Buine Group on Health 2003) etimate that 30 percen
t of all direct healthcare cot in the United State today are the reult of po
or-quality care. According to the latter tudy etimate, $420 billion i pent
each year on overue, miue, and wate in the healthcare ytem. The cot of qu
ality i uually eparated into four part: 1. External failure: cot aociate
d with failure after the cutomer receive the product or ervice (e.g., bad pub
licity, liability, cutomer ervice) 2. Internal failure: cot aociated with
failure before the cutomer receive the product or ervice (e.g., rework, repla
cement, correction of problem) 3. Appraial: cot aociated with inpecting a
nd evaluating the quality of upplie or the final product/ervice (e.g., audit
, accreditation preparation, and review) 4. Prevention: cot incurred to elimin
ate or minimize appraial and failure cot (e.g., training, invetment in yte
m) Often, the cot of prevention are een a expene while the other, le ap
parent cot of appraial and failure are hidden in the ytem (Suver, Neumann,
and Bole 1992). However, it i uually le cotly to prevent quality problem
than to fix quality failure. Striving for continuou improvement in quality can
not only improve quality, but it can alo improve an organization financial it
uation.
Quality Program
Thi book focue on the Six Sigma methodology becaue of it popularity and gro
wing utilization, but there are many other, equally valid program for quality m
anagement and continuou improvement. TQM, ISO 9000 certification, the Baldrige
criteria, and Lean can all provide a framework for organizational improvement. A
lthough each of thee program or methodologie ha a lightly different focu,
their underlying principle are baically the ame.
Total Quality Management
Total quality management (TQM)or continuou quality improvement (CQI), a it i
ometime referred to in the healthcare arena (Shortell et al. 1995)i baically
imilar to Six Sigma. Baed on the philoophie of quality guru uch a Deming a
nd Juran, TQM ha been defined a a management philoophy that focue on meetin
g cutomer need and on continuou improvement. To accomplih thi, TQM relie o
n the active involvement of all functional area and all employee. Although the
re i no one definition of TQM, mot expert would agree on the underlying princ
iple that provide it framework (Hackman and Wageman 1995; Powell 1995; Ro an
d Perry 1999; Samon and Terziovki 1999; Tar 2005):
232
Focu on the cutomer. Quality i baed on atifying, and even delighting, the
cutomer. Top-management leaderhip and upport. Management i ultimately repon
ible for quality and mut communicate the quality viion to the organization an
d provide upport for it. Employee involvement. Employee are key to quality. Th
ey mut buy into the viion and be empowered to reach the goal. Sytem thinking
. Everything i related to everything ele. Quality will not be achieved without
cro-functional involvement. Continuou improvement. Quality i a moving goal.
To achieve it, the organization mut continuouly improve. Data-baed deciion
making. The tool and technique of quality management are needed to enure qual
ity.
ISO 9000
The International Organization for Standardization (ISO) 9000 erie (2006) i p
rimarily concerned with enuring that organization maintain conitently high l
evel of quality. The ISO tandard are proce oriented and require evidence of
outcome achievement. Many organization require that their vendor be ISO certi
fied, meaning that they have demontrated to an accredited regitrar (a certifie
d third-party organization) compliance with the requirement pecified in the t
andard(). Organization eeking certification mut implement formal policie, p
rocee, and procedure to enure that the need of their cutomer are being m
et. To do thi, all apect of their quality ytem need to be documented, moni
tored (or audited), and controlled o that continuou improvement can be made.
ISO 9000:2000 provide a quality management ytem that take into account the m
eaure, etting, ervice, and function of both clinical and adminitrative a
ctivitie within the healthcare indutry. ISO 9000-IWA1 provide baic definitio
n and guideline that can help healthcare provider elect and ue the appropri
ate tandard. ISO 9001:2000 tandard are ytem requirement, wherea ISO 9004:
2000 i a guideline that goe beyond the other tandard to help organization i
mprove the efficiency of their operation. The five ection of the ISO 9001:200
0 tandard are baed on quality management principle and define what hould be
done to conitently provide product and ervice that meet cutomer requiremen
t. Organization mut demontrate appropriate application of the tandard thro
ugh extenive documentation. The five ection of the ISO 9001:2000 tandard are
: 1. Quality management ytem. The etablihment and documentation of a quality
management ytem, including quality policy, quality goal, and a ytem of con
trol. 2. Management reponibility. Management mut continually analyze, review,
and improve the quality management ytem.
a
233
3. Reource management. Employee mut be qualified and trained for the job they
are doing. Other reource (facilitie, equipment, upport ervice) mut be u
fficient to produce quality product and ervice. 4. Meaurement, analyi, and
improvement. The organization will carry out inpection, teting, meaurement,
analyi, and improvement procee. 5. Product realization. The procee that
produce the organization product or ervice mut be well controlled. ISO 9000 pr
ovide a methodology and framework for enuring that an organization ha efficie
ntly and effectively defined, organized, integrated, and ynchronized it qualit
y management ytem to optimize performance and enure cutomer atifaction. Dr
. Paul M. Schyve (2000), enior vice preident of the Joint Commiion, believe
that accreditation, ISO 9000, and the Baldrige criteria can complement and augm
ent one another a part of a healthcare organization overall effort to achieve
performance excellence.
Baldrige Criteria
The Malcolm Baldrige National Quality Award wa etablihed to recognize U.S. or
ganization for their achievement in quality. The award ha raied awarene ab
out the importance of quality a a competitive priority and helped to dieminat
e bet practice in achieving world-cla quality by providing example of how t
o achieve quality and performance excellence. The Baldrige ytem (Baldrige Nati
onal Quality Program 2005) focue on continual improvement of key procee to
deliver exceptional cutomer value. Cutomer are key to the Baldrige concept; h
owever, employee are the one who actually deliver that exceptional value to th
e cutomer, and the Baldrige ytem alo place ignificant emphai on building
and retaining a killed, motivated, and atified workforce. Finally, quality i
not deired for quality ake, but rather it i needed to drive buine perfor
mance; the Baldrige criteria place almot 50 percent of their weight on outcome
. Thee outcome include trategie that lead to improved market performance, in
creaed market hare, cutomer retention, and atifaction. Organization are en
couraged to ue financial information uch a profit trend to analyze the conne
ction to overall performance. The Baldrige criteria conit of even categorie
of excellence (Figure 8.1) and have been ued by thouand of organization a
a elf-aement and improvement tool. Many organization do not even apply for
the award, but till ue the framework a an internal aement and improvemen
t tool. Healthcare organization that have received the award (Baldrige National
Quality Program 2006a) include: Bronon Methodit Hopital, Kalamazoo, Michigan
(2005); Robert Wood Johnon Univerity Hopital Hamilton, Hamilton, New Jerey
(2004); Baptit Hopital, Inc., and Saint Luke Hopital of Kana City (2003); a
nd SSM Healthcare, St. Loui (2002).
234
FIGURE 8.1 Baldrige Criteria for Performance Excellence Framework: A Sytem Per
pective
1 Leaderhip
Organizational profile: Environment, relationhip, and challenge
2 Strategic planning
5 Human reource focu 7 Reult
3 Cutomer and market focu
6 Proce management
4 Meaurement, analyi, and knowledge management
SOURCE: Baldrige National Quality Program (2006b).
Six Sigma
Six Sigma began in the 1980 at Motorola a the organization in-houe quality im
provement program. The program ha ince been adopted by many companie, and tra
ining in the Six Sigma methodology i offered by many conultant, including Mot
orola. Six Sigma ha been defined a a philoophy, methodology, et of tool, an
d goal. It i een a a buine trategy with a focu on eliminating defect th
rough prevention and proce improvement. The Six Sigma philoophy tranform th
e culture of the organization, and it methodology employ a project teambaed ap
proach to proce improvement uing the definemeaure-analyze-improve-control (D
MAIC) cycle. Six Sigma i compoed of a et of quantitative and qualitative tat
itically baed tool ued to provide management with fact to allow improvement
of an organization performance. Finally, Six Sigma alo repreent a goal of no
more than 3.4 defect per million opportunitie (DPMO). Six Sigma program can
take many form, depending on the organization, but the ucceful one all ha
re ome common theme: Top management upport for Six Sigma a a buine trate
gy; Extenive training at all level of the organization in the methodology and
ue of tool and technique;
a
235
Emphai on the DMAIC approach and ue of quantitative meaure of project ucce
; and Team-baed project for improvement that directly affect the organizatio
n financial well-being. The Six Sigma methodology ha been ued uccefully to
improve quality and buine performance in many healthcare organization, both
a a tand-alone program (Barry, Murcko, and Brubaker 2002; Ettinger 2001; Farre
ll and Sima 2005; Lazaru and Stamp 2002) and a an extenion of current CQI o
r TQM initiative (Revere and Black 2003).
Culture
Six Sigma, like all ucceful change initiative, both require and upport cu
ltural change within the organization. The culture of the organization can be th
ought of a it peronality, made up of the aumption, value, norm, and beli
ef of the whole of the organization member. It i evidenced by how tak are p
erformed, how problem are olved, and how employee interact with one another a
nd the outide world. Leader and employee of organization both hape and are
haped by the culture of the organization.
Leaderhip Six Sigma focue on leaderhip and the need for top management to u
pport and drive the initiative. Without thi upport, any initiative i doomed t
o failure. With both financial and ideological upport, Six Sigma can tranform
an organization. Training Succeful Six Sigma initiative require a high level
of proficiency in the application of the method qualitative and quantitative too
l and technique. To achieve thi, Six Sigma initiative involve an extenive a
mount of training at all level of the organization. A employee receive more t
raining and become more proficient, they are deignated a green belt, black be
lt, and mater black belt. Green belt typically undergo five day of claroo
m training covering quality management and control, problem olving, data analy
i, group facilitation, and project management. To obtain certification, they mu
t uually pa a written exam and uccefully complete and defend a Six Sigma
project. Green belt continue to perform their uual job in addition to Six Sig
ma project. Many organization have a goal of training all their employee to t
he green-belt level. Black belt have more Six Sigma project leaderhip experien
ce than green belt. They have more training in higher-level tatitical method
and mentor green belt. Black belt pend a limited amount of time, typically o
ne or two year, working on Six Sigma project throughout the organization and t
hen return to their uual job.
236
At the highet level are mater black belt, qualified to train green belt and
black belt and therefore given more extenive training in tatitical method a
well a communication and teaching kill. Mater black belt provide mentorh
ip and expertie to green and black belt in their organization. Thi training
ytem erve everal purpoe: It provide the organization with in-houe exper
t; enable everyone in the organization to peak the ame language, to underta
nd exactly what Six Sigma and Six Sigma project are all about; and, by uing bl
ack belt for a limited amount of time and then returning them to their uual po
ition within the organization, eed the organization with Six Sigma diciple
.
Define-Meaure-Analyze-Improve-Control
DMAIC i the acronym for the five phae of a Six Sigma project: define, meaure
, analyze, improve, and control. The DMAIC framework, or improvement cycle (Figu
re 8.2), i ued almot univerally to guide Six Sigma proce improvement proje
ct. DMAIC i baed on the plan-do-check-act continuou improvement cycle develo
ped by Shewart and Deming but i much more pecific. There are other improvement
and problem-olving framework, uch a the eight dicipline proce (Lazlo 20
03), but thi text focue on the more popular Six Sigma DMAIC proce.
Define In the definition phae, the Six Sigma team chooe a project baed on th
e trategic objective of the buine and the need or requirement of the cut
omer of the proce. The problem to be olved (or proce to be improved) i op
erationally defined in term of meaurable reult. Good Six Sigma project typica
lly have the following attribute:
FIGURE 8.2 DMAIC Proce
CONTROL DEFINE ACT PLAN
IMPROVE
CHECK
DO
MEASURE
ANALYZE
a
237
The project will ave or make money for the organization. The deired proce ou
tcome are meaurable. The problem i important to the buine, ha a clear rel
ationhip to organizational trategy, and i (or will be) upported by the organ
ization. A benchmarking tudy of project election (iSixSigma 2005) found that m
ot organization (89 percent of repondent) prioritize Six Sigma project on t
he bai of financial aving. The urvey alo found that the exitence of forma
l project election procee, proce documentation, and rigorou requirement
for project approval were all important to the ucce of Six Sigma project. In
the definition phae, internal and external cutomer of the proce are identi
fied and their critical to quality (CTQ) characteritic are determined. CTQ are
the key meaurable characteritic of a product or proce for which minimum per
formance tandard deired by the cutomer can be determined. Often, CTQ mut b
e tranlated from a qualitative cutomer tatement to a quantitative pecificati
on. In thi phae, the team alo define project boundarie and map the proce
(Chapter 5 and 6).
Meaure In the meaurement phae, the team determine the current capability and
tability of the proce. Key proce output variable (KPOV) are identified,
and valid, reliable metric are determined for them (ee Figure 8.3). The input
to the proce are identified and prioritized. Root-caue analyi (RCA) or fai
lure mode and effect analyi (FMEA) i ometime ued here to determine the ke
y proce input variable (KPIV). Valid, reliable metric are determined for th
e KPIV a well. A data collection plan for the proce i determined and implem
ented related to the KPIV and KPOV. The purpoe of thi phae of the project i
to etablih the current tate of the proce to evaluate the effect of any ch
ange to it. Analyze In the analyi phae, the team analyze the data that have
been collected to determine true root caue, or KPIV, to decide how bet to e
liminate variation or failure in the proce and improve the outcome.
FIGURE 8.3 Six Sigma Proce Metric
INPUT Key Proce Input Variable (KPIV) PROCESS Key Proce Output Variable (K
POV) OUTPUT Critical to Quality (CTQ) CUSTOMERS
238
Improve In the improvement phae, the team identifie, evaluate, and implement
the improvement olution. Poible olution are identified and evaluated in t
erm of their probability of ucceful implementation. A plan for deployment of
olution i developed, and the olution are implemented. Here, actual reult
hould be meaured to quantify the effect of the project. Control In the contro
l phae, control are put in place to enure that proce improvement gain are
maintained and the proce doe not revert to the old way. The improvement are in
titutionalized through modification of tructure and ytem (training, incent
ive, monitoring, etc.)
Seven Baic Quality Tool
The even fundamental tool ued in quality management and Six Sigma were firt
popularized by Kauro Ihikawa (1985), who believed that up to 95 percent of qual
ity-related problem could be olved with thee fundamental tool (ee Figure 8.
4): Fihbone diagram: tool for analyzing and illutrating the root caue of an
effect (Chapter 6). Check heet: imple form ued to collect data. Hatch mark a
re ued to record frequency of occurrence for variou categorie. Frequently, ch
eck heet data are ued to produce hitogram and Pareto chart. Hitogram: grap
h ued to how frequency ditribution (Chapter 7). Pareto chart: orted hitogr
am. Pareto chart are ued to eparate the vital few from the trivial many, bae
d on the idea that 80 percent of quality problem are due to 20 percent of caue
(Chapter 7). Flowchart: proce map (Chapter 6). Scatter plot: graphical techn
ique to analyze the relationhip between two variable (Chapter 7). FIGURE 8.4 S
even Quality Tool
Check heet Fihbone diagram Run chart
Hitogram Pareto chart Flowchart Scatter diagram
a
239
Run chart: plot of a proce characteritic, in chronologic equence, ued to ex
amine trend. A control chart, dicued in the next ection, i a type of run c
hart.
Statitical Proce Control
Statitical proce control (SPC) i a tatitic-baed methodology for determin
ing when a proce i moving out of control. All procee have variation in out
put, ome of it due to factor that can be identified and managed (aignable or
pecial) and ome of it inherent in the proce (common). SPC i aimed at dico
vering variation due to aignable caue o that adjutment can be made and bad
output i not produced. In SPC, ample of proce output are taken over time, m
eaured, and plotted on a control chart. From tatitic theory, the ample mean
will follow a normal ditribution. From the central limit theorem, 99.7 percen
t of ample mean will have a ample mean within 3 tandard error of the overall
mean and 0.3 percent will have a ample mean outide thoe limit. If the proce
i working a it hould, only three time out of 1,000 would a ample mean ou
tide the three tandard error limit be obtained. Thee three tandard error li
mit ( X ) are the control limit on a control chart. If the ample mean fall o
utide the control limit (or follow tatitically unuual pattern), the proce
i likely experiencing variation due to aignable or pecial caue and i ou
t of control. The pecial caue hould be found and corrected. After the proce
i fixed, the ample mean hould fall within the control limit, and the proc
e hould again be in control. Some tatitically unuual pattern that indicate
that a proce i out of control are hown in Figure 8.5. A more complete lit c
an be found in Sytma (1997). Often, the ample mean (X , called X-bar) chart i
ued in conjunction with a range (R) chart. R-chart follow many of the ame ru
le a X-bar chart and can be ued a an additional check on the tatu of a pr
oce. There are alo C-chart, ued when the meaured proce output i the cou
nt of dicrete event (e.g., number of occurrence in a day), and p-chart, ued
when the output i a proportion. There are alo more ophiticated type of con
trol chart that can be ued in healthcare organization (Lim 2003). A control c
hart could alo be et up uing individual value rather than ample mean. Howe
ver, thi i not often done for two reaon. Firt, the individual value mut b
e normally ditributed. Second, data collection can be expenive. It uually co
t le to collect ample of the data than to collect all of the data.
Riverview Clinic SPC The Riverview Clinic of Vincent Valley Hopital and Health
Sytem (VVH) i undertaking a Six Sigma project to reduce it waiting time. In
the meaurement phae of the project, data were collected on waiting time, and
240
FIGURE 8.5 Out-of-Control Pattern
One ample more than 3 tandard error from mean UCL = 3
X=0
LCL = 3 Obervation 8 or more ample above (or below) mean UCL = 3
X=0
LCL = 3 Obervation 14 or more ample ocillating UCL = 3
X=0
LCL = 3 Obervation 6 or more ample increaing (or decreaing) UCL = 3
X=0
LCL = 3 Obervation
a
241
Riverview decided that a control chart would be helpful for undertanding the cu
rrent ituation. Six obervation of waiting time were made over 20 day. At ran
domly choen time throughout each of 20 day, the next patient to enter the cli
nic wa choen. The time from when thi patient entered the clinic until he exit
ed wa recorded (Table 8.1). Riverview ued the tandard deviation of all of the
obervation to etimate the tandard deviation of the population. The three-i
gma control limit for the X-bar chart are: x z / 2 x x + z / 2 x x z /
n x + z / 2 4.4 s n
30 + 3 6 6 30 5.4 30 + 5.4 24.6 35.4
4.4
Day 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Observations of Wait Times (minutes) Observation 1 2 3 4 5 29 24 28 26 36 26 22
40 32 34 35 31 36 25 38 35 26 22 33 26 29 29 33 31 29 27 33 29 32 26 30 39 24 23
43 29 29 29 33 26 22 40 25 38 24 32 30 26 21 35 29 32 30 29 37 30 20 26 34 34 3
1 26 26 30 29 25 31 29 34 27 30 32 29 31 35 25 33 30 37 34 29 36 28 23 26 30 37
32 28 31 31 30 31 25 38 28 30 36 28 25
6 31 30 33 28 32 29 34 30 29 26 27 31 26 23 32 30 28 28 30 36
Samle Mean Samle Range 28.50 30.83 28.83 29.33 29.33 28.17 31.17 31.00 29.33 2
9.83 30.33 32.50 29.33 26.00 37.17 29.50 27.67 28.50 32.50 30.17 9 16 8 15 12 7
15 14 13 9 8 9 12 8 11 10 13 14 9 11
TABLE 8.1 Riverview Clinic Wait Times, in Minutes
Standard Deviation = 4.42
Overall Mean = 30.00
242
Mean wait time (minutes)
FIGURE 8.6 Riverview Clinic Wait Times: X Bar Control Chart
40
Out of control samle
35
30
+/ 1
+/ 2
+/ 3
25
20 0 5 10 15 Day 20 25 30
Looking at th control chart (Figur 8.6), it app ars that Day 15 was out of con
trol. An inv stigation found that on Day 15 th clinic was short-staff d b caus
of a school holiday. Th control chart cannot b us d as is b caus of th outof-control point. Riv rvi w could choos to continu to coll ct data until all p
oints ar in control, or it could r calculat th control chart limits, xcludin
g Day 15. Riv rvi w chos to r calculat , and th n w thr -sigma limits ar : x
z / 2 x x + z / 2 x x z / 2 30 3 0 s n x + z / 2 4.1 s n
30 + 3 6 6 30 5.0 30 + 5.0 25.0 35.0
4.1
Unless the system is changed, 50 ercent of Riverview atients will exerience a
wait time longer than 30 minutes (50 ercent will exerience a wait time of les
s than 30 minutes), and 10 ercent of Riverview atients will exerience a wait
time of greater than 35.3 minutes (90 ercent will exerience a wait time of les
s than 35.3 minutes).
x + z x x + z s; z 0.9 = 1.3 30 + ( 1.3 4.1) 30 + 5.3 35.3
C h p t e r 8 : Q u l i t y M n g e m e n t : Fo c u s o n S i x S i g m

243
If Riverviews gol for its Six Sigm project is to ensure tht 90 percent of pti
ents experience wit time of no more thn 30 minutes, the clinic needs to impr
ove the system. The Six Sigm tem would need to reduce men wit time to 24.7 m
inutes if the process vrition remins the sme (Figure 8.7).
x + z x x + z s ; z 0.9 = 1.3 30 24.7 + 5.3; x = 24.7
Process Cpility nd Six Sigm Qulity
Process cpility is mesure of how well process cn produce output tht me
ets desired stndrds or specifictions. This is done y compring the nturl (
or common) vriility of n in-control process, the process width, to the speci
fiction width. Specifictions re determined y outside forces (customers, mn
gement, etc.), ut process vriility is not determinedit is simply nturl p
rt of ny process. A cple process is process tht produces few defects, whe
re defect is defined s n output outside specifiction limits. There re two
common mesures of process cpility, Cp nd Cpk. Cp is used when the process i
s centered on the specifiction limits; the men of the process is the sme s t
he men of the specifiction limits. Cpk is used when the process is not centere
d. A cple process will hve Cp or Cpk greter thn 1. At Cp of 1, the pro
cess will produce out three defects per 1,000.
Cp = C pk
USL LSL USL LSL and is estimated by C = 6s 6 x LSL USL x = min or 3 3

x LSL USL x m or and is estimated by C k = min 3s 3
NOTE: USL = upper pecified limit and LSL = lower pecified limit
CURRENT WAIT TIME
50% of patient wait more than 30 minute
WAIT TIME GOAL 10% of patient wait more than 30 minute
FIGURE 8.7 Riverview Clinic Wait Time
15
20
25
30
35
40
45
15
20
25
30
35
40
45
Wait time (minute)
Wait time (minute)
244
Six Sigma quality i defined a fewer than 3.4 defect per million opportunitie
(DPMO). Thi definition can be omewhat confuing, a it actually correpond t
o the 4.5-igma one-tail probability limit for the normal ditribution. Six Sigm
a allow for a 1.5-igma hift in the mean of the proce and Cpk = 1.5 (Figure
8.8).
Riverview Clinic Proce Capability

Riverview Clinic management ha decided that no patient hould wait more than 40
minute, a waiting time upper pecification limit (USL) of 40 minute. The Six
Sigma team want to determine if the proce i capable. Note that becaue there
i no lower pecification limit (waiting time le than ome lower limit would
not be a defect), Cpk i the correct meaure of proce capability. USL x 40 30 10
C k = = = 0.81 = 3 3 4.1 12.3 The Cpk i le than 1. Therefore, the proce i
not capable, and 7,000 DPMO are expected [x N(30, 16.8), P(x > 40) = 0.007]. T
he team determine that to enure Six Sigma quality, the pecification limit wou
ld need to be 48.8 minute: USL x USL 30 48.8 30 USL = 48.8 C k = 1.5 = = 1
12.3 . 3
FIGURE 8.8 Six Sigma Proce Capability Limit
1.5 hift
Lower Specification Limit
Upper Specification Limit
3.4 DPMO
7
6
5
4
3
2
1
0
1
2
3
4
5
6
a
245
If the Riverview Six Sigma team determined that Six Sigma quality with a pecifi
cation limit of 40 minute wa a reaonable goal, it could reduce average wait t
ime, reduce the variation in the proce, or do ome combination of both. USL x U
SL x 40 21 C k = = 12.3 = 12.3 x = 21 3 10 USL x 40 30 C k = 1.5
3s 3 3 2.2 Average wait time would need to be reduced to 21 minute, or the t
andard deviation of the proce would need to be reduced to 2.2 minute to reach
the goal.
Rolled Throughput Yield
Rolled throughput yield (RTY) i a meaure of overall proce performance. It i
the probability that a unit (of product or ervice) will pa through all proce
tep free of defect. For example, conider a proce compriing four tep
or ubprocee. If each tep ha a 5 percent probability of producing an error
or defect (95 percent probability of an error-free outcome), the RTY of the over
all proce i 81 percentconiderably lower than that in the individual tep (Fi
gure 8.9).
Additional Quality Tool
In addition to the quality tool and technique commonly aociated with Six Sig
ma, many other tool can be ued in proce improvement. Quality function deploy
ment (QFD) and Taguchi method are often ued in the development of new product
or procee to enure quality outcome. However, they can alo be ued to impr
ove exiting product and procee. Benchmarking i ued to help to determine b
et practice and adapt them to the organization to achieve uperior performance
. Mitake-proofing, or poka-yoke, i ued to minimize the poibility of an erro
r occurring. All of thee tool are an eential part of an organization quality
toolbox. FIGURE 8.9 Rolled Throughput Yield
Step 1 100 in, 95 errorfree product out
246
Quality Function Deployment
QFD i a tructured proce for identifying cutomer need and want and tranla
ting them into a product or proce that meet thoe need. Thi tool i mot of
ten ued in the development phae of a new product or proce, but it can alo b
e ued to redeign an exiting product or proce. Typically, QFD i found in a
Deign for Six Sigma (DFSS) project, where the goal i to deign the proce to
meet Six Sigma goal. The QFD proce ue a matrix called the houe of quality
(Figure 8.10) to organize data in a uable fahion. The firt tep in QFD i to
determine cutomer requirement. Cutomer requirement repreent the voice of th
e cutomer (VOC) and are often tated in cutomer term, not technical term. Th
ere can be many cutomer for a particular product or ervice, and the VOC of al
l mut be heard. Market reearch tool are ued to capture the VOC. The many cu
tomer need found are organized into a few key cutomer requirement, which are
weighted baed on their relative importance to the cutomer. Typically, a cale
of one to five i ued, with five repreenting the mot important. The cutomer
requirement and their related importance are lited on the left ide of the QFD
diagram. A competitive analyi baed on the identified cutomer need i alo
performed. The quetion here i how well competitor meet cutomer need. Typica
lly, a cale of one to five i ued here a well, with five indicating that the
competitor completely meet thi need. The competitive analyi i ued to focu
the development of the ervice or product to determine where there are opporFIGURE 8.10 Houe of Quality
Correlation matrix Technical requirement
Cutomer requirement
Importance
Relationhip matrix
Competitive aement
Importance weight Specification or target value
a
247
tunitie to gain competitive advantage and where the organization i at a compet
itive diadvantage. Thi can help the development team to focu on important tr
ategic characteritic of the product or ervice. The competitor core on each
of the cutomer requirement are lited on the right ide of the QFD diagram. Te
chnical requirement of the product or proce that relate to cutomer requireme
nt are determined. For example, if cutomer want peedy ervice time, a relate
d technical requirement might be 90 percent of all ervice time be le than 20
minute. The technical requirement are lited horizontally acro the top of t
he QFD diagram. The relationhip between the cutomer requirement and technical
requirement i evaluated. Uually, the relationhip are evaluated a trong,
medium, or weak, and ymbol are ued to repreent thee relationhip in the re
lationhip matrix. Numeric value are aigned to the relative weight (5 = tro
ng, 3 = medium, 1 = weak), and thee value are placed in the matrix. Poitive a
nd negative interaction among the technical requirement are evaluated a tron
gly poitive, poitive, trongly negative, and negative. Symbol are ued to rep
reent thee relationhip in the roof, or correlation matrix, of the houe of qua
lity. Thi make clear the trade-off involved in product and proce deign. Cu
tomer importance weight are multiplied by relationhip weight and ummed for
each technical requirement to determine the importance weight. Target value ar
e developed baed on the houe of quality. Hitorically, QFD wa a phaed proce
. The aforementioned i the planning phae; for product development, planning i
followed by additional houe of quality related to part, proce, and produc
tion. For ervice development and improvement, uing only the firt houe of qua
lity or the firt houe of quality followed by the proce houe i often uffic
ient. For example of QFD application in healthcare environment, ee Chaplin a
nd Terninko (2000), Mazur, Gibon, and Harrie (1995), and Omachonu and Barach (
2005).
Riverview Clinic QFD Many diabete patient at the Riverview Clinic were not ret
urning for routine preventive exam. The team formed to addre thi problem dec
ided to ue QFD to improve the proce and began by oliciting the VOC via focu
group. The team found the following patient need and want:
To know (or be reminded) that they need to chedule a preventive exam; To know w
hy an office viit i needed; A convenient mean to chedule their appointment;
That their appointment be on time; and To know that their appointment will la
t a certain length of time.
Patient ranking of the importance of thee need and want wa determined via p
atient urvey.
248
A competitive analyi of Riverview Clinic two main competitor wa conducted ba
ed on the determined need and want of Riverview patient with diabete. The t
eam developed related technical requirement and evaluated the interaction betw
een them. The houe of quality developed i hown in Figure 8.11. On-time appoin
tment have the highet importance ranking becaue they affect both appointment
time and appointment length. The team then evaluated variou proce change and
improvement related to the determined technical requirement. To meet thee te
chnical requirement, they decided to notify patient via potcard and follow th
i with e-mail and phone notification if needed. The potcard and e-mail contain
ed information related to the need for an office viit and directed patient to
the clinic web ite for more information. Appointment cheduling wa made availa
ble via the Internet a well a by phone. Staffing level and appointment time
were adjuted to enure that appointment were on time and approximately the am
e length. Training wa conducted to help phyician and nure undertand the ne
ed to maintain contant appointment length and provide them with tool to enur
e conitent length. Table 8.2 outline thee proce change and related techni
cal requirement. After the change were implemented, the team checked to enure
that the technical requirement determined with the houe of quality were being
met. FIGURE 8.11 Riverview Clinic Houe of Quality for Patient with Diabete
+
Initial notification Subequent notification Information on need Time to chedul
e Appointment length range
+
On-time appointment Our ervice Competitor A Competitor B
Time knowledge Why knowledge Convenient Appointment length Appointment time
5 3 4 3 4
5
3 5 5 5 3 3 5 29 90%
3 3 3 2 2
5 5 3 3 4
3 3 3 3 3
25 Ye
15
15 3 channel
20 8 minute
27 8 minute
3
a
249
Technical Requirement Initial notification Subequent notification Information
on need Time to chedule Appointment length range On-time appointment
Proce Change Potcard mailed E-mail and/or phone call Web ite Web ite and ph
one Staff level adjuted Staff training
TABLE 8.2 Riverview QFD Technical Requirement and Related Proce Change
Taguchi Method
Taguchi method refer to two related idea firt introduced by Genichi Taguchi (
Taguchi, Chowdhury and Wu 2004). Rather than the quality of a product or ervice
being good or bad, where good i within ome pecified tolerance limit and bad
i outide thoe limit, quality i related to the ditance from ome target va
lue; further from the target i wore. Taguchi developed experimental deign tec
hnique (deign of experiment, or DOE; Chapter 7) where not only the target val
ue, but alo the aociated variation, i important. The optimal proce deign
i not necearily where the target value i maximized but where variation i mi
nimal in relation to the target. In other word, the proce i robut and will
perform well under le-than-ideal condition. Taguchi method are often ued in
DFSS where the product or ervice i deigned to be error free while meeting or
exceeding the need of the cutomer. Rather than fixing an exiting product or e
rvice, the deign proce of the product or ervice hould enure quality from t
he tart.
Benchmarking
According to the American Productivity and Quality Center (2005), benchmarking i
the proce of identifying, undertanding, and adapting outtanding practice a
nd procee from organization anywhere in the world to help your organization
improve it performance. Benchmarking i focued on how to improve any given proc
e by finding, tudying, and implementing bet practice. Thee bet practice
may be found within the organization, in competitor organization, and even in o
rganization outide the particular market or indutry. Bet practice are every
wherethe challenge i to find them and adapt them to the organization. The benchm
arking proce conit of deciding what to benchmark, determining how to meaur
e it, gathering information and data, and then implementing the bet practice wi
thin the organization. Benchmarking can
250
be an important part of a quality improvement initiative, and many healthcare or
ganization are involved in benchmarking (Biognano and Caldwell 1995). The jour
nal Healthcare Benchmark and Quality Improvement ha information on many of the
e initiative.
Poka-Yoke
Poka-yoke (a Japanee phrae meaning to avoid inadvertent error), or mitake-pr
oofing, i a way to prevent error from happening. A poka-yoke i a mechanim th
at either prevent a mitake from being made or make the mitake immediately ob
viou o that no advere outcome are experienced. For example, all of the intr
ument required in a urgical procedure are found on an intrument tray with ind
entation for each intrument. After the procedure i complete, the intrument
are replaced in the tray. Thi provide a quick viual check to enure that all
intrument are removed before cloing the patient inciion. Locating the contro
l for a mammography machine o that the technician cannot tart the machine unl
e he i hielded from radiation via a wall provide another example of mitak
e-proofing a proce. In FMEA, identified fail point are good candidate for po
ka-yoke. Technology can often enable poka-yoke. When patient data are input into
a ytem, the oftware i often programmed to provide an error meage if the d
ata are incorrect. For example, a Social Security number i nine digit long; no m
ore than nine digit can be input into the Social Security field, and an error m
eage appear if fewer than nine digit are entered. In the pat, urgical pon
ge were counted before and after a procedure to enure that none were left in a
patient. Now, the ponge can be radio frequency identification tagged, elimina
ting the error-prone counting proce, and a imple can can be ued to determin
e if any ponge remain in the patient.
Riverview Clinic Six Sigma Generic Drug Project
Management determined that a trategic objective for Riverview Clinic wa meetin
g pay-for-performance goal related to precribing generic drug, and a project
team wa organized to meet thi goal. Benchmarking wa done to help the team det
ermine which pay-for-performance meaure to focu on and define reaonable goal
for the project. The team found that 10 percent of nongeneric precription drug
could be replaced with generic drug and that other clinic had uccefully m
et thi goal.
Define
In the definition phae, the team articulated the project goal, cope, and bui
ne cae. Thi included developing the project charter, determining cu-
a
251
tomer requirement, and diagramming a proce map. The charter for thi project
i found in Chapter 5; it define the project goal, cope, and buine cae. T
he team identified the health plan and patient a being cutomer of the proce
. The output of the proce were identified a precription and the efficacy
of thoe precription. The proce input were phyician judgment and the info
rmation technology (IT) ytem for drug lit. Additionally, pharmaceutical firm
provide input on drug efficacy. The proce map developed by the team i hown
in Figure 8.12.
Meaure
The team determined that the percentage of generic (veru nongeneric) drug pre
cribed and the percentage of precription change after precription of a gener
ic drug would be ued to quantify the outcome. Additionally, the team decided t
hat it would need to track and record data on all nongeneric drug precribed by
each individual clinician for one month.
Analyze
After one month, the team analyzed the data and found that, overall, clinician
precribed 65 percent generic drug (Figure 8.13), and precription change were
needed for 3 percent of all precription. An example of the data collected i
hown in Table 8.3. The team generated a Pareto analyi by clinician and drug t
o determine if particular drug or clinician were more problematic. The analyi
howed that ome drug were more problematic but that all clinician were abou
t the ame (Figure 8.14). FIGURE 8.12 Riverview Clinic Precription Proce
Information on drug
Drug doent work
Patient need drug
Clinician precribe drug
Type of drug
Generic
Drug efficacy
Drug work End
Nongeneric Drug doent work Drug efficacy Drug work End
252
FIGURE 8.13 Riverview Generic Drug Project: Drug Type and Availability
Generic 20% 65% 35% 15% Nongeneric, generic available Nongeneric, generic not av
ailable
The team reexamined it tated goal of increaing generic drug precription by
4 percent in light of the data collected. If all precription for the top four
nongeneric drug for which a generic drug i available could be changed to gener
ic, Riverview would increae generic precription by 5 percent. Therefore, the
y decided that the original goal wa till reaonable.
TABLE 8.3 Riverview Clinic Six Sigma Generic Drug Project Example Data
Date 1-Jan 1-Jan 1-Jan 1-Jan 1-Jan 1-Jan 1-Jan 1-Jan 1-Jan 1-Jan . . . 31-Jan 31
-Jan 31-Jan 31-Jan 31-Jan 31-Jan 31-Jan 31-Jan 31-Jan 31-Jan
Clinician Smith Davi Jone Anderon Swanon Smith Swanon Jone Jone Swanon .
. . Anderon Anderon Davi Smith Jone Swanon Swanon Smith Davi Anderon
Drug F G L F R S U P S A . . . F E T Y D J I T G H
Drug Type Nongeneric Generic Generic Nongeneric Generic Nongeneric Generic Gener
ic Nongeneric Generic . . . Nongeneric Nongeneric Generic Generic Generic Generi
c Nongeneric Generic Generic Generic
Generic Available Ye Ye Ye No Ye Ye Ye Ye No Ye . . . Ye No Ye Ye Ye
Ye Ye Ye Ye Ye
Reprecribe No No No No Ye No No No No No . . . No No No No No No No No No No
a
253
Clinician Precription Nongeneric precription where there i a generic availa
ble/month 20 15 10 5 0 Davi Jone Smith Clinician Swanon Anderon
FIGURE 8.14 Riverview Clinic Generic Drug Project: Pareto Diagram
Non-Generic Precription Where There i a Generic Available Precription/Month
20 15 10 5 0 O W J V B H M A C I G Drug D K L T U N P Q R
Improve
The team conducted an RCA of the reaon for precribing nongeneric drug and de
termined that the major caue wa the clinician lack of awarene of a generic r
eplacement for the precribed drug. In addition to the IT ytem that will ident
ify approved generic drug, the team decided to publih, baed on data from the
previou month, a monthly top five lit of nongeneric drug for which an approve
d generic exit. The team continued to collect and analyze data after thee cha
nge were implemented and found that precription for generic drug had rien b
y 4.5 percent after 6 month.
Control
To meaure progre and enure continued compliance, the team et up a weekly co
ntrol chart for generic precription and continued to monitor and publih the t
op five lit. The team conducted an end-of-project evaluation to document the t
ep taken and reult achieved, and to enure that learning from the project wa
retained in the organization.
Concluion
The Six Sigma DMAIC proce i a framework for improvement. At any point in the
proce, it may be neceary to backtrack to enure improvement. For example, wh
at the team thought wa the root caue of the problem may
254
be found not to be the true root caue. Or, when attempting to analyze the data,
it may be that inufficient data, or the wrong data, were taken. In both cae,
it may be neceary to go back in the DMAIC proce to enure a ucceful proj
ect. At each top in the DMAIC proce, variou tool could be ued. The choice
of tool i related to the problem and poible olution. Table 8.4 outline ug
getion for when to chooe a particular tool or technique. Thi i only a guide
lineue whatever tool i mot appropriate for the ituation. TABLE 8.4 Quality To
ol and Technique Selector Chart
Tool or Technique 7 Quality control tool Caue-and-effect diagram Run chart Che
ck heet Hitogram Pareto chart Scatter plot Flowchart Other tool and technique
Mind-mapping/ braintorming 5 Why/RCA FMEA Turnip graph Pie chart Hypothei
teting Regreion analyi Control chart Proce capability QFD DOE/Taguchi met
hod Benchmarking Poka-yoke Gantt chart Project planning Charter Tree diagram S
imulation Force field analyi Balanced corecard
Define
Meaure
Analyze
Improve
Control
x x x x x
x x x x x
x x
x
x x x x x x x x x x
x x
x x
x
x
x x x x x x
x x x x x x x x x x x x
x x
x
x x
a
255
Dicuion Quetion
1. Read the Executive Summary of the IOM report To Err I Human (http://fermat.n
ap.edu/openbook.php?record_id=9728&page=1) and anwer the following quetion: W
hy did thi report pur an interet in quality management in the healthcare indu
try? What doe IOM recommend to addre thee problem? How much ha been done
ince 1999? 2. What doe quality in healthcare mean to your organization? To you
peronally? 3. Dicu a real example of each of the four cot of quality in a
healthcare organization. 4. Compare and contrat total quality management/conti
nuou quality improvement, ISO 9000, the Baldrige criteria, and Six Sigma. (More
information on each of thee program i linked from the companion web ite.) W
hich would you find mot appropriate to your organization? Why? 5. Think of at l
eat three poka-yoke currently ued in the healthcare indutry. Can you think o
f a new one for your organization?
Chapter Exercie
1. Clinician at VVH have been complaining about the turnaround time for blood w
ork. The laboratory manager decide to invetigate the problem and collect turn
around time data on five randomly elected requet every day for one month.
Obervation Day 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 1 44 28 54 57 30 42 26 54 46
49 53 75 74 91 59 2 41 32 83 53 50 40 17 39 62 71 64 43 19 40 32 3 80 58 59 63
62 50 50 39 53 34 12 43 52 66 59 4 51 42 50 15 68 49 47 82 64 42 35 50 55 15 49
5 25 18 46 52 42 73 91 28 57 43 43 64 59 73 71 Day 16 17 18 19 20 21 22 23 24 25
26 27 28 29 30 1 14 52 28 25 46 33 64 53 15 64 36 24 75 60 52 2 44 84 20 23 74
54 55 49 16 9 21 58 66 42 28 Obervation 3 35 55 67 35 24 62 62 72 18 51 51 19 3
4 20 85 4 52 63 76 21 10 40 14 49 35 47 40 88 27 59 39 5 76 15 69 23 47 27 72 61
78 70 57 16 71 60 67
256
a. Contruct an X-bar chart uing the tandard deviation of the obervation to
etimate the population tandard deviation. Contruct an Xbar chart and R-chart
uing the range to calculate the control limit. (The Excel template on the comp
anion web ite will do thi for you.) b. I the proce in control? Explain. c.
If the clinician feel that any time over 100 minute i unacceptable, what are
the Cp and Cpk of thi proce? d. What are the next tep for the laboratory ma
nager? 2. Riverview Clinic ha tarted a cutomer atifaction program. In addit
ion to other quetion, each patient i aked if he i atified with hi overal
l experience at the clinic. Patient can repond ye, they were atified, or no, th
ey were not atified. Typically, 200 patient are een at the clinic each day.
The data collected for two month are hown below.
Day 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Proportion of patient who were unatified 0.17 0.13 0.15 0.22 0.16 0.13 0.17 0
.17 0.11 0.16 0.15 0.17 0.17 0.12
Day 15 16 17 18 19 20 21 22 23 24 25 26 27
.15 0.12 0.16 0.18 0.14 0.17
Day 28 29 30 31 31 33 34 35 36 37 38 39 40
.17 0.15 0.15 0.15 0.14 0.19
a. Contruct a p-chart uing the collected data. b. I the proce in control? c
. On average, how many patient are atified with Riverview Clinic ervice? If
Riverview want 90 percent (on average) of patient to be atified, what hould
the clinic do next? 3. Think of a problem in your organization that Six Sigma c
ould help to olve. Map the proce and determine the KPIV, KPOV, CTQ, and ex
actly how you could meaure them. 4. Ue QFD to develop a houe of quality for t
he VVH emergency department (you may need to gue the number you do not know).
An Excel
a
257
template (QFD.xl) available on the companion web ite may be helpful for thi p
roblem.
Reference
American Productivity and Quality Center. 2005. Gloary of Benchmarking Term. [O
nline information; retrieved 1/30/06.] www.apqc.org/portal/apqc/kn/ Gloar y%2
0of%20Benchmarking%20Term.pdf?paf_gear_id=content gearhome&paf_dm=full&pageele
ct=contentitem&docid=119519. Baldrige National Quality Program. 2006a. 19882004 Aw
ard Recipient Contact and Profile. [Online information; retrieved 1/15/06.] www
.quality.nit.gov/ Contact_Profile.htm. . 2006b. 2006 Criteria for Performance Ex
cellence. Gaitherburg, MD: National Intitute of Standard and Technology. . 2005.
Frequently Aked Quetion about the Malcolm Baldrige National Quality Award. [On
line information; retrieved 1/15/06.] www.nit.gov/ public_affair/factheet/bal
dfaq.htm. Barry, R., A. Murcko, and C. E. Brubaker. 2002. Six Sigma Book for He
althcare: Improving Outcome by Reducing Error. Milwaukee, WI: ASQ Quality Pre
. Beaver, R. 2004. Six Sigma Succe in Health Care. Quality Diget March: 3134. B
iognano, M. A., and C. Caldwell. 1995. Bet Practice in Quality Leaderhip. The
Quality Letter for Healthcare Leader 7 (6): 1621. Center for the Study of Health
care Management. 2002. Deploying Six Sigma in a Healthcare Sytem. [Online white p
aper; retrieved 8/13/06.] www.hpm.umn.edu/ reearch/centerline/8260.pdf. Chaplin
, E. M. D., and J. Terninko. 2000. Cutomer Driven Healthcare: QFD for Proce I
mprovement and Cot Reduction. Milwaukee, WI: American Society for Quality Pre
. Cleverley, W. O., and R. K. Harvey. 1992. I There a Link Between Hopital Prof
it and Quality? Healthcare Financial Management 46 (9): 4044. Croby, P. B. 1979.
Quality I Free: The Art of Making Quality Certain. Boton: McGraw-Hill. Ettinge
r, W. H. 2001. Six Sigma: Adapting GE Leon to Health Care. Trutee 54 (8): 1015.
Farrell, B., and T. Sima. 2005. Tracking Quality Improvement in Health Care. Qua
lity Diget 25 (9). [Online information; retrieved 10/31/07.] www.quality diget
.com/ept05/article/02_article.html. Garvin, D. A. 1987. Competing on the Eight
Dimenion of Quality. Harvard Buine Review 65 (6): 10110. Hackman, J. R., and
R. Wageman. 1995. Total Quality Management: Empirical, Conceptual, and Practical
Iue. Adminitrative Science Quarterly 40 (2): 30942.
258
Intitute of Medicine. 2001. Croing the Quality ChamA New Health Sytem for th
e 21t Century. Wahington, D.C.: National Academie Pre. . 1999. To Err I Human
: Building a Safer Health Sytem. Wahington, D.C.: National Academie Pre. In
ternational Organization for Standardization. 2006. ISO Homepage. [Online informat
ion; retrieved 1/15/06.] www.io.org/io/en/ISOOnline.frontpage. Ihikawa, K. 19
85. What I Total Quality Control? Tranlated by D. J. Lu. Englewood Cliff, NJ:
Prentice-Hall. iSixSigma. 2005. Six Sigma Project Selection: How Organization C
hooe the Mot Important Project. [Online information; retrieved 1/10/06.] www.i
ix igma.com/library/content/df.ap. Juran, J. M., and A. B. Godfrey. 1998. J
uran Quality Handbook, 5th ed. New York: McGraw-Hill. Lazlo, C. 2003. The Suta
inable Company: How to Create Lating Value Through Social and Environmental Per
formance. Wahington, D.C.: Iland Pre. Lazaru, I. R., and B. Stamp. 2002. Th
e Promie of Six Sigma (Special Report). Managed Healthcare Executive 12 (1): 2730
. Lim, T. O. 2003. Statitical Proce Control Tool for Monitoring Clinical Perf
ormance. International Journal for Quality in Health Care 15 (1): 34. Mazur, G., J
. Gibon, and B. Harrie. 1995. QFD Application in Health Care and Quality of Wo
rk Life. Paper preented at the Firt International Sympoium on QFD, Tokyo, Japa
n, November 11. Midwet Buine Group on Health. 2003. Reducing the Cot of Po
or-Quality Health Care Through Reponible Purchaing Leaderhip. Chicago: Midwe
t Buine Group on Health, PPNEW04. Omachonu, V., and P. Barach. 2005. QFD in a
Managed Care Organization. Quality Progre 38 (11): 3641. Parauraman, A., V. A.
Zeithaml, and L. L. Berry. 1988. SERVQUAL: A MultipleItem Scale for Meauring Co
numer Perception of Service Quality. Journal of Retailing 64 (4): 1240. Powell,
T. C. 1995. Total Quality Management a Competitive Advantage: A Review and Empir
ical Study. Strategic Management Journal 16 (1): 1537. Quality Aurance Project.
2003. Dimenion of Quality. [Online information; retrieved 1/15/06.] www.qaprojec
t.org/method/redimenion.html. Revere, L., and A. K. Black. 2003. Integrating S
ix Sigma with Total Quality Management: A Cae Example for Meauring Medication
Error. Journal of Healthcare Management 48 (6): 37791. Ro, J. E., and S. Perry.
1999. Total Quality Management: Text, Cae and Reading. Oxford, UK: CRC Pre
. Samon, D., and M. Terziovki. 1999. The Relationhip Between Total Quality Man
agement Practice and Operational Performance. Journal of Operation Management 1
7 (4): 393409.
a
259
Schyve, P. M. 2000. A Trio for Quality: Accreditation, Baldrige and ISO 9000 Can
Play a Role in Reducing Medical Error. Quality Progre 33 (6): 5355. Shortell, S
. M., J. L. OBrien, J. M. Carman, R. W. Foter, E. F. Hughe, H. Boertler, and E
. J. OConnor. 1995. Aeing the Impact of Continuou Quality Improvement/Total Q
uality Management: Concept Veru Implementation. Health Service Reearch 30 (2)
: 377401. Suver, J. D., B. R. Neumann, and K. E. Bole. 1992. Accounting for the C
ot of Quality. Healthcare Financial Management 46 (9): 2837. Sytma, S. 1997. Qua
lity and Statitical Proce Control. [Online information; retrieved 2/4/06.] www
.ytma.com/tqmtool/ctlchtprinciple.html. Taguchi, G., S. Chowdhury, and Y. Wu
. 2004. Taguchi Quality Engineering Handbook. Hoboken, NJ: Wiley-Intercience. T
ar, J. J. 2005. Component of Succeful Total Quality Management. The TQM Magazine
17 (2): 18294.
CHAPTER
9
THE LEAN ENTERPRISE
CHAPTER OUTLINE
Operation Management in Action Overview What I Lean? Type of Wate Kaizen Val
ue Stream Mapping Vincent Valley Hopital and Health Sytem Value Stream Mapping
Meaure and Tool Takt Time Throughput Time Riverview Clinic Five S Spaghetti
Diagram Kaizen Blitz or Event VVH Kaizen Event Standardized Work Jidoka and And
on Kanban Single-Minute Exchange of Die Flow and Pull Heijunka and Advanced Acce
Concluion Dicuion Quetion Chapter Exercie Reference
260

% value added advanced acce cheduling andon care path computed tomography (C
T) continuou quality improvement coronary artery bypa graft (CABG) cycle time
echocardiograph (echo) five S heijunka jidoka Joint Commiion jut in time ka
izen blitz, ee kaizen event kaizen event kaizen philoophy kanban Lean Sigma m
uda neceary activity nonvalue added operating room (OR) ingle-minute exchange
of die (SMED) Six Sigma paghetti diagram tandardized work takt time throughput
time total quality management (TQM) Toyota Production Sytem (TPS) type of wa
te unneceary activity value added value tream mapping work-in-proce (WIP)
261

The author of Going Lean in Healthcare provide the following example of Lean ma
nagement principle applied in healthcare: Virginia Maon Medical Center in Seat
tle, Wahington, ha been uing lean management principle ince 2002. By workin
g to eliminate wate, Virginia Maon created more capacity in exiting program
and practice o that planned expanion were crapped, aving ignificant capit
al expene: $1 million for an additional hyperbaric chamber that wa no longer
needed; $1 to $3 million for endocopy uite that no longer needed to be reloca
ted; $6 million for new urgery uite that were no longer neceary. Depite a
no-layoff policy, a key tenet of Lean management, taffing trend at Virginia Ma
on howed a decreae in 2003 and 2004, after ix year of annual increae in t
he number of full-time equivalent (FTE). Uing Lean principle, taff, provide
r, and patient have continuouly improved or redeigned procee to eliminate
wate, requiring fewer taff member and le rework, and reulting in better q
uality. Conequently, a employee retire or leave for other reaon, improved p
roductivity allow for them not to be replaced. All 5,000 Virginia Maon employe
e are required to attend an Introduction to Lean coure, and many have particip
ated in Rapid Proce Improvement Week (RPIW). RPIW are intenive weeklong e
ion in which team analyze procee and propoe, tet, and implement improvem
ent. The reult from the 175 RPIW that were conducted from January 2002 throu
gh March 2004 are hown in Table 9.1.
SOURCE: Womack et al. (2005).
TABLE 9.1 Reult of 175 RPIW at Virginia Maon Medical Center
Category Inventory Productivity Floor pace Lead time People ditance Product di
tance Setup time
2004 Reult (After 2 Year of Lean) $1,350,000 158 22,324 23,082 Traveled 267,793
Traveled 272,262 7,744
Metric Dollar FTE Sq. Ft. Hour Feet Feet Hour
Change from 2002 Down 53% 36% redeployed to other open poition Down 41% Down 6
5% Down 44% Down 72% Down 82%
SOURCE: Womack et al. (2005). Ued with permiion.
262
C h a p t e r 9 : T h e Le a n E n t e r p r i e
263
Overview
Lean tool and technique have been employed extenively in manufacturing organi
zation ince the 1990 to improve the efficiency and effectivene of thoe org
anization. More recently, many healthcare organization have begun to realize t
he tranformative potential of Lean (Panchak 2003; Womack et al. 2005). The heal
thcare indutry i facing increaing preure to employ reource in an effectiv
e manner to reduce cot and increae patient atifaction. Thi chapter provide
an introduction to the Lean philoophy a well a the variou Lean tool and t
echnique ued by many healthcare organization today. The major topic covered
include: The Lean philoophy; Defining wate; Kaizen; Value tream mapping; and
Other Lean tool, technique, and idea, including the five S, paghetti diagra
m, kaizen event, takt time, kanban, ingle-minute exchange of die, heijunka,
jidoka, andon, tandardized work, ingle-piece flow, and pull.
After completing thi chapter, reader hould have a baic undertanding of Lean
tool, technique, and philoophy. Thi hould help them undertand how Lean co
uld be ued in their organization and enable them to begin to employ it tool
and technique to facilitate continuou improvement.
What I Lean?
A decribed in Chapter 2, Lean production wa developed by Taiichi Ohno, Toyota
chief of production after World War II. The Toyota Production Sytem (TPS) wa
tudied by reearcher at MIT and documented in the book The Machine that Change
d the World (Womack, Jone, and Roo 1990). TPS wa much more effective and effi
cient than traditional ma production, and the MIT reearcher coined the term
Lean production to capture thi radically different approach. Although Lean prod
uction wa initially ued only in manufacturing operation, it ha pread to er
vice and many healthcare organization (Caldwell, Brexler, and Gillem 2005; Cha
lice 2005; Spear 2005). A Lean organization i focued on eliminating all type
of wate. Like Six Sigma, Lean ha been defined a a philoophy, methodology, an
d et of tool. The Lean philoophy i to produce only what i needed, when
264
it i needed, and with no wate. The Lean methodology begin by examining the y
tem or proce to determine where in the proce value i added and where it i
not; tep in the proce that do not add value are eliminated, and thoe that
do add value are optimized. Lean tool include value tream mapping, the five S
, paghetti diagram, kaizen event, kanban, ingle-minute exchange of die, hei
junka, jidoka, and tandardized work. Where Six Sigma and total quality manageme
nt (TQM)/continuou quality improvement are focued on eliminating defect and r
educing variability, Lean i focued on eliminating wate and enuring wift, ev
en flow through the proce or ytem. However, defect are wate, and variabili
ty caue defect. Although the two program are different, their methodologie,
tool, and outcome are imilar. Both program have Japanee root, a evidence
d by the terminology aociated with them, and ue many of the ame tool and te
chnique.
Type of Wate
In Lean, wate i called muda, which come from the Japanee term for wate. Man
y type of wate are found in organization. Ohno defined even categorie of wa
te related to manufacturing, and they can be reinterpreted for ervice and hea
lthcare: 1. Overproduction: producing more than i demanded or producing before
the product i needed to meet demand. Printing report, label, and o forth, an
d preparing meal when they are not needed are example of overproduction in hea
lthcare. 2. Waiting: time during which value i not being added to the product o
r ervice. For ervice, thi include both cutomer waiting time and time where
the ervice provider i idle. 3. Tranportation: unneceary travel or conveyan
ce of the part being manufactured or the peron being provided the ervice. In h
ealthcare, wated tranportation include patient travel and equipment movement.
4. Inventory: holding or purchaing raw material, work-in-proce (WIP), and f
inihed good that are not immediately needed. Thi tie up capital. In healthca
re, wated inventory include upplie and pharmaceutical that are not immediat
ely needed. 5. Motion: action of provider or operator that do not add value t
o the product. Thi could alo include repetitive-type motion that caue injury
. In healthcare, wated motion include unneceary travel of the ervice provid
er to obtain upplie or information.
265
6. Overproceing: unneceary proceing or tep and procedure that do not ad
d value to the product or ervice. There are numerou example of overproceing
in healthcare related to recordkeeping and documentation. 7. Defect: productio
n of a part or ervice that i crapped or require rework. In healthcare, defec
t wate include mundane error, uch a mifiling document, and eriou error
reulting in the death of a patient. Lean focue on eliminating all wate from
the ytem through continuou improvement.
Kaizen
Kaizen i the Japanee term for change for the better, or continuou improvement
. Although kaizen i more typically aociated with the Lean philoophy, the qua
lity philoophie (TQM, Six Sigma, etc.) alo focu on the need for continuou i
mprovement. The philoophy of kaizen involve all employee in making uggetion
for improvement, then implementing thoe uggetion quickly. Kaizen i baed
on the aumption that everything can be improved and that many mall incrementa
l change will reult in a better ytem. Prior to adopting a kaizen philoophy
organization ued to believe, If it int broken, leave it alone; with a kaizen phi
loophy, organization believe, Even if it int broken, it can be improved. An orga
nization that doe not focu on continuou improvement will not be able to compe
te with other organization that are continuouly improving. Kaizen i viewed a
both a general philoophy of improvement focuing on the entire ytem or value
tream and a pecific improvement technique focued on a particular proce. Th
e kaizen philoophy of continuou improvement conit of five baic tep: 1. S
pecify value: Identify activitie that provide value from the cutomer perpecti
ve. 2. Map and improve the value tream: Determine the equence of activitie or
current tate of the proce and the deired future tate. Eliminate nonvalue-ad
ded tep and other wate. 3. Flow: Enable the proce to flow a moothly and q
uickly a poible. 4. Pull: Enable the cutomer to drive product or ervice.
5. Perfection: Repeat the proce to enure a focu on continuou improvement. T
he kaizen philoophy i enabled by the variou tool and technique of Lean.
266
Value Stream Mapping
Value tream mapping i a type of proce mapping or flowcharting of the value
tream, which include all of the tep, both the value-adding and the nonvalue-ad
ding tep, in producing and delivering a product or ervice. Both information p
roceing and tranformational proceing tep are included in a value team ma
p. The value tream map how proce flow from a ytem perpective and can he
lp in determining how to meaure and improve the ytem or proce of interet.
The value tream map can enable the organization to focu on the entire value t
ream rather than jut a pecific tep or piece of the tream. Without a view of
the entire tream, it i poible, even likely, that individual part of the y
tem will be optimized according to the need of thoe part, but the reulting
ytem will be uboptimal. Value tream mapping in healthcare i typically done f
rom the perpective of the patient, where the goal i to optimize her journey th
rough the ytem. Information, material, and patient flow are captured in the v
alue tream map. Each tep in the proce i claified a value added or nonvalu
e added. Value-added activitie are thoe that change the item being worked on i
n ome way that the cutomer deire. The key quetion to ak when determining i
f an activity i value added i, Would the cutomer be willing to pay for thi ac
tivity? If the anwer i no, the activity i nonvalue added. In healthcare, an exa
mple of a nonvalue-added activity i waiting timepatient would not pay to wait. N
onvalue-added activitie can be further claified a neceary or unneceary. A
n example of a neceary nonvalue-added activity that organization mut perform
i payroll. Payroll activitie do not add value for cutomer, but employee mu
t be paid. Activitie that are claified a nonvalue added and unneceary houl
d be eliminated. Activitie that are neceary but nonvalue added hould be exami
ned to determine if they can be made unneceary and eliminated. Value-added and
neceary but nonvalue-added activitie are candidate for improvement and wate
reduction. The value tream map enable organization to ee all of the activit
ie in a value tream and focu their improvement effort (Womak and Jone 1999)
. A common meaurement for the progre of Lean initiative i percent value add
ed. The total time for the proce to be completed i alo meaured. Thi can be
accomplihed by meauring the time a ingle item, cutomer, or patient pend t
o complete the entire proce. At each tep in the proce, the value-added time
i meaured: Value-added time 100 Total time in ytem
% Value added =
267
The goal of Lean i to increae percent value added by increaing thi ratio. Ma
ny procee have a percent value added of 5 percent or le. Once the value tr
eam map ha been generated, kaizen activitie can be identified that will allow
the organization to increae the percent value-added time and employ reource i
n the mot effective manner poible.
Vincent Valley Hopital and Health Sytem Value Stream Mapping
Vincent Valley Hopital and Health Sytem (VVH) identified it birthing center a
an area in need of improvement and decided to ue Lean tool and technique to
accomplih it objective. The goal for the Lean initiative were to decreae c
ot and increae patient atifaction. Project management tool (Chapter 5) wer
e ued to enure ucce. Initially, a team wa formed to improve the operation
of the birthing center. The team conited of the manager of the birthing unit
(the project manager), two phyician, three nure (one from triage, one from l
abor and delivery, and one from potpartum), and the manager of admiion. All
team member were trained in Lean tool and technique and tarted the project b
y developing a high-level value tream map over the coure of everal week (Fig
ure 9.1). The team mapped patient and information flow in the birthing center a
nd collected data related to taffing type and level a well a length of time f
or the variou proce tep. The high-level value tream map wa ued to help t
he team decide where to focu it effort. The team then developed a plan for th
e coming year baed on the opportunitie identified.
Meaure and Tool
Takt Time
Takt i a German word meaning rhythm or beat. It i often aociated with the rh
ythm et by a conductor to enure that the orchetra play in union. Takt time
determine the peed with which cutomer mut be erved to atify demand for t
he ervice. Takt time = Available work time/Day Cutomer demand/Day
Cycle time i the actual time to accomplih a tak in a ytem. Cycle time for a
ytem will be equal to the longet tak-cycle time in that ytem. In a Lean
ytem, cycle time and takt time are equal. If cycle time i greater than takt ti
me, demand will not be atified and cutomer or patient will wait. If cycle t
ime i le than takt time in a manufacturing environment, inventory i generate
d; in a ervice environment, reource are underutilized. In a Lean ytem, the
rate
268
FIGURE 9.1
VVH Birthing Center Value Stream Map
Room not available Slow turnaround Education late Long wait after cleared to di
charge
Nure time pent on nonpatient care
Supplie
LOS
Houekeeping Radiology Lab Pharmacy Dietetic
Anetheiology
Social ervice
Vaginal v. C-ection
Food quality
Kitchen
Stabilize
Room unavailable
Porter
Incorrect patient form
Patient
Admitting
Triage
Labor and Delivery
# % % hr Cycle time On-time delivery Wait time Changeover time 1 3 hr % min min
na 87 5 4 90 % hr % hr hr 1 8 hr 1 60 hr n % FTE 90 #
Potpartum
Dicharge
Arrival rate 6 # % % min Cycle time On-time delivery 92 25 45 Wait time Changeov
er time 0 2 hr 1 3 hr % min min 1 88 90 85 82 20 90 20 na FTE 20
30
#/ day
FTE
FTE
75 85 80 Cycle time On-time delivery Wait time Changeover time 50 75 5 4
# % % hr % hr hr 1 5 hr 20 80 hr

FTE
10 85 83 Cycle time On-time delivery Wait time Changeover time 2 75 2 na
# % % hr % hr hr Firt time correct Patient atifaction Firt time correct P
atient atifaction
Firt time correct Patient atifaction
Cycle time
On-time delivery
Wait time
Changeover time
30 90 min
3 hr
Created with eVSM oftware from GumhoeKI, Inc., a Microoft Viio add-on.
269
at which a product or ervice can be produced i et by cutomer demand, not by
the organization ability (or inability) to upply the product or ervice.
Throughput Time
Throughput time i the time for an item to complete the entire proce. It inclu
de waiting time and tranport time a well a actual proceing time. In a heal
thcare clinic, for example, throughput time would be the total time the patient
pend at the clinic, tarting when he walk through the door and ending when he
walk out; thi include not only the time the patient i actually interacting
with a clinician, but alo time pent waiting in the waiting room and examining
room. In a perfectly Lean ytem, there would be no waiting time and throughput
time would be minimized.
Riverview Clinic
VVH Riverview Clinic collected the data hown in Figure 9.2 for a typical patien
t viit. Here, the phyician exam and conultation ha the longet tak time, 20
minute; therefore, the cycle time for thi proce i 20 minute. Auming tha
t phyician time i not contrained, every phyician could output one patient fr
om thi proce every 20 minute. However, the throughput time i equal to the t
otal amount of time a patient pend in the ytem. Here, it i: 3 + 15 + 2 + 15
+ 5 + 10 + 20 = 70 minute The available work time per phyician day i 5 hour
(phyician work 10 hour per day, but only 50 percent of that time i pent wi
th patient), there are 8 phyician, and 100 patient are expected at the clini
c every day. FIGURE 9.2 Riverview Clinic Cycle, Throughput, and Takt Time
Patient check-in 3 minute
Wait 15 minute
Move to examining room 2 minute
Wait 15 minute
Nure doe preliminary exam 5 minute
Wait 10 minute
Phyician exam and conultation 20 minute
Viit complete
NOTE: Created with Microoft Viio.
270
Takt Time =
8 Phyician 5 Hour/Day = 0.4 Phyician hour/Patient 100 Patient/Day = 24 Phy
ician minute/Patient
Therefore, to meet demand, the clinic need to erve one patient every 24 minute
. Becaue cycle time i le than takt time, the clinic can meet demand. Aumi
ng that patient check-in i neceary but nonvalue added and that both the nure
preliminary exam and the phyician exam and conultation are value-added tak,
the value-added time for thi proce i: 5 Minute (Nure preliminary exam) + 2
0 Minute (Phyician exam and conultation) = 25 Minute and the percent value-a
dded time i: 25 Minute/70 Minute = 36% A Lean ytem work toward decreaing
throughput time and increaing percent value-added time. The tool dicued in
the following ection can aid in achieving thee goal.
Five S
The five S are workplace practice that contitute the foundation of other Lean
activitie; the Japanee word for thee practice all begin with S. The five S
are eentially way to enure a clean and organized workplace. Often, they ar
e een a obviou and elf-evident; a clean and organized workplace will be more
efficient. However, without a continuing focu on thee five practice, workpla
ce often become diorganized and inefficient. The five practice, with their Ja
panee name and the Englih term typically ued to decribe them, are: 1. Seir
i (ort): Separate neceary from unneceary item, including tool, part, mat
erial, and paperwork, and remove the unneceary item. 2. Seiton (traighten):
Arrange the neceary item neatly, providing viual cue to where item hould
be placed. 3. Seio (weep): Clean the work area. 4. Seiketu (tandardize): St
andardize the firt three S o that cleanline i maintained. 5. Shituke (u
tain): Enure that the firt four S continue to be performed on a regular bai
.
271
The five S are often the firt tep an organization take in it Lean journey b
ecaue o much wate can be eliminated by enuring an organized and efficient wo
rkplace.
Spaghetti Diagram
A paghetti diagram i a viual repreentation of the movement or travel of mate
rial, employee, or cutomer. In healthcare, a paghetti diagram i often ued
to document or invetigate the movement of caregiver or patient. Typically,
the patient or caregiver pend a ignificant amount of time moving from place t
o place and often backtrack. A paghetti diagram (Figure 9.3) can help to find
and eliminate wated movement in the ytem.
Kaizen Event or Blitz
A kaizen event or blitz i a focued, hort-term project aimed at improving a pa
rticular proce. A kaizen event i uually performed by a cro-functional team
of eight to ten people, alway including at leat one peron who actually work
with or in the proce. The ret of the team hould include peronnel from othe
r functional area and even nonemployee with an interet in improving the proce
. In healthcare organization, taff, nure, doctor, and other profeional
, a well a management peronnel from acro department, hould be repreented
. Typically, a kaizen event conit of the following tep, baed on the plan-d
o-check-act (PDCA) improvement cycle of Deming and Juran: Determine and define t
he objective(). Determine the current tate of the proce by mapping and meau
ring the proce. Meaurement will be related to the deired objective but oft
en include uch thing a cycle time, waiting time, WIP, throughput time, and tr
avel ditance. Determine the requirement of the proce (takt time), develop ta
rget goal, and deign the future tate or ideal tate of the proce. Create a
plan for implementation, including who, what, when, and o on. Implement the imp
rovement. Check the effectivene of the improvement. Document and tandardize
the improved proce. Continue the cycle. The kaizen event i baed on the noti
on that mot procee can be quickly (and relatively inexpenively) improved, i
n which cae it make ene to jut do it rather than be paralyzed by reitance t
o change. A kaizen event i typically one week long and begin with training in
the tool of Lean, followed by analyi and meaurement of the current proce a
nd generation of poible idea for improvement. By midweek, a propoal for chan
ge to improve the proce hould be ready. The propoal hould include the impr
oved proce flow
272
FIGURE 9.3 Spaghetti Diagram: VVH Potpartum Nuring
20
401
402
403
3 5
404
6 8
405
406
407
408
409
410
411
412
10 13 14 18
19 9 1
21 11
Office
7
12
16
Office Locker Room
4
17
Nuring Station
Supplie and Equipment
15
Kitchen
Med Room
2
424
423
422
421
420
419
418
417
416
415
414
413
Note:
21 poition change Multiple trip to upply room Multiple trip to med room Sea
rch time (upplie and paperwork)
Time (min)
5 0 Food upplie Find info Get med NSLT
and metric for determining the effect of the change. The propoed change are
implemented and teted during the remainder of the week. By the following week,
the new proce hould be in place. A kaizen event can be a powerful way to qui
ckly and inexpenively improve procee. The reult are uually a ignificantl
y improved proce and increaed employee pride and atifaction. Table 9.1 how
the improvement experienced by Virginia Maon Medical Center in Seattle, two
year and 175 kaizen event (RPIW) after beginning their lean journey.
7. 5
30 25 20 15 10
12 .5
10
30
273
VVH Kaizen Event
The value tream map developed for the VVH birthing center highlighted the fact
that nuring taff wa pending a ignificant amount of time on activitie not r
elated to actual patient care. Thi not only reulted in diatified patient,
phyician, and nure, but it alo increaed taffing cot to the hopital. A
kaizen blitz wa planned to addre thi problem in the potpartum area of the b
irthing center. The nuring adminitrator wa charged with leading the kaizen ev
ent. She put together a team coniting of a phyician, a houekeeper, two nure
aitant, and two nure. On Monday morning, the team began the kaizen event
with four hour of Lean training. On Monday afternoon, they developed a paghett
i diagram for a typical nure (Figure 9.3) and began collecting data related to
the amount of time nuring taff pent on variou activitie. They alo collecte
d hitorical data on patient load and taffing level. On Tueday morning, the t
eam continued to collect data. On Tueday afternoon, they analyzed the data and
found that nuring taff pent only 50 percent of their time in actual patient c
are. A ignificant amount of time one hour per eight-hour hiftwa pent locating
equipment, upplie, and information. The team decided that a 50 percent reducti
on in thi time wa a reaonable goal for the kaizen event. On Wedneday morning
, the team ued root-caue analyi to determine the reaon nuring taff pent
o much time locating and moving equipment and upplie. They determined that o
ne of the major caue wa general diorder in the upply/equipment room and pat
ient room. On Wedneday afternoon, the team organized the upply/equipment room
. They began by determining what upplie and equipment were neceary and remov
ing thoe that were unneceary. Next, they organized the upply/equipment room
by determining which item were often needed together and locating thoe item c
loe to one another. All torage area were labeled, and pecific location for
equipment were deignated viually. White board were intalled to enable the tr
acking and location of equipment. The team alo developed and poted a map of th
e room o that the location of equipment and upplie could be eaily determined
. On Thurday, the team worked on reorganizing all of the patient room, tandar
dizing the layout and location of item in each one. Firt, the team oberved on
e of the patient room and determined the equipment and upply need of phyicia
n and nure. All noneential item were removed, creating more pace. Additio
nally, room were tocked with upplie ued on a routine bai to reduce trip
to the central upply room. A procedure wa alo etablihed to retock upplie
on a daily bai. On Friday morning, the kaizen team again collected data on th
e amount of time nuring taff pent on variou activitie. They found that afte
r implementing the change, the time nuring taff pent locating and moving
274
upplie and equipment wa reduced to approximately 20 minute in an eight-hour

hift, a 66 percent reduction. Friday afternoon wa pent documenting the kaizen
event and putting ytem in place to enure that the new procedure and organi
zation would be maintained.
Standardized Work
Standardized work i an eential part of Lean and provide the baeline for con
tinuou improvement. Standardized work i written documentation of the precie w
ay in which every tep in a proce hould be performed. It hould not be een a
a rigid ytem of compliance, but rather a a mean of communicating and codif
ying current bet practice within the organization to provide a baeline from w
hich improvement can be made to the proce. All relevant takeholder of the p
roce hould be involved in etablihing tandard work. Standardizing work in t
hi way aume that the people mot intimately involved with the proce have t
he mot knowledge of how to bet perform the work. It promote employee buy-in a
nd ownerhip of the proce and reponibility for improvement. Clear documentat
ion and pecific work intruction enure that variation and wate are minimized
. Standardized work hould be een a a tep on the road to improvement. A bett
er method are identified, the work tandard are modified to reflect the change
. In the healthcare indutry, example of tandardized work include treatment p
rotocol and the etablihment of care path. (Care path are alo example of e
vidence-baed medicine, which i explored in more depth in Chapter 3.) Maachu
ett General Hopital developed and implemented a care path for coronary artery
bypa graft (CABG) urgery that reulted in an average length of tay reduction
of 1.5 day; ignificant cot aving were aociated with that reduction (Whee
lwright and Weber 2004). A part of an overall program to improve practice and
reduce cot, Maachuett General identified the etablihment of a care path
for CABG urgery. A care path i an optimal equencing and timing of intervention
by phyician, nure, and other taff for a particular diagnoi or procedure
, deigned to minimize delay and reource utilization and at the ame time maxi
mize the quality of care (Wheelwright and Weber 2004). A care path define and do
cument pecifically what hould happen to a patient the day before urgery, the
day after urgery, and on poturgical day. The care path wa not intended to
dictate medical treatment, but to tandardize procedure a much a poible to
reduce variability and improve the quality of outcome. There were 25 participan
t repreenting the variou area involved in treatment on the team that develop
ed the care path. The team took more than a year to develop the initial care pat
h, but becaue of it breadth of
275
incluion and applicability, reitance to implementation wa minimal. After the
ucceful implementation of the CABG urgical care path, Maachuett General
Hopital developed and implemented more than 50 care path related to urgical
procedure and medical treatment (Wheelwright and Weber 2004). Standard work pr
ocee can be ued in clinical, upport, and adminitrative operation of healt
hcare organization. The development and documentation of tandardized procee
and procedure can be a powerful tool for engaging and involving everyone in th
e organization in continuou improvement.
Jidoka and Andon
In Lean ytem, jidoka refer to the ability to top the proce in the event o
f a problem. The term tem from the weaving loom invented by Sakichi Toyoda, Fo
under of the Toyota Group. The loom topped itelf if a thread broke, eliminatin
g the poibility that defective cloth would be produced. Jidoka prevent defect
from being paed from one tep in the ytem to the next and enable the wif
t detection and correction of error. If the ytem or proce i topped when a
problem i found, everyone in the proce work quickly to identify and elimina
te the ource of the error. In ancient Japan, an andon wa a paper lantern ued
a a ignal; in a Lean ytem, an andon i a viual or audible ignaling device
ued to indicate a problem in the proce. Andon are typically ued in conjunct
ion with jidoka. Virginia Maon Medical Center implemented a jidoka-andon ytem
called the Patient Safety Alert Sytem (Womack et al. 2005). If a caregiver bel
ieve that omething i not right in the care proce, he not only can, but mu
t, top the proce. The peron who ha noticed the problem call (or alert, vi
a the Internet) the patient afety department. The appropriate proce takehold
er or relevant manager immediately come to determine and correct the root cau
e of the problem. After two year, the number of alert per month roe from 3 to
17, enabling Virginia Maon to correct mot problem in the proce before they
become more eriou. The alert are primarily related to ytem iue, medica
tion error, and problem with equipment or facilitie.
Kanban
Kanban i a Japanee term meaning ignal. A kanban ytem in a manufacturing org
anization typically ue container of a certain ize to ignal the need for mor
e production. When a container i emptied at a worktation, it i ent uptream
to ignal to the previou operation that work can begin. Kanban are the control
ling mechanim in a pull production ytem. The cutomer indicate that he want
product, a kanban i releaed to the lat operation in the ytem to ignal the
cutomer demand, and that tation begin to produce product in
276
repone. A incoming material i conumed at the lat worktation, another kanb
an will be emptied and ent to the previou worktation to ignal that productio
n hould begin at that tation. The empty kanban go backward through the produc
tion ytem to ignal the need to produce in repone to cutomer demand (ee Fi
gure 9.4). Thi enure that production i only done in repone to cutomer dem
and, not imply becaue production capacity exit. In a healthcare environment,
kanban can be ued for upplie or pharmaceutical to ignal the need to order
more. For example, a pharmacy would have two kanban; when the firt kanban i
emptied, thi ignal the need to order more and an order i placed. The econd
kanban i emptied while waiting for the order to arrive. Ideally, the firt kanb
an i received from the upplier at the point that the econd kanban i empty an
d the cycle continue. The ize of the kanban i related to demand for the phar
maceutical during lead time for the order. The number and ize of the kanban de
termine the amount of inventory in the ytem. In a healthcare environment, kanb
an could alo be ued to control the flow of patient, enuring continuou flow
. For example, for patient needing both an echocardiography (echo) procedure an
d a computed tomography (CT) can, where the echo procedure i to be performed b
efore the CT can, the CT can could pull patient through the proce. When a C
T i performed, a patient i taken from the pool of patient between CT and echo
. A kanban (ignal) i ent to the echo tation to indicate that another patient
hould receive an echo (ee Figure 9.5). Thi enure that a contant pool of p
atient i kept between the two procee. The patient pool hould be large enou
gh to enure that the CT i buy even when there are diturbance in the echo pr
oce. However, thi mut be balanced with the need to keep patient from waitin
g for long period. Eventually, in a Lean ytem, the pool ize would be reduced
to one.
FIGURE 9.4 Kanban Sytem
Empty Kanban
Empty Kanban
Tak 1 Worktation 1
Full Kanban
Tak 2 Worktation 2
Full Kanban Cutomer Order
277
Signal
Signal
FIGURE 9.5 Kanban for Echo/CT Scan
Patient
Echo
Patient
CT
Single-Minute Exchange of Die
The ingle-minute exchange of die (SMED) ytem wa developed by Shigeo Shingo (
1985) of Toyota. Originally, it wa ued by manufacturing organization to reduc
e changeover or etup time, the time between producing the lat good part of one
product and the firt good part of a different product. Currently, the techniqu
e i ued to reduce etup time for both manufacturing and ervice. In healthcar
e environment, etup i the time needed, or taken, between the completion of on
e procedure and the tart of the next. The SMED technique conit of three tep
: 1. Separating internal activitie from external activitie; 2. Converting int
ernal etup activitie to external activitie; and 3. Streamlining all etup act
ivitie. Internal activitie are thoe activitie that mut be done in the yte
m; they could not be done offline. For example cleaning an operating room (OR) p
rior to the next urgery i an internal etup activity; it could not be complete
d outide the OR. However, organizing the urgical intrument for the next urg
ery i an external etup, a thi could be done outide the OR to allow for pee
dier changeover of the OR. Setup include finding and organizing intrument, ge
tting upplie, cleaning room, and obtaining paperwork. In the healthcare envir
onment, SMED ha been ued to treamline the proce for changing over from one
patient to the next in diagnotic imaging area, urgical uite, and hopital r
oom.
Flow and Pull
Continuou, or ingle-piece, flow conit of moving item to be worked on (job
, patient, product) through the tep of the proce one at a time without int
erruption or waiting. Pull, or jut-in-time, production refer to a
278
ytem in which product or ervice are not produced until the downtream cuto
mer demand them. A Lean organization continually work toward achieving both of
thee goal. Enuring efficient patient flow ha become increaingly important
in healthcare. In 2005, the Joint Commiion introduced a new tandard on managi
ng patient flow: The leader develop and implement plan to identify and mitigate
impediment to efficient patient flow throughout the hopital. To meet thi tan
dard, hopital need to have policie, procee, and procedure in place to en
ure continuou improvement in patient flow. The Intitute for Healthcare Improve
ment (IHI 2006a) ha extenive reource available on the Internet to help healt
hcare organization improve patient flow; thee reource include meaure, tool
, and cae tudie. IHI model for improvement i baed on the PDCA cycle and in
corporate many Lean tool and technique.
Heijunka and Advanced Acce
Heijunka i a Japanee term meaning make flat and level. It refer to eliminating
variation in volume and variety of production to reduce wate. In healthcare en
vironment, thi often mean determining how to level patient demand. Producing
good or ervice at a level rate allow organization to be more reponive to
cutomer and better ue their own reource. In healthcare, advanced acce pro
vide a good example of the benefit of heijunka. Several tudie have hown tha
t many people are unable to obtain timely primary care appointment (Murray and
Berwick 2003). Advanced acce cheduling reduce the time between cheduling an
appointment for care and the actual appointment. It i baed on the principle
of Lean and aim for wift, even patient flow through the ytem. In clinical e
tting, reducing the wait time for appointment ha been hown to decreae no-h
ow rate (Kennedy and Hu 2003) and improve both patient and taff atifaction
(Radel et al. 2001). Revenue are increaed (OHare and Corlett 2004) a a reult
of higher patient volume and increaed taff and clinician productivity. Additi
onally, greater continuity of care (Belardi, Weir, and Craig 2004) hould increa
e the quality of care and reult in more poitive outcome for patient. Althou
gh the benefit of advanced acce are great, implementation can be difficult be
caue the concept challenge etablihed practice and belief. However, if the
delay between making an appointment and the actual appointment i relatively con
tant, it hould be poible to implement advanced acce. Centra, a multiite p
rimary care clinic located in Chicago, began advanced acce cheduling in 2001.
The organization wa able to reduce acce time to 3 day or le, patient ati
faction increaed from 72 percent to 85 percent, and continuity of care wa ig
nificantly increaed (75 percent of viit occurred with a patient primary phyi
cian, compared to 40 percent
279
prior to advanced acce). The bigget iue encountered wa the greater demand
for popular clinician and the need to addre thi inequity on an ongoing bai
(Murray et al. 2003). Succeful implementation of advanced acce require tha
t upply and demand be balanced. To do thi, accurate etimate of both upply a
nd demand are needed, backlog mut be reduced or eliminated, and the variety of
appointment type need to be minimized. Once upply and demand are known, deman
d profile may need to be adjuted and the availability of bottleneck reource
increaed (Murray and Berwick 2003). IHI (2006b) offer extenive online reourc
e to aid healthcare organization in implementing advanced acce, and Chapter
12 dicue the concept in more detail.
Concluion
Recently, organization have begun combining the philoophie and tool of both
Lean and Six Sigma into Lean Sigma. Although proponent of Lean or Six Sigma might
ay that thee initiative are very different and champion one or the other, th
ey are actually complementary method, and combining them can be effective for m
any organization. Lean i focued on eliminating wate and treamlining flow. S
ix Sigma i focued on eliminating defect and reducing variation. However, both
Lean and Six Sigma are ultimately focued on continuou improvement of the yt
em. The kaizen philoophy of Lean begin with determining what cutomer value,
followed by mapping and improving the proce to achieve flow and pull. Six Sigm
a i focued on reducing defect and eliminating variation in the procee. Lea
n thinking enable identification of the area cauing inefficiencie. However,
to truly achieve Lean, variation in the procee mut be eliminatedSix Sigma wil
l help to do thi. Focuing on the cutomer and eliminating wate will not only
reult in increaed cutomer atifaction, but it will alo reduce cot and inc
reae the profitability of the organization. Together, Lean and Six Sigma can pr
ovide the philoophie and tool needed to enure that the organization i conti
nuouly improving.
Dicuion Quetion
1. What are the driver behind the healthcare indutry new focu on employing reo
urce in an effective manner and on patient atifaction? 2. What are the differ
ence between Lean and Six Sigma? The imilaritie? Would you like to ee both i
n your organization? Why or why not? 3. From your own experience, dicu a pe
cific example of each of the even type of wate.
280
4. From your own experience, decribe a pecific intance in which tandardized
work, kanban, jidoka and andon, and SMED would enable an organization to become
more effective or efficient. 5. Doe your primary care clinic have advanced acc
e? Should it? To determine upply and demand and track progre, what meaure
would you recommend to your clinic? 6. Are there any drawback to Lean Sigma? E
xplain.
Chapter Exercie
1. A imple value tream map for patient requiring a colonocopy at an endocop
y clinic i hown in Figure 9.6. Aume that patient recover in the ame room w
here the colonocopy i performed and there are two colonocopy room. What i t
he cycle time for the proce? What i the throughput time? What i the percent
value added in thi proce? If the clinic operate 10 hour a day and demand i
12 patient per day, what i the takt time? If demand i 20 patient per day, w
hat i the takt time? What would you do in the econd ituation? FIGURE 9.6 Valu
e Stream Map for Colonocopie
Colonocopy patient
Demand Value No./ day
Patient check-in
Patient prep
Colonocopy
Patient recovery
Dicharge
20 Min 5 Min 15 Min
15 Min 30 Min
0 Min 40 Min
10 Min 5 Min
NOTE: Created with eVSM oftware from GumhoeKI, Inc., a Microoft Viio add-on.
2. Draw a high-level value tream map for your organization (or a part of your o
rganization). Pick a part of thi map and draw a more detailed value tream map
for it. On each map, be ure to identify the information you would need to compl
ete the map and exactly how you might obtain it. What are the takt and throughpu
t time of your proce? Identify at leat three kaizen opportunitie on your ma
p. (Advanced: Ue eVSM.) 3. For one of your kaizen opportunitie decribed above
, decribe the kaizen event you would plan if you were the kaizen leader.
Reference
Belardi, F. G., S. Weir, and F. W. Craig. 2004. A Controlled Trial of an Advanced
Acce Appointment Sytem in a Reidency Family Medicine Center. Family Medicine
36: 34145.
281
Caldwell, C., J. Brexler, and T. Gillem. 2005. Lean-Six Sigma for Healthcare: A
Senior Leader Guide to Improving Cot and Throughput. Milwaukee, WI: ASQ Quality
Pre. Chalice, R. 2005. Stop Riing Healthcare Cot Uing Toyota Lean Product
ion Method: 38 Step for Improvement. Milwaukee, WI: ASQ Quality Pre. Intitu
te for Healthcare Improvement. 2006a. Flow. [Online information; retrieved 7/17/06
.] www.ihi.org/IHI/Topic/Flow/. . 2006b. Primary Care Acce. [Online information; r
etrieved 7/17/06.] www.ihi.org/IHI/Topic/OfficePractice/Acce/. Joint Commi
ion on Accreditation of Healthcare Organization. 2005. New Standard LD.3.11. [Onl
ine information; retrieved 7/17/06.] www.jcrinc.com/ ubcriber/perpective.a
p?durki=6640&ite=10&return=6065. Kennedy, J. G., and J. T. Hu. 2003. Implementa
tion of an Open Acce Scheduling Sytem in a Reidency Training Program. Family
Medicine 35: 66670. Murray, M., and D. M. Berwick. 2003. Advanced Acce: Reducing
Waiting and Delay in Primary Care. JAMA 290: 33234. Murray, M., T. Bodenheimer,
D. Rittenhoue, and K. Grumbach. 2003. Improving Timely Acce to Primary Care: C
ae Studie in the Advanced Acce Model. JAMA 289: 104246. OHare, C. D., and J. Co
rlett. 2004. The Outcome of Open-Acce Scheduling. Family Practice Management 11
(2): 3538. Panchak, P. 2003. Lean Health Care? It Work. Indutry Week 252 (11): 3
438. Radel, S. J., A. M. Norman, J. C. Notaro, and D. R. Horrigan. 2001. Redeigni
ng Clinical Office Practice to Improve Performance Level in an Individual Prac
tice Aociation Model HMO. Journal of Healthcare Quality 23 (2): 1115. Shingo, S.
1985. A Revolution in Manufacturing: The SMED Sytem. Tranlated by A. Dillon.
New York: Productivity Pre. Spear, S. J. 2005. Fixing Health Care from the Ini
de, Today. Harvard Buine Review 83 (9): 7891. Wheelwright, S., and J. Weber. 20
04. Maachuett General Hopital: CABG Surgery (A). Harvard Buine Review Ca
e 9-696-015. Boton: Harvard Buine School Publihing. Womack, J. P., A. P. B
yrne, O. J. Fiume, G. S. Kaplan, and J. Touaint. 2005. Going Lean in Health Ca
re. Cambridge, MA: Intitute for Healthcare Improvement. [Online information; re
trieved 7/17/06.] www.ihi.org/IHI/Reult/ WhitePaper/GoingLeaninHealthCare.htm
. Womack, J., and D. Jone. 1999. Learning to See. Brookline, MA: The Lean Enter
prie Intitute. Womack, J. P., D. T. Jone, and D. Roo. 1990. The Machine that
Changed the World: Baed on the Maachuett Intitute of Technology 5-Million
Dollar 5-Year Study on the Future of the Automobile. New York: Rawon Aociate
.
CHAPTER
10
SIMULATION
CHAPTER OUTLINE
Operation Management in Action Overview Ue of Simulation The Simulation Proce
Model Development Model Validation Simulation and Output Analyi Monte Carlo
Simulation Vincent Valley Hopital and Health Sytem Example Dicrete Event Sim
ulation Queueing (Waiting Line) Theory Queueing notation Queueing olution VVH
M/M/1 Queue Dicrete Event Simulation Detail VVH M/M/1 Queue Advanced DES Concl
uion Dicuion Quetion Chapter Exercie Reference
282

arrival proce bottleneck capacity utilization () cadiopulmonay esuscitation
(CPR) community acquied pneumonia (CAP) Decisioneeing Cystal Ball diagnosis-
elated goup (DRG) discete event simulation (DES) entity event exponential dist
ibution fist come, fist seved flow time, see thoughput time Institute of Me
dicine (IOM) inteaival time Kaizen event Littles law M/M/1 queueing system mag
netic esonance imaging (MRI) mean aival ate () mean service rate () Microsoft
Visio mode vaidation Monte Caro simuation Nationa Academy of Engineering (N
AE) Paisade @Risk Paisade BestFit Poisson distribution queue queue discipin
e queueing notation queueing theory resource or server shortest processing time
Simu8 state task cyce time throughput time tornado graph waiting ine theory,
see queueing theory
283

Surgeons at BC Chidrens Hospita have a way to shorten wait ists and get their
patients through the hospita system more efficienty and with ess stress. Drs.
Geoff Bair and Jacques Lebanc, head and assistant head of surgery, respective
y, have created a simuation mode not unike those used by airines and the mi
itary during wartime, that aows for the movement of arge numbers of peope.
Using a branch of mathematica anaysis known as the science of decision making, i
ncuding game theory and queueing theory, Drs. Bair and Lebanc beieved it was
possibe to simuate and test how changes to patient fow and surgica schedui
ng woud affect throughput, patient waiting times, and budgets, without adverse
y affecting ongoing operations. They created a mode to simuate the processes f
rom the moment a patient enters BC Chidrens for surgery to the time they eave.
This incudes OR prep time, how ong each procedure takes per surgeon, how patie
nts are moved from the ICU to the appropriate unit, avaiabe bed space and actu
a recovery times. Programmers added random factors provided by hospita staff s
uch as unschedued emergency cases, unpanned fuctuations in staff eves due t
o iness, equipment breakdowns, and cerica probems. They aso came up with a
bock schedue anaysis too the hospita coud use to test various schedue ch
anges, and assess their impact on future wait times, staffing, and costs. The te
am used aiases of more than 30,000 actua patient fies from the previous three
years to ensure confidentiay was maintained during the simuation process. Th
ey checked the accuracy of the program by running a one year simuation of patie
nt fow, and the resuts compared to what actuay transpired during that period
. The vaue of the simuation is that it can effectivey get past the what if? whi
e staff try to come up with ways to dea with growing wait ists and escaating
costs, says Dr. Bair. In heathcare, weve been reying on our puny human minds t
o try to dea with a these systems of queues and decision making. And the inte
raction of everything has to be brought to bear to get a kid at Chidrens Hospita
who needs surgery into the operating room, with everybody who needs to be ther
e, with a the equipment that needs to be there, on time, at an appropriate tim
e that refects the chids medica condition. Its not unike the same compexity a
irines face when faced with ticketing, boarding passes, security, baggage hand
ing, runway avaiabiity, 747s unoading their passengers whie another 747 is b
oarding passengers. So why not use the same sort of approach? he asked.
284
Chapter 10: Simuation

285
The mode has been a boon for bock scheduing, he says. OR times have been set
since the hospita opened in 1982, but with the aid of the simuation too, thos
e surgeries with the ongest wait ists can be reaocated without causing a og
jam. The human tendency is to avoid changing, because we dont know whats going to h
appen, said Bair. Theres a rea inertia in heathcare. And governments are increas
ingy demanding more proof that we deserve more money, to prove that we are bein
g as efficient as possibe. What we can do with this bock scheduer [is] simua
te what wi happen because we have a the data entered in rea timeevery patien
t who is waiting is now in the system and we can then take time away from orthope
dics, cardiac surgery and give it to ophthamoogy, say, and then we just run it
. Within minutes, we can see graphicay exacty what wi happen to the waiting
ists in cardiac surgery, orthopedics, and ophthamoogy. It aows us to tweak
time in areas that are not in bad shape and not inconvenience anybody; we can p
redict scientificay, in a vaid way, what wi happen if we make this change.
This removes the fear of change. The Simuation Too, as its caed, wi aso be
used to better manage outpatient cinics as we as OR time. Its aso going to be
appied to better management of beds. In one striking scenario used by the prog
rammers, Wednesday morning rounds with surgeons and nurses were changed to the a
fternoon and resuted in an additiona 54 surgeries being performed over the cou
rse of a year. This isnt a cure for cancer, said Bair. But it coud go a ong way t
o curing some of the major probems that affect us every day in our attempts to de
iver care to our kids here.
SOURCE: Haey, L. 2005. Simuation Aims to Speed Patients Through Surgeries. Medic
a Post 41 (1): 23. Reprinted with permission.
Overview
Simuation is the process of modeing reaity to gain a better understanding of
the phenomena or system being studied. Simuation aows the user to ask and ans
wer what if questions in an environment that is more cost effective, ess dangerou
s, faster, or more practica than the rea word. Athough the simuation techni
ques outined in this chapter are computer based mathematic modes, simuation d
oes not require mathematic modes or computer anaysis. A fire dri, for examp
e, is a simuation of what woud or coud happen in the event of a rea fire. Th
e dri is run to address
286
any probems that might arise if there were a rea fire, without the danger asso
ciated with a rea fire. There are many different types of simuations appropria
te for many different contexts. The report Buiding a Better Deivery System: A
New Engineering/ Heath Care Partnership, a joint effort between the Nationa Ac
ademy of Engineering and the Institute of Medicine (2005), identifies engineerin
g toos and technoogies that coud be empoyed to hep to overcome the current
safety, quaity, cost, and access crises faced by the heathcare industry. This
report specificay cites systems modeing and simuation as toos that have the
power to enabe heathcare organizations to improve quaity, efficiency, and sa
fety. This chapter provides an introduction to simuation and the theories under
ying it. The major topics covered incude: How, where, and why simuation can b
e used; The simuation process; Monte Caro simuation; Queueing theory; and Dis
crete event simuation.
After competing this chapter, readers shoud have a basic understanding of simu
ation. This shoud hep them understand how simuation coud be used in their o
rganizations to evauate choices and optimize processes and systems.
Uses of Simuation
Simuation can be used for many different purposes, incuding performance, proof
, discovery, entertainment, training, education, and prediction (Axerod 2006).
Performance simuation can actuay carry out some task for an organization or e
ntity. It is reated to artificia inteigence and usuay simuates a human be
havior. Voice recognition and robotic assisted surgery are exampes of performan
ce simuation. Simuation can be used to discover new reationships and princip
es and to provide proof of a theory. For exampe, Conways Game of Life (Berekamp
, Conway, and Guy 2003; Poundstone 1985) is used to discover and prove that simp
e rues can resut in compex behavior. Simuations can aso be used to enterta
in, as with virtua reaity video games. Simuation is often used for education
and training purposes. Increasingy, simuators are being used to educate heath
care professionas in medica concepts and decision making as we as to train t
hem in therapeutic and diagnostic procedures. For exampe, the mannequin simuat
or Resusci Anne has been used for CPR training since the 1960s (Grenvik and Scha
efer

287
2004). Since then, simuations reated to medica training and education have be
come increasingy sophisticated. Training simuations aow users to practice de
cisions and techniques in a safe environment where an incorrect decision does no
t have serious consequences. Predictive simuation can be used to evauate the d
esign of new products, systems, or procedures as we as to anayze and improve
existing products, systems, or procedures. This chapter focuses on predictive si
muationspecificay, Monte Caro simuation and discrete event simuation.
The Simuation Process
Simuation begins with deveopment of a mode. Once the mode has been buit and
vaidated, the output of the simuation is anayzed to address the origina que
stion or probem (Figure 10.1).
Mode Deveopment
The first step in mode deveopment is to define the probem or question to be a
nswered with the simuation. The usefuness of the simuation wi be driven by
the accuracy of the probem definition. The next step in deveoping a simuation
mode is defining the conceptua mode. Here, the system is described anaytica
y or mathematicay; inputs and outputs are determined, and reationships are
defined. Because rea word systems are compex and difficut to represent anay
ticay, assumptions about the system must be made. A perfect mode of the syste
m is sedom possibe, and approximations appropriate for the study must be made.
There is usuay a trade off between mode vaidity and mode compexity; a t
hings being equa, however, a simper mode is better. Once the conceptua mode
has been defined, information required for the simuation must be coected. Da
ta reated to the probabiity distributions of random variabes in the system, d
ata defining the reationships in the
Mode Deveopment Probem/ question definition Conceptua mode definition Data
coection
Mode Vaidation Quantitative comparison Expert opinion
Simuation and Anaysis Design of experiments Repication Data coection, stora
ge, and organization Anaysis
FIGURE 10.1 Simuation Process
288
simuation, and data reated to the output behavior of existing systems are co
ected. These data wi be used in running and vaidating the simuation. The fin
a step in mode deveopment is actuay buiding the computer mode. In the pas
t, this meant coding the software for the mode. Today, many commerciay avaia
be software packages make this step reativey simpe.
Mode Vaidation
The vaidity of a simuation is reated to how cosey the simuation mirrors re
aity and answers the question that was asked. Simuations can be deveoped that
are technicay correct but do not accuratey refect reaity or do not address
the intended question or probem. Therefore, assessing the vaidity of the simu
ation is an essentia, but often difficut, step. Ideay, the simuation is ru
n and quantitative output data of the simuation are compared to output data fro
m the rea system to determine whether they are simiar. Aternativey, experts
are asked to determine if the design and output of the simuation make sense to
them. If the simuation is not deemed vaid, the mode must be redeveoped.
Simuation and Output Anaysis
Here, the simuation mode is actuay run and output data are coected. If a n
umber of different variabes and variabe states are of interest, experimenta d
esign can be used to determine the specifications of those variabes so that the
experiments can be optimay run in a timey, cost effective manner. Ensuring r
eiabe resuts may require many repications of the simuation. The resuts of
the simuations must be coected, organized, and stored. Finay, the output da
ta must be anayzed to determine the answer to the origina question or probem.
Monte Caro Simuation
Monte Caro simuation was pioneered by John von Nuemann, Stanisaw Uam, and Ni
choas Metropois in the 1940s whie they were working on the Manhattan Project
to deveop the first atomic bomb (Metropois 1987). This group conceived of the
idea of modeing the output of a system by using input variabes that coud not
be known exacty but coud be represented with probabiity distributions. Many r
epetitions of the mode were run, and the behavior of the rea system coud be e
stimated based on the outcomes of the many repications. This technique came to
be known as Monte Caro simuation because of its use of probabiity distributio
ns (discussed in Chapter 7) and their reationship to the games of chance found
in the famous casino.

289
The Monte Caro method consists of defining the reationships in a system mathem
aticay. The random variabes in the mode (those that are uncertain and have a
range of possibe vaues) are not characterized by a singe number but by a pro
babiity distribution. The probabiity distribution used to characterize the ran
dom variabe is chosen based on historica data or expert knowedge of the situa
tion. Many soutions to the mode are determined, each one found by samping fro
m the probabiity distributions associated with the random variabes. The souti
on is not a singe number but a distribution of possibe outcomes that can be ch
aracterized statisticay. A simpe, noncomputer based exampe of Monte Caro sim
uation is provided here to aid in understanding. A cinic wishes to estimate th
e amount of revenue that wi be coected from each patient. Fifty percent of t
he cinics patients do not pay for their services, and it is equay ikey that
they wi pay or not pay. The cinic has coected information on charges incurr
ed by the most recent 360 patients (Figure 10.2). The payment per patient is mod
eed by: Probabiity of payment Charges/Patient = Payment/Patient A deterministi
c soution to this probem woud be: 0.50 $70/Patient = $35 per Patient A Monte
Caro simuation of this probem is based on cacuating the payment from many i
ndividua simuated patients. The probabiity of payment can be simuated by fi
pping a coin to represent the probabiity that a FIGURE 10.2 Payment Information
for Monte Caro Simuation
Charges ($) $ 20 $ 30 $ 40 $ 50 $ 60 $ 70 $ 80 $ 90 $ 100 $ 110 $ 120 $ 70
Number of patients (frequency)
Number of patients (frequency) 10 20 30 40 50 60 50 40 30 20 10 Tota 360 Averag
e
70 60 50 40 30 20 10 0
0
0
0
0
0
0
0
0
00
10 $1
$2
$3
$4
$5
$6
$7
$8
$9
$1
Charges
$1
20
290
patient wi pay, where heads represents payment and equas 1, and tais represe
nts nonpayment and equas 0. The charges incurred can be simuated by roing tw
o 6 sided dice and mutipying their tota by $10. The payment for the simuated
patient is cacuated as 0 or 1 times the dice tota times $10. This process is
repeated many times to determine the payment per patient. For exampe, in the f
irst tria a heads is roed and the die tota is 7. The patient payment is: 1 7
$10 = $70 The first ten trias of this simuation are shown in Tabe 10.1, and
a frequency diagram of the output of 100 trias is shown in Figure 10.3. Athoug
h on average each patient pays $35, 50 percent of the patients pay nothing, a sm
a percentage pay as much as $120, and no individua patient pays $35. Using av
erages or most ikey vaues can mask a significant amount of information. Just
as a person can easiy drown in a river that is, on average, ony three feet dee
p, decisions based on averages are often fawed (Savage 2002). Monte Caro simu
ation can revea hidden information and is often used in business and economics
to gain a cearer understanding of the risks and rewards of a situation or decis
ion. Fipping coins, roing dice, and cacuating output for thousands of tria
s woud be tedious; however, a computer can accompish these tasks quicky and e
asiy. There are many commerciay avaiabe software packages for Monte Caro s
imuation; Crysta Ba (Decisioneering Corporation 2006) and @Risk (Paisade C
orporation 2006) are two of the more popuar Microsoft Excebased packages.
TABLE 10.1 Simuation Trias
Tria No. 1 2 3 4 5 6 7 8 9 10
Coin Fip H T H T H T H T H T
Payment 1 0 1 0 1 0 1 0 1 0
Die Tota 7 10 8 8 9 8 7 10 9 10
Charges $70.00 $100.00 $80.00 $80.00 $90.00 $80.00 $70.00 $100.00 $90.00 $100.00
Patient Payment $70.00 $ $80.00 $ $90.00 $ $70.00 $ $90.00 $

291
60 50 40 30 20 10 0
Number of trias (frequency)
FIGURE 10.3 Frequency Diagram (100 Trias)
$0
0
0
0
0
0
0
0
0
00
$2
$3
$4
$5
$6
$7
$8
$9
$1
$1
Payment
Vincent Vaey Hospita and Heath System Exampe
Vincent Vaey Hospita and Heath System (VVH) is trying to decide if it shoud
participate in a new pay for performance (Chapter 3) incentive program offered
by the three argest heath pans in the area. This program focuses on community
acquired pneumonia (CAP), and hospita performance is evauated based on compos
ite measures of quaity reated to this condition. Incentives are based on perfo
rmance: Hospitas scoring above 0.90 wi receive a 2 percent bonus payment on t
op of the standard diagnosisreated group (DRG) payment, hospitas scoring above
0.85 wi receive a 1 percent bonus, hospitas scoring beow 0.65 wi incur a
2 percent reduction in the standard DRG payment, and hospitas scoring beow 0.7
0 wi incur a 1 percent reduction. VVH has coected payment data reated to th
is condition for the previous 36 months. The organization used BestFit to find t
he probabiity distribution that wi best represent these data and determined t
hat the data can be represented fairy we by a norma distribution (Figure 10.
4) with mean () = $250,000/month and standard deviation () = $50,000/month. Becau
e VVH ha no hitorical data related to the ucce or cot of a program to incr
eae quality core related to CAP, expert opinion mut be ued to etimate them
. After dicuion with expert, the VVH team believe that reaching a core of
greater than 0.90 will cot the organization between $10,000 and $50,000, with
any cot in the range being equally likely (i.e., a uniform ditribution).
$1
20
10
292
FIGURE 10.4 CAP Payment Data
NOTE: Created with BetFit 4.5.
VVH alo need to etimate the probability of receiving variou quality core.
A triangular ditribution i fairly eay to etimate from expert opinion. All th
at i needed to define thi ditribution are high, low, and mot likely value.
Again, after talking with variou internal and external expert, VVH believe th
at the mot likely core it will receive will be 0.88, but that core could be a
high a 0.95 or a low a 0.60. A determinitic evaluation of thi ituation,
uing point etimate for all value, reult in the following analyi: Profit
= Revenue Cot Revenue = (Revenue/Month 12 Month/Year) Quality bonu or penalty
= ($250,000/Month 12 Month/Year) 0.01 = $30,000/year Cot = $30,000/year Profi
t = $30,000/year $30,000/year = $0 From thi analyi, it appear that there wil
l be no net gain or lo hould VVH decide to participate in the pay-for-perform
ance program. However, a Monte Carlo imulation of the ituation reveal ome ad
ditional information.
Chapter 10: Simulation

293
The Excel preadheet for thi cenario i hown in Figure 10.5, and the reult
of the firt few imulation trial are hown in Table 10.2. On the firt trial,
@Rik randomly elect from a normal ditribution with mean of $250,000 and ta
ndard deviation of $50,000 and find that the firt month revenue i $155,687 fo
r CAP-related condition. @Rik repeat thi proce for the remaining 11 month
and um thoe month to obtain a total of $2,699,013 for the year. Random ele
ction from a uniform
Point Etimate CAP Quality Score 0.00 0.65 0.70 0.85 0.90 Pay for Performance Fa
ctor 0.02 0.01 0.00 0.01 0.02 Profit = Revenue Cot $ FIGURE 10.5 Monte Carlo Simulation of CAP Pay-forPerformance Program
Quality Score 0.88 Payment or Penalty 0.01 Bonu Payment $ 30,000.00 Cot/ Year
$ 30,000.00
Revenue/ Month $ 250,000.00
Revenue/ Year $ 3,000,000.00
Simulation Uing @Rik Bonu Payment or Penalty Charge 0 = Revenue/ Year * Payment
or Penalty Factor Profit = Revenue Cot $ (30,000.00) = RikOutput() Revenue Cot
Revenue/ Month $ 250,000.00 = RikNormal (250,000, 50,000)
Revenue/ Year $ 3,000,000.00 = Revenue/ Month * 12
Quality Score 0.81 = RikTriang (0.6, 0.88, 0.95)
Payment or Penalty Factor 0.00 = VLOOKUP (D15,B4:C8,2, TRUE)
Cot/ Year $ 30,000.00 = RikUniform (10,000, 50,000)
NOTE: Created with @Rik 4.5.
Output Simulation Iteration/Cell
Revenue/ Month 1 $B$14
Revenue/ Year 1 $C$14
Quality Score 1 $D$14
Cot/ Year 1 $G$14
Profit = Revenue Costs 1 $H$14
1 2 3 4 5
155,687.1563 244,965.375 257,408.3125 335,716.8438 232,497.8281
$2,699,013.25 $2,903,593.00 $2,924,186.25 $3,441,799.25 $2,857,697.00
0.840952277 0.764673352 0.785211325 0.652704477 0.823849738
17,032.68359 15,443.74902 26,655.60938 31,370.79883 46,067.85156
$ (17,032.68) $ (15,443.75) $ (26,655.61) $ (65,788.80) $ (46,067.85)

TABLE 10.2 CAP Pay forPerformance Program Simulation Trials
NOTE: Created with @Risk 4.5.
294
distribution of costs finds a cost of $17,032, and random selection from a trian
gular distribution of quality scores determines a score of 0.84. In this trial V
VH would not be charged a enalty, nor would it receive a bonus. Therefore, VVH
would exerience a total loss (negative rofit) of $17,032, equal to the cost of
the rogram. @Risk erforms many of these trials, collects outut data, and ro
vides statistics and grahs that can be used to analyze this situation. Figure 1
0.6 shows the distribution of rofits for this situation. In contrast to the oi
nt estimate obtained in the deterministic analysis, this analysis shows that if
VVH chooses to articiate in the ay for erformance rogram, it will break eve
n or make a rofit less than 20 ercent of the time. More than 80 ercent of the
time, VVH will incur a loss that could be as large as $108,000. A tornado grah
showing the correlation between each inut and the outut can be a useful analy
tic tool (Figure 10.7). Here, the quality score received has the most imortant
relationshi to rofit (or loss). In this case, FIGURE 10.6 Simulated Distributi
on of Profits for CAP Pay forPerformance Program
DISTRIBUTION FOR PROFIT = REVENUE COSTS/H14 2.500 Mean = 19,998.71 2.000
Values in 10 5
1.500
1.000
0.500
0.000 120 0% 110
90
60 30 0 Values in thousands
30 17.87%
60
0
NOTE: Ceated with @Risk 4.5.
Chapte 10: Simulation

295
REGRESSION SENSITIVITY FOR PROFIT = REVENUE COSTS/H14
Quality scoe/D14
0.815
FIGURE 10.7 Tonado Gaph fo CAP Pay-foPefomance Pogam
0.414
Costs/yea/G14
Revenue/month
12/T14
0.097
1 0.75 0.5 0.25
0
0.25 0.5 0.75
1
Standad coelation coefficient
NOTE: Ceated with @Risk 4.5.
it highlights the impotance of VVHs investment esulting in the achievement of a
high quality scoe. If VVH believes that a $30,000 investment will esult in th
e quality scoes indicated, it may not want to paticipate in the pay-fo-pefo
mance pogam. Altenatively, if VVH decides to paticipate, it is impeative th
at the investment esults in high quality scoes. The VVH team may want to evalu
ate diffeent types of pogams o stategies to incease the pobability of ach
ieving a high quality scoe. Monte Calo analyses ae paticulaly useful in ena
bling oganizations to evaluate the pobable outcomes of decisions, evaluate the
isks involved with those decisions, and hopefully develop stategies to mitiga
te those isks.
Discete Event Simulation
Like Monte Calo simulation, discete event simulation (DES) is based on using p
obability distibutions to epesent andom vaiables. DES, howeve, has its o
ots in queueing, o waiting line, theoy.
Queueing (Waiting Line) Theoy
Although most people ae familia with waiting in line, few ae familia, o eve
n awae of, queueing theoy o the theoy of waiting line. Most peoples expeienc
e with waiting lines is when they ae actually pat of those lines, fo example,
when waiting to check out in a etail envionment. In a manufac-
296
Pefomance Impovement Tools, Techniques, and Pogams
tuing envionment, items wait in line to be woked on. In a sevice envionment
, customes wait fo a sevice to be pefomed. Queues, o lines, fom because t
he esouces needed to seve them (seves) ae limitedit is economically unfeasi
ble to have unlimited esouces. Queueing theoy is used to study systems to det
emine the best balance between sevice to customes (shot o no waiting lines,
implying many esouces o seves) and economic consideations (few seves, i
mplying long lines). A simple queuing system is illustated in Figue 10.8. Cust
omes (often efeed to as entities) aive and ae eithe seved (if thee is
no line) o ente the queue (if othes ae waiting to be seved). Once they ae
seved, customes exit the system. Some impotant chaacteistics of this system
ae discussed below. The custome population, o input souce, can be eithe fi
nite o infinite. If the souce is effectively infinite, the analysis of the sys
tem is easie because simplifying assumptions can be made. The aival pocess i
s chaacteized by the aival patten, the ate at which customes aive (numb
e of customes/unit of time), o inteaival time (time between aivals) and
the distibution in time of those aivals. The distibution of aivals can be
constant o vaiable. A constant aival distibution would have a fixed intea
ival time. A vaiable, o andom, aival patten is descibed by a pobability
distibution. Often, the Poisson distibution (Chapte 7) is used to model ai
vals, and is the mean arriva rate. The queue discipine is the method by which
customers are seected from the queue to be served. Often, customers are served
in the order in which they arrivedfirst come, first served. However, many other q
ueue discipines are possibe, and choice of a particuar discipine can greaty
affect system performance. For exampe, choosing the customer whose service can
be competed most quicky (shortest processing time) usuay minimizes the aver
age time customers spend waiting in ine. This is one reason urgent care centers
are often ocated near an emergency departmenturgent issues can usuay be hand
ed more quicky than true emergencies. The service process is characterized by t
he number of servers and service time. Like arrivas, the distribution of servic
e times can be constant or
FIGURE 10.8 Simpe Queueing System
Buffer or queue Customer popuation, input source Arriva Server(s) Exit

297
variabe. Often, the exponentia distribution (Chapter 7) is used to mode varia
be service times, and is the mean service rate.
Queueing notation The type of queueing system is identified with specific notati
on of the form A/B/c/D/E. The A represents the interarriva time distribution, a
nd B represents the service time distribution. For exampe, A and B are represen
ted with an M (for Markovian) when the arriva distribution is Poisson exponenti
a interarriva distribution (see Chapter 7 for a compete description of these
probabiity distributions) and the service time distribution is exponentia; A a
nd B woud be represented with a D for deterministic, or constant, rates. The c
represents the number of servers, D represents the maximum queue size, and E is
the size of the input popuation. When both queue and input popuation are assum
ed to be infinite, D and E are typicay omitted. An M/M/1 queueing system, ther
efore, has a Poisson arriva distribution, exponentia service time distribution
, singe server, infinite possibe queue ength, and infinite input popuation,
and it is assumed that ony one queue exists. Queueing soutions Anaytic souti
ons for some simpe queueing systems at equiibrium, or steady state (after the
system has been running for some time and is unchanging, often referred to as a
stabe system), have been determined. The derivation of these resuts is outside
the scope of this text; see Cooper (1981) for a compete derivation and resuts
for many other types of queuing systems. Here, the resuts for an M/M/1 queuing
system with < the arriva rate ess than the service rateare presented. Note that
if (customers arrive faster than they are served), the queue wi become infini
tey ong, the number of customers in the system wi become infinite, waiting t
ime wi become infinite, and the server wi experience 100 percent capacity ut
iization. The foowing formuas can be used to determine some characteristics
of the queuing system at steady state.
Capacity utiization = Percentage of time the server is busy =
=
Mean arriva rate 1 Mean time between arrivas a = = e Mean service rate 1/Mean
service time Mean service time = Mean time between arrivas
Average waiting time in queue = Wq =
( )
298
Average time in the system = Average waiting time in queue + average service tim
e =
W s = Wq +
1 1 =
Average ength of queue (or Average number in queue) =
Lq =
2 = ( )
Average tota number of customers in the system =
Ls =
= Ws = Arriva rate Time in the system
This ast resut is caed Littes aw and appies to a types of queueing syste
ms and subsystems. To summarize this resut in pain anguage, in a stabe syste
m or process, the number of things in the system is equa to the rate at which t
hings arrive to the system mutipied by the time they spend in the system. In a
stabe system, the average rate at which things arrive to the system is equa t
o the average rate at which things eave the system. If this were not true, the
system woud not be stabe. Littes aw can aso be restated using other termino
ogy: Inventory (Things in the system) = Arriva rate (or Departure rate) Through
put time (Fow time) and Throughput time = Inventory/Arriva rate Knowedge of t
wo of the variabes in Littes aw aows cacuation of the third variabe. Cons
ider a cinic that services 200 patients in an eighthour day, or an average of 2
5 patients an hour (). The average number of patients in the cinic (waiting room
, exams rooms, etc.) is 15 (I). Therefore, the average throughput time is: T = I
/ 15 Patients = 25 Patients Hour = 0.6 hour Hence, each patient spends an average
of 36 minutes in the cinic.

299
Littes aw has important impications for process improvement and can be seen as
the basis of many improvement techniques. Throughput time can be decreased by d
ecreasing inventory or increasing departure rate. Lean initiatives often focus o
n decreasing throughput time (or increasing throughput rate) by decreasing inven
tory. The theory of constraints (Chapter 6) focuses on identifying and eiminati
ng system bottenecks. The departure rate in any system wi be equa to 1/task
cyce time of the sowest task in the system or process (the botteneck). Decrea
sing the amount of time an object spends at the botteneck task therefore increa
ses the departure rate of the system and decreases throughput time.
VVH M/M/1 queue VVH had been receiving compaints from patients reated to crowd
ed conditions in the waiting area for magnetic resonance imaging (MRI) procedure
s. VVH wanted an average of ony one patient waiting in ine for the MRI. The or
ganization coected data on arriva and service rates and found that for MRI th
e mean service rate () was four patients per hour, exponentiay distributed. VVH
aso found that arrivas foowed a Poisson distribution and that mean arriva
rate was three patients per hour ().
Capacity utiization of MRI = Percentage of time the MRI is busy =
=
3 1 15 Minutes = = 75% or = = = 75% 4 1 20 Minutes
If one customer arrives every 20 minutes, and it takes 15 minutes to perform the
MRI, the MRI wi be busy 75 percent of the time. Average time waiting in ine
= 3 3 Wq = = = = 0.75 Hour ( ) 4(4 3) 4 Average time in the system =
Ws =
1 1 = = 1 Hour 43
Average tota number of patients in the system =
3 = = 3 Patients or 43 Ls = Ws = Arriva rate Time in the system = 3 Patients/Hour

1 Hour = 3 Patients Ls =
300
Average number of patients in waiting ine =
Lq =
=
3 3 2 = = ( ) 4 4 3 32 9 = =
To decrease the average number of patients waiting, VVH needs to decrease the ut
iization, = /, of the MRI process. In other words, the service rate must be incre
ased or the arriva rate decreased. VVH coud increase the service rate by makin
g the MRI process more efficient so that the average time to perform the procedu
re is decreased, and MRIs can be performed on a greater number of patients in an
hour. Aternativey, VVH coud decrease the arriva rate by scheduing fewer pa
tients per hour. To achieve its goa (assuming that the service rate is not incr
eased), VVH needs to decrease the arriva rate to:
Lq =
2 2 = =1 ( ) 4(4 )
2 = 4 (4 ) = 16 4 2 + 4 16 = 0 = 2.47
Aternativey (assuming that the arriva rate is not decreased), VVH coud incre
ase the service rate to: 32 32 = =1 ( 3) ( 3)
Lq =
( 3) = 2 3 = 32 = 9 2 3 9 = 0 = 4.85
VVH coud aso impement some combination of decreasing arriva rate and increas
ing service rate. In a cases, utiization of the MRI woud be reduced to = / = 2
.47/4.00, or 3.00/4.85 = 0.62. Rea systems are sedom as simpe as an M/M/1 que
ueing system and rarey reach equiibrium; often, simuation is needed to study
these more compicated systems.

301
Discrete Event Simuation Detais
DES is typicay performed using commerciay avaiabe software packages. Like
Monte Caro simuation, it is possibe to perform DES simuation by hand; howeve
r, this woud be tedious. Two of the more popuar software packages are Arena (R
ockwe Automation 2006) and Simu8 (Simu8 Corporation 2006). This text focuses
on Arena (an academic version of the software is avaiabe on the CD accompanyi
ng this text), but this shoud not imit students, as most DES software packages
use the same type of ogic as Arena. The terminoogy and ogic of DES are based
on queueing theory. A basic simuation mode consists of entities, queues, and
resources, a of which can have various attributes. Entities are the objects th
at fow through the system; in the heathcare arena entities woud typicay be
patients, but they coud be any object on which some service or task wi be per
formed. For exampe, bood sampes in the hematoogy ab coud aso be entities.
Queues are the waiting ines that hod the entities whie they are waiting for
service. Resources (referred to as servers previousy) coud be peope, equipmen
t, or space for which entities compete. The ogic of a simuation mode is based
on states (variabes that describe the system at a point in time) and events (v
ariabes that change the state of the system). Events are controed by the simu
ation executive, and data are coected on the state of the system as events oc
cur. The simuation jumps through time from event to event. A simpe exampe bas
ed on the VVH M/M/1 MRI queueing discussion wi hep to show the ogic behind D
ES software. Tabe 10.3 contains a ist of the events as they happen in the simu
ation. The arriva rate is three patients per hour, and the service rate is fou
r patients per hour. Random interrarriva times are generated using an exponenti
a distribution with a mean of 0.33 hour. Random service times are generated usi
ng an exponentia distribution with a mean of 0.25 hour (shown at the bottom of
Tabe 10.3). The simuation starts at time 0.00. The first event is the arriva
of the first patient (entity); there is no ine (queue), so this patient enters
service. Upcoming events are the arriva of the next patient at 0.17 hours (the
interarriva between Patients 1 and 2 is 0.17 hour) and the competion of the fi
rst patients service at 0.21 hours. The next event is the arriva of Patient 2 at
0.17 hours. Because the MRI on Patient 1 is not compete, Patient 2 enters the
queue. The MRI has been busy since the start of the simuation, so the utiizati
on of the MRI is 100 percent. Upcoming events are the competion of the first pa
tients service at 0.21 hours and the arriva of Patient 3 at 0.54 hours (the inte
rarriva between Patients 2 and 3 is 0.37 hour, and Patient 2 arrived at 0.17 ho
urs).
302
TABLE 10.3
Attributes Statistics Upcoming Events
Simuation Event List
Just Finished
Variabes
Entity # Utiization 0 1.00 1.00 1.00 1.00 0.34 0.30 1.00 0.88 0.00 0.17 0.00 0.
54 0.17 0.77 0.90 3 0.27 3 0.27 0.21 2 0.27 0.35 1 0.04 0.07 0.00 1 0.04 0.19 0
0 0 0.00 0 0 0
Time
Event Type
Length of Queue Entity # Time
Server Busy
Arriva Time in Queue Average Queue Length
Arriva Time in Service
Number Compete Waits in Queue
Tota Wait Time in Queue
Event
1
0.00
Arr
0
1
2
0.17
Arr
1
1
1
0.21
Dep
0
1
3
0.54
Arr
1
1
2
0.77
Dep
0
1
2 1 1 3 3 2 2 4 3 4 4 4 4 5
0.17 0.21 0.21 0.54 0.54 0.77 0.77 0.90 0.79 0.90 0.90 1.27 1.27 1.49
Arr Dep Dep Arr Arr Dep Dep Arr Dep Arr Arr Dep Dep Arr
3
0.79
Dep
0
0
4
0.90
Arr
0
1
1
2
3
4
5
6
7
8
Interarriva time Time of arriva Service time 0.00 0.21
Expon (0.33)
Expon (0.25)
0.17 0.17 0.56
0.37 0.54 0.02
0.36 0.90 0.37
0.59 1.49 0.34
0.14 1.63 0.11
0.17 1.80 1.02
0.24 2.04 0.01
0.06 2.10 0.20
Arr = Arriva; Dep = Departure; Expon = Exponenet.

303
The first patients MRI is competed at 0.21 hours. There is no one in the queue a
t this point because once Patient 1 has competed service, Patient 2 can enter s
ervice. The tota waiting time in the queue for a patients is 0.04 hour (the d
ifference between when Patient 2 entered the queue and entered service). The ave
rage queue ength is 0.19 patient. There were no peope in ine for 0.17 hour an
d one person in ine for 0.04 hour. 0 Peope 0.17 Hours + 1 Person 0.04 Hours =
0.19 Peope 0.21 Hours Upcoming events are the arriva of Patient 3 at 0.54 hour
s and the departure of Patient 2 at 0.77 hours (Patient 2 entered service at 0.2
1 hours, and service takes 0.56 hour). Patient 3 arrives at 0.54 hours and joins
the queue because the MRI is sti busy with Patient 2. The average queue engt
h has decreased from the previous event because more time has passed with no one
in the queue there has sti ony been one person in the queue for 0.04 hour, bu
t tota time in the simuation is 0.54 hour. Upcoming events are the departure o
f Patient 2 at 0.77 hours and the arriva of Patient 4 at 0.90 hours. Patient 2
departs at 0.77 hours. There is no one in the queue at this point because Patien
t 3 has entered service. A tota of two peope have departed the system. The tot
a wait time in the queue for a patients is 0.04 hour for Patient 2 pus 0.17
hour for Patient 3 (0.77 hour 0.54 hour), a tota of 0.21 hour. The average queu
e ength is: 0 Peope 0.50 Hour + 1 Person 0.21 Hours = 0.35 Peope 0.77 Hours T
he MRI utiization is sti 100 percent because the MRI has been busy constanty
since the start of the simuation. Upcoming events are the departure of Patient
3 at 0.79 hours (Patient 3 arrived at 0.77 hours, and service takes 0.02 hour)
and the arriva of Patient 4 at 0.90 hours. Patient 3 departs at 0.79 hours. Bec
ause no patients are waiting for the MRI, it becomes ide. Upcoming events are t
he arriva of Patient 4 at 0.90 hours and the departure of Patient 4 at 1.27 hou
rs. Patient 4 arrives at 0.90 hours and enters service. The utiization of the M
RI has decreased to 88 percent because it was ide for 0.11 hour of the 0.90 hou
r the simuation has run. Upcoming events are the departure of Patient 4 at 1.27
hours and the arriva of Patient 5 at 1.49 hours. The simuation continues in t
his manner unti the desired stop time is reached.
304
Even for this simpe mode, performing these cacuations by hand woud obvious
y take a ong time. Additionay, one of the advantages of simuation is that it
uses process mapping; many simuation software package such as Arena are abe t
o import and use Microsoft Visio process and vaue stream maps. Using DES softwa
re makes it possibe to buid, run, and anayze simpe modes in imited time. T
herefore, Arena was used to buid and simuate the present mode (Figure 10.9).
(Step by step instructions for buiding and running this mode are avaiabe on
the companion web site at ache.org/books/OpsManagement.) As before, the arriva
rate is three patients per hour, the service rate is four patients per hour, and
both are exponentiay distributed. Averages over time for queue ength, wait t
ime, and utiization for a singe repication are shown in the pots in Figure 1
0.9. Each repication of the simuation is run for 200 hours, and there are 30 r
epications. Repications are needed to determine confidence intervas for the r
eported vaues. Some of the output from this simuation is shown in Figure 10.10
. The sampe mean the hafwidth gives the 95 percent confidence interva for the
mean. Increasing the number of repications wi reduce the haf width.
FIGURE 10.9 Arena Simuation of VVH MRI M/M/1 Queueing Exampe
Patient demand 5 9 1
MRI exam 2 SCANNER
Exit 5 8 9
03 : 57 : 26
MRI UTILIZATION 1.0
0.0 0.0 AVERAGE NUMBER IN QUEUE 3.0 2.0 200.0 AVERAGE WAIT IN QUEUE AND SYSTEM
0.0 0.0 200.0 0.0 0.0 200.0
NOTE: Figure created in Arena, Rockwe Automation.

305
Category Overview 8:22:36 AM
Vaues across a repications
Juy 26, 2006
MRI Exampe Repications: 30 Time unit: Hours
Key Performance Indicators
System Number out Entity Time Wait Time
Patient Average 0.7241 HafWidth 0.08 HafWidth 0.08 Minimum Average 0.5009 Mini
mum Average 0.7427 Maximum Average 1.3496 Maximum Average 1.6174 Minimum Vaue 0
.00 Minimum Vaue 0.00001961 Maximum Vaue 7.3900 Maximum Vaue 7.4140
Average 601
FIGURE 10.10 Arena Output for VVH MRI M/M/1 Queueing Exampe: 200 Hours
Tota Time
Patient
Average 0.9734
Queue Other
Number Waiting MRI exam queue Average 2.1944 HafWidth 0.25 Minimum Average 1.43
26 Maximum Average 4.2851 Minimum Vaue 0.00 Maximum Vaue 29.0000
Resource Usage
Instantaneous Utiization MRI Average 0.7488 HafWidth 0.01 Minimum Average 0.67
Arriva rate = 3 patients/hour; service rate = 4 patients/hour.
The resuts of this simuation agree fairy cosey with the cacuated steady s
tate resuts because the process was assumed to run continuousy for a significa
nt period, 200 hours. A more reaistic assumption might be that MRI procedures a
re ony done for 10 hours every day. The Arena simuation was rerun with this as
sumption, and the resuts are shown in Figure 10.11. The average wait times, que
ue ength, and utiization are ower than the steady state vaues.
VVH M/M/1 Queue
VVH determined that a steady state anaysis was not appropriate for its situatio
n because MRIs are ony offered 10 hours a day. The team decided to anayze this
situation using simuation. Once the mode was buit and run, the mode and sim
uation resuts were compared with actua data and eva
306
FIGURE 10.11 Arena Output for VVH MRI M/M/1 Queueing Exampe: 10 Hours
Category Overview 12:19:03 PM
Juy 26, 2006
System Number out Entity Time Wait Time
Patient Average 0.4778 HafWidth 0.15 HafWidth 0.16 Minimum Average 0.02803444
Minimum Average 0.2407 Maximum Average 1.4312 Maximum Average 1.7611 Minimum Va
ue 0.00 Minimum Vaue 0.00082680 Maximum Vaue 2.9818 Maximum Vaue 3.3129
Average 28
Tota Time
Patient
Average 0.7304
Queue Other
Resource Usage
Arriva rate = 3 patients/hour; service rate = 4 patients/hour.
uated by reevant staff to ensure that the mode accuratey refected reaity. A
staff agreed that the mode was vaid and coud be used to determine what nee
ded to be done to achieve the stated goa. If the mode had not been thought to
be vaid, the team woud have needed to buid and vaidate a new mode. The resu
ts of the simuation (Figure 10.11) indicate that VVH has an average of 1.5 pat
ients in the queue. To reach the desired goa of an average of ony one patient
waiting, VVH needs to decrease the arriva rate or increase the service rate. Us
ing tria and error in the simuation, the organization found that decreasing th
e arriva rate to 2.7 or increasing the service rate to 4.4 woud aow the goa
to be achieved.

307
Even using the improvement toos in this text, the team beieved that it woud o
ny be possibe to increase the service rate of the MRI to 4.2 patients per hour
. Therefore, to reach the goa the arriva rate must aso be decreased. Again us
ing the simuation, VVH found that it woud need to decrease the arriva rate to
2.8 patients per hour. Figure 10.12 shows the resuts of this simuation. The t
eam recommended that a kaizen event be hed for the MRI process to increase serv
ice rate and that appointments for the MRI be reduced to decrease the arriva ra
te. However, the team aso noted that impementing these changes woud reduce th
e average number of patients served from 28 to 26 and reduce the utiization of
the MRI from 0.72 to 0.69. More positivey, however, average patient wait time w
oud be reduced from 0.48 hour to 0.35 hour.
Category Overview 8:24:44 AM
Juy 26, 2006
System Number out Entity Wait Time
Patient Average 0.3507 HafWidth 0.12 HafWidth 0.14 Minimum Average 0.02449931
Minimum Average 0.1899 Maximum Average 1.4202 Maximum Average 1.7825 Minimum Va
ue 0.00 Minimum Vaue 0.00097591 Maximum Vaue 3.4973 Maximum Vaue 4.2210
Average 26
FIGURE 10.12 Arena Output for VVH MRI M/M/1 Queuing Exampe: Decreased Arriva R
ate, Increased Service Rate
Tota Time
Patient
Average 0.6008
Queue Other
Resource Usage
Arriva rate = 2.8 patients/hour; service rate = 4.2 patients/hour; 10 hours sim
uated.
308
VVH was abe to increase the service rate to 4.2 patients per hour and decrease
the arriva rate to 2.8 patients per hour, and the resuts were as predicted by
the simuation. The team aso began to ook at other soutions that woud enabe
VVH to increase MRI utiization whie maintaining wait times and queue ength.
Advanced DES
DES is typicay used to mode more compex situations, processes, and systems t
han the simpe M/M/1 queueing system anayzed previousy. The aim of this text i
s to present an introduction to the modeing process and underying theory to en
abe a basic understanding of simuation. More advanced treatments of Monte Car
o simuation can be found in Practica Management Science: Spreadsheet Modeing
and Appications (Winston and Abright 2005); more advanced discussion of DES ca
n be found in Simuation with Arena (Keton, Sadowski, and Sturrock 2004).
Concusion
Simuation is a powerfu too for modeing processes and systems to evauate cho
ices and opportunities. As is true of a of the toos and techniques presented
in this text, simuation can be used in conjunction with other initiatives such
as Lean or Six Sigma to enabe continuous improvement of systems and processes.
1. Think of at east three simuations that you personay have used or been a p
art of. Why was simuation used? What questions did the simuation hep you to a
nswer? 2. Why is simuation not used more extensivey in heathcare organization
s? 3. What are the advantages of simuation? What are its imitations? 4. Expai
n the faw of averages and provide a specific exampe where this faw adversey af
fected your organization. 5. Discuss at east three opportunities in your organi
zation in which simuation might be usefu in anayzing and improving the situat
ion. 6. Describe severa paces or times in your organization where peope or ob
jects (paperwork, tests, etc.) wait in ine. How do the characteristics of each
exampe differ?

309
Chapter Exercises
1. VVH is considering a pharmacy managed medication assistance program. This pro
gram woud hep patients who are unabe to afford pharmaceuticas obtain free dr
ugs from pharmaceutica manufacturers. VVH woud save the cost of the drugs but
incur costs to manage the program, and the organization woud ike to know if th
e program woud be cost beneficia. VVH beieves that between 60 and 120 patient
s wi use this program, with equa probabiity over this range (a uniform distr
ibution); the average vaue of a patients drugs obtained per year wi most ike
y be $200, but coud be as ow as $0 or as high as $1,000 (a trianguar distribu
tion). The time to administer the program is expected to foow a norma distrib
ution, with a mean of four hours per week and a standard deviation of 0.5 hour p
er week (but never ess than zero hours). VVH aso beieves that there is an 80
percent probabiity the program coud be administered by pharmacy research feo
ws who receive wages and benefits of $30 per hour, but hospita pharmacists may
need to administer the program ($60 per hour wages and benefits). Anayze this s
ituation using @Risk. What shoud VVH do? Why? 2. The hematoogy ab manager has
been receiving compaints that the turnaround time for bood tests is too ong.
Data from the past month show that the arriva rate of bood sampes to one tec
hnician in the ab is five per hour and the service rate is six per hour. Use qu
eueing theory and assume that both rates are exponentiay distributed and that
the ab is at steady state to answer the foowing questions. What is the capaci
ty utiization of the ab? Average number of bood sampes in the ab? Average t
ime that a sampe waits in the queue? Average number of bood sampes waiting fo
r testing? Average time that a bood sampe spends in the ab? 3. Use Arena to a
nayze the above situation. Assume that the ab operates 7 days a week, 24 hours
a day; that arrivas are exponentiay distributed; and that service rate foo
ws a trianguar distribution, with 10 minutes being the most ikey time to comp
ete a bood test (but the test coud require as few as 5 minutes or as ong as
20 minutes). Are the compaints justified? What shoud the ab manager do? 4. De
veop a more reaistic mode for exercise 3. Assume that the ab is open 10 hour
s a day. Assume that bood sampes arrive at a rate of 25 per hour and that five
technicians are avaiabe to run the bood tests. 5. Appying one of the previo
us exampes to your organization, deveop a mode of the process. What data wou
d you need to coect to buid a simuation?
310
References
Axerod, R. 2006. Advancing the Art of Simuation in the Socia Sciences. In Handb
ook of Research on Nature Inspired Computing for Economics and Management, edite
d by J. Rennard. Hershey, PA: Idea Group Reference. Berekamp, E. R., J. H. Conw
ay, and R. K. Guy. 2003. Winning Ways (for Your Mathematica Pays), 2nd ed., Vo
. 2. Weesey, MA: AK Peters. Cooper, R. B. 1981. Introduction to Queueing The
ory, 2nd ed. New York: NorthHoand. [Onine information; retrieved 7/17/06.] ww
w.cse.fau.edu/ %7Ebob/pubications/IntroToQueueingTheory_Cooper.pdf. Decisioneer
ing Corporation. 2006. Crysta Ba. [Onine information; retrieved 5/25/06.] www.
decisioneering.com/. Grenvik, A., and J. Schaefer. 2004. From Resusci Anne to Sim
Man: The Evoution of Simuators in Medicine. Critica Care Medicine 32 (2 Supp
.): S56S57. Haey, L. 2005. Simuation Aims to Speed Patients Through Surgeries. Me
dica Post 41 (1): 23. Keton, W. D., R. P. Sadowski, and D. T. Sturrock. 2004. S
imuation with Arena, 3rd ed. Burr Ridge, IL: McGraw Hi Higher Education. Metr
opois, N. C. 1987. The Beginning of the Monte Caro Method. Los Aamos Science 15
(Specia Issue): 12530. Nationa Academy of Engineering and Institute of Medicin
e. 2005. Buiding a Better Deivery System: A New Engineering/Heath Care Partne
rship, edited by P. P. Reid, W. D. Compton, J. H. Grossman, and G. Fanjiang. Was
hington, D.C.: Nationa Academies Press. [Onine information; retrieved 7/28/06.
] www.nap.edu/cataog/11378.htm. Paisade Corporation. 2006. @Risk 4.5. [Onine
information; retrieved 5/25/06.] www.paisade.com/risk/. Poundstone, W. 1985. Th
e Recursive Universe. Chicago: Contemporary Books. Rockwe Automation. 2006. Are
na. [Onine information; retrieved 6/6/06.] www.arenasimuation.com/products/basi
c_edition.asp. Savage, S. 2002. The Faw of Averages. Harvard Business Review 80 (
11): 2021. Simu8 Corporation. 2006. Simu8. [Onine information; retrieved 6/6/06.
] www.simu8.com/index.htm. Winston, W. L., and S. C. Abright. 2005. Practica
Management Science: Spreadsheet Modeing and Appications (with CD ROM Update),
2nd ed. Bemont, CA: Duxbury.
PART
IV
appications to contemporary heathcare operations issues
CHAPTER
11
PROCESS IMPROVEMENT AND PATIENT FLOW
CHAPTER OUTLINE
Operations Management in Action Overview Probem Types Patient Fow Quaity Proc
ess Improvement Approaches Probem Definition and Process Mapping Process Mappin
g: VVH Emergency Department Process Measurements Toos for Process Improvement E
iminate nonvaue added activities Eiminate dupicate activities Combine reated
activities Process in parae Load baancing Aternative process fow paths an
d contingency pans Critica path Information feedback and reatime contro Qua
ity at the source Let the patient do the work Theory of constraints Process Impr
ovement in Practice Six Sigma Lean Simuation Process Improvement Project: VVH E
mergency Department Phase I Phase II Phase III: Simuation Concusion Discussion
Questions Chapter Exercises References Recommended Reading
312

activity and roe ane map capacity utiization contro oop discrete event simu
ation eectronic medica record (EMR) Emergency Department (ED) entity heijunka
ide or wait time Littes aw oad baancing, eveing nonvaue added time number
of defects or errors patient fow patients in process (PIP) pan do check act (
PDCA) process mapping quaity at the source queue route service or cyce time se
tup time simuation station subprocesses systems tasks theory of constraints thi
ngs in process (TIP) throughput work in process (WIP)
313
Appications to Contemporary Heathcare Operations Issues

The staff of Carondeet St. Marys Hospita in Tucson, Arizona, undertook an impro
vement effort aimed at emergency center operations. A number of operationa chan
ges were impemented to achieve improvements in patient fow. The hospita: Crea
ted a cuture for in house nursing units using pu system versus push system as a c
oncept to promote patient fow out of the emergency center. Deveoped a centrai
zed admitting process with an eectronic tracking board. Impemented a bed contr
o nurse position. Deveoped a mobie admission team to perform bedside admissio
ns. Deveoped and impemented a process for direct admit admission abs to impro
ve turnaround time. Deveoped and impemented a protoco to bypass the emergency
center triage when a bed is avaiabe. Adjusted triage staffing to voume peaks
. Adjusted emergency center staffing to meet voume demand. Deveoped physician
driven triage protocos. Impemented bedside registration in the emergency cente
r when patients bypass triage or arrive by ambuance. Refocused the inpatient un
it charge nurse position to frontine operations with responsibiity for unit f
ow and throughput. Used the express admission unit as a discharge ounge to open
beds and improve patient fow when census is high. Standardized the channes of
communication for foow up with primary care providers whose patients are care
d for by the hospitaist through information systems interfacing. Provided porta
access for physicians to readiy obtain patient information. Opened a fast tra
ck patient care area in the new emergency center to improve fow of ow acuity p
atients. Deveoped a focus team to decrease externa and interna transportation
deays. Standardized surgica admission and preadmission testing processes. Imp
emented processes to reduce operating room turnover time. Standardized the reco
very process between inpatients and outpatients. Deveoped baancing measures to
ensure that changes made to improve patient fow did not adversey affect quai
ty indicators.
314
C h a p t e r 1 1 : Pro c e s s I m p rove m e n t a n d Pa t i e n t F ow
315
These process improvements resuted in the foowing changes: Reduced the emerge
ncy center ength of stay by 7 percent; Increased the emergency centers monthy v
oume by 5 percent; Increased the inpatient daiy census by 20 percent; and Impr
oved the hospitas net operating margin by 1.3 percent above budget.
SOURCE: Schmidt and Messer (2005). Reprinted with permission.
Overview
At the core of a organizations are their operating systems. Exceent organiza
tions continuousy measure, study, and make improvements to these systems. This
chapter provides a methodoogy for measuring and improving systems using a seec
ted set of the toos contained in the preceding chapters. The terminoogy associ
ated with process improvement can be confusing. Typicay, tasks combine to form
subprocesses, subprocesses combine to form processes, and processes combine to
form a system. The boundaries of a particuar system are defined by what is of i
nterest. For exampe, the boundaries of a suppy chain system woud be more enco
mpassing than a hospita system that is part of that suppy chain. The term proce
ss improvement refers to improvement at any of these eves, from the task eve
to the systems eve. This chapter focuses on process and systems improvement. P
rocess improvement foows the cassic pan do check act (PDCA) cyce (Chapters
8 and 9), with the foowing, more specific key steps: Pan: Define the entire p
rocess to be improved using process mapping. Coect and anayze appropriate dat
a for each eement of the process. Do: Use process improvement too(s) to improv
e the process. Check: Measure the resuts of the process improvement. Act to ho
d the gains: If the process improvement resuts are satisfactory, hod the gains
(Chapter 14). If the resuts are not satisfactory, repeat the PDCA cyce. This
chapter discusses the types of probems or issues being faced by heathcare orga
nizations, reviews many of the operations toos discussed in earier chapters, a
nd iustrates how these toos can be appied to process improvement. The reeva
nt toos incude: Basic process improvement toos; Six Sigma and Lean toos; and
Simuation with Arena.
316
Probem Types
Continuous process improvement is essentia for organizations to meet the chae
nges of todays heathcare environment. The theory of swift, even fow (Schmenner
2001, 2004; Schmenner and Swink 1998) asserts that a process is more productive
as the stream of materias (customers or information) fows more swifty and eve
ny. Productivity rises as the speed of fow through the process increases and t
he variabiity associated with that process decreases. It shoud be noted that t
hese phenomena are not independent. Often, decreasing system variabiity wi in
crease fow, and increasing fow wi decrease variabiity. For exampe, advance
d access scheduing increases fow by decreasing the eapsed time between when a
patient schedues an appointment and when she has competed her visit to the pr
ovider. This can aso decrease variabiity by decreasing the number of patient n
o shows. Soutions to many of the probems facing heathcare organizations can b
e found in increasing fow or decreasing variabiity. For exampe, a key operati
ng chaenge in most heathcare environments is the efficient movement of patien
ts within a hospita or cinic, commony caed patient fow. Various approaches t
o process improvement wi be iustrated with the patient fow probem. Optimiz
ing patient fow through emergency departments has become a top priority of many
hospitas; therefore, the Vincent Vaey Hospita and Heath System (VVH) examp
e at the end of this chapter wi focus on improving patient fow through that
organizations emergency department. Another key issue facing heathcare organizat
ions is the need to increase the eve of quaity and eiminate errors in system
s and processes. In other words, variation must be decreased. Finay, increasin
g cost pressures resut in the need for heathcare organizations to not ony imp
rove processes but aso to improve processes whie reducing costs. The toos and
techniques presented in this book are aimed at enabing cost effective process
improvement. Athough this chapter focuses on patient fow and eimination of er
rors reated to patient outcomes, the discussion woud be equay appicabe to
other types of fow probems (e.g., information, paperwork) and other types of e
rrors (e.g., biing). Some toos are more appicabe to increasing fow and oth
ers to decreasing variation, eiminating errors, or improving quaity, but a o
f the toos can be used for process improvement.
Patient Fow
Efficient patient movement in heathcare faciities can significanty improve th
e quaity of care that patients receive and substantiay improve financia perf
ormance. A patient receiving timey diagnosis and treatment has a higher ikeih
ood of obtaining a desired cinica outcome. Because most current
317
payment systems are based on fixed payments per episode of treatment, a patient
moving more quicky through a system wi generate ower costs and, therefore, h
igher margins. Increased interest in patient fow in hospitas has been stimuat
ed by overcrowding in emergency departments and ambuance diversions caused by h
ospitas being fu. In the United States, the average waiting time for nonurgent
visits to emergency departments increased by 33 percent between 1997 and 2000. M
ore than 75 miion Americans reside in areas where ambuances are routiney div
erted from their primary destinations because hospitas do not have avaiabe be
ds (Institute for Heathcare Improvement 2003). The cause of emergency departmen
t crowding goes deeper than the department itsef and invoves patient fow issu
es throughout the hospita. If an operating room is backed up because of schedu
ed surgery deays, an emergency patient needing surgery may have to wait. If a
ab resut is deayed, a patient may wait in the emergency department much onger
than is needed for optima care. Patient fow optimization opportunities occur
in many heathcare settings. Exampes incude operating suites, imaging departme
nts, urgent care centers, and immunization cinics. Advanced access (same day sc
heduing) is a specia case of patient fow and is examined in depth in Chapter
12. There are many causes of poor patient fow, but a cuprit discovered by many
investigators is variabiity of schedued demand. For exampe, if an operating
room is schedued for a surgery but the procedure does not take pace at the sch
edued time, or takes onger than it was schedued to take, the rest of the surg
ery schedue becomes deayed. These deays rippe through the entire hospita, i
ncuding the emergency department. As expained by Dr. Eugene Litvak (Institute
for Heathcare Improvement 2006),
You have two patient fows competing for hospita bedsICU or patient foor beds.
The first fow is schedued admissions. Most of them are surgica. The second f
ow is medica, usuay patients through the emergency department. So when you ha
ve a peak in eective surgica demand, a of a sudden your resources are being
consumed by those patients. You dont have enough beds to accommodate medica dema
nd.
If schedued surgica demand varies unpredictaby, the ikeihood of inpatient o
vercrowding, emergency department backogs, and ambuance diversions increases d
ramaticay. Even when patients are admitted into a bed, they may not be receivi
ng a vaue added service. A British study (Waey, Sivester, and Steyn 2006) fo
und that, athough Nationa Heath Service hospitas were running at an occupanc
y rate of 90 percent, ony 60 to 70 percent of the patients occupying those beds
were receiving active treatment and the rest are either waiting for initia visi
ts by doctors, stuck in the system because of deays, or not i but have
318
not eft the hospita for some reason. Ceary, significant opportunities to impr
ove capacity and reduce cost by improving patient fow exist. A number of manage
ment soutions have been introduced to improve patient fow. Separating ow acui
ty patients into a unique treatment stream can reduce time in the emergency depa
rtment and improve patient satisfaction (Rodi, Grau, and Orsini 2006). Once a pa
tient is admitted to the hospita, other toos have been empoyed to improve fo
w, especiay around the discharge process. These incude creating a uniform dis
charge time (e.g., 11 a.m.), writing discharge orders the night before, communic
ating discharge pans eary in the patients care, centraizing oversight of censu
s and patient movement, changing physician rounding times, aerting anciary de
partments when their testing procedures are critica to a patients discharge, and
improving discharge coordination with socia services (Cark 2005). However, fo
r patient fow to be carefuy managed and improved, the forma methods of proce
ss improvement outined beow need to be widey empoyed.
Quaity
Ensuring quaity of care is becoming an increasingy important probem for the h
eathcare industry. It was estimated that in 1999 more than 98,000 individuas d
ied in hospitas because of errors made by heathcare professionas (Institute o
f Medicine 1999). Both consumers and heathcare providers have begun to recogniz
e that errors in patient care need to be reduced or eiminated. Because the admi
nistration and deivery of heathcare invoves the interaction of providers and
patients, this is a more difficut probem than is encountered in most environme
nts. Human invovement in the production process and the intangibiity of the resu
ting product make it difficut to standardize and contro heathcare. As noted
in To Err Is Human (Institute of Medicine 1999), heathcare differs from a syste
matic production process mosty because of huge variabiity in patients and circu
mstances, the need to adapt processes quicky, the rapidy changing knowedge ba
se, and the importance of highy trained professionas who must use expert judgm
ent in dynamic settings. However, because the consequences of faiurepatient injur
y or deathare so great, the need to reduce or eiminate faiures is even more imp
ortant in heathcare than in manufacturing or other service industries. Again, t
o ensure quaity and reduce or eiminate errors, the forma methods of process i
mprovement outined beow need to be widey empoyed.
Process Improvement Approaches
Process improvement projects can use a variety of approaches and toos. Typica
y, they begin with process mapping and measurement. Some simpe toos
319
can be initiay appied to identify opportunities for improvements. Identifying
and eiminating or aeviating bottenecks in a system (theory of constraints)
can quicky improve overa system performance. In addition, the Six Sigma toos
described in Chapter 8 can be used to reduce variabiity in process output, and
the Lean toos discussed in Chapter 9 can be used to identify and eiminate was
te. Finay, simuation (Chapter 10) provides a powerfu too to understand and
optimize fow in a system. A major process improvement projects shoud use the
forma project management methodoogy outined in Chapter 5. An important first
step is to identify a systems owner: For a system to be managed effectivey over
time it must have an owner, someone who monitors the system as it operates, co
ects performance data, and eads teams to improve the system. Many systems in h
eathcare do not have an owner and, therefore, operate inefficienty. For examp
e, a patient may enter an emergency department, see the triage nurse, move to th
e admitting department, take a chair in the waiting area, be moved to an exam ro
om, be seen by a nurse, have his bood drawn, and finay be examined by a physi
cian. From the patients point of view this is one system, but many hospita depar
tments may be operating autonomousy. System ownership probems can be remedied
by mutidepartment teams with one individua designated as the overa system or
process owner.
Probem Definition and Process Mapping
The first step in improving a system is process description and mapping. However
, the team shoud first ensure that the correct probem is being addressed. Mind
mapping or root cause anaysis shoud be empoyed to ensure that the probem is
identified and framed correcty; much time and money can be wasted finding an o
ptima soution to something that is not a probem. For exampe, suppose that a
project team was given the task of improving customer satisfaction with the emer
gency department. If the team assumes that customer satisfaction is ow because
of high throughput time, they might proceed to optimize patient fow in the emer
gency department. However, if an anaysis of customer satisfaction showed that c
ustomers were dissatisfied because of a ack of parking, the soution to the pro
bem woud foow a different path. The probem must be ceary understood and d
efined to determine what process to map. Processes can be described in a number
of ways. The most common is the written procedure or protoco, typicay constru
cted in the directions stye, which is sufficient for simpe proceduresfor exampe,
Turn right at Em Street, go two bocks, and turn eft at Vine Avenue. Ceary wr
itten procedures are an important part of defining standard work, as described i
n Chapter 9. However, when processes are inked to form systems they become more
compex. These inked processes benefit from process mapping because process ma
ps:
320
Provide a visua representation that offers an opportunity for process improveme
nt through inspection; Aow for branching in a process; Provide the abiity to
assign and measure the resources in each task in a process; and Are the basis fo
r modeing the process via computer simuation software. Chapter 6 provides an i
ntroduction to process mapping. To review, the steps in process mapping are: 1.
Assembe and train the team. 2. Determine the boundaries of the process (where d
oes it start and end?) and the eve of detai desired. 3. Brainstorm the major
process tasks and ist them in order. (Sticky notes are often hepfu here.) 4.
Once an initia process map (aso caed a fowchart) has been generated, the ch
art can be formay drawn using standard symbos for process mapping. 5. The for
ma fowchart shoud be checked for accuracy by a reevant personne. 6. Depen
ding on the purpose of the fowchart, data may need to be coected or more info
rmation added.
Process Mapping: VVH Emergency Department
A basic process map iustrating patient fow in the VVH emergency department is
dispayed in Figure 11.1. In this basic process map, the patient arrives at the
emergency department and is examined by the triage nurse. If the patient is ver
y i (high compexity eve), she is immediatey sent to the intensive care sec
tion of the emergency department. If not, she is sent to admitting and then to t
he routine care section of the emergency department. The simpe process map show
n here ends with this step. In actuaity, other processes woud now begin, such
as admission into an inpatient bed or a discharge from the emergency department
to home with a schedued cinica foow up. The VVH emergency department proces
s improvement project is detaied at the end of this chapter.
Process Measurements
Once a process map is deveoped, reevant data shoud be coected and anayzed.
The situation wi dictate which specific data and measures shoud be empoyed.
Important measures and data for possibe coection and anaysis incude:
321
Patient arrives at the ED
Triage cinica
Compexity
High
Intensive ED care
FIGURE 11.1 VVH Emergency Department (ED) Patient Fow Process Map
Low Admitting private insurance Yes Triage financia Private insurance No
Waiting
Waiting
Admitting Medicaid
Nurse history/ compaint
Waiting
Exam/ treatment
Waiting
Discharge
End
NOTE: Created with Microsoft Visio.
Capacity of a process is the maximum possibe amount of output (goods or service
s) that a process or resource can produce or transform. Capacity measures can be
based on outputs or on the avaiabiity of inputs. The capacity of a series of
tasks is determined by the owest capacity task in the series. Capacity utiizat
ion is the proportion of capacity actuay being used. It is measured as actua
output/maximum possibe output. Throughput time is the average time a unit spend
s in the process. Throughput time incudes both processing time and waiting time
and is determined by the critica (ongest) path through the process. Throughpu
t rate, sometimes referred to as drip rate, is the average number of units that
can be processed per unit of time.
322
Service time or cyce time is the time to process one unit. The cyce time of a
process is equa to the ongest task cyce time in that process. The probabiity
distribution of service times may aso be of interest. Ide or wait time is the
time a unit spends waiting to be processed. Arriva rate is the rate at which u
nits arrive to the process. The probabiity distribution of arriva rates may a
so be of interest. Work in process, things in process, patients in process, or i
nventory describe the tota number of units in the process. Setup time is the am
ount of time spent getting ready to process the next unit. Vaue added time is t
he time a unit spends in the process where vaue is actuay being added to the
unit. Nonvaue added time is the time a unit spends in the process where no vaue
is being added. Wait time is nonvaue added time. Number of defects or errors. T
he art in process mapping is to provide enough detai to be abe to measure over
a system performance, determine areas for improvement, and measure the effect
of these changes.
Toos for Process Improvement
Once a system has been mapped, a number of techniques can be used to improve the
process. These improvements wi resut in a reduction in the duration, cost, o
r waste in a system.
Eiminate nonvaue added activities The first step after a system has been mapped
is to evauate every eement to ascertain whether each one is truy necessary a
nd provides vaue (to the customer or patient). If a system has been in pace fo
r a ong period and has not been evauated through a forma process improvement
project, a number of eements of the system can ikey be easiy eiminated. Thi
s step is sometimes termed harvesting the ow hanging fruit. Eiminate dupicate a
ctivities Many processes in systems have been added on top of existing systems w
ithout formay evauating the tota system, frequenty resuting in dupicate a
ctivities. The most infamous redundant process step in heathcare is asking pati
ents repeatedy for their contact information. Dupicate activities increase bot
h time and cost in a system and shoud be eiminated whenever possibe. Combine
reated activities Process improvement teams shoud examine both the process map
and the activity and roe ane map. If a patient moves back and forth between d
epart
323
ments, the movement shoud be reduced by combining these activities so he ony n
eeds to be in each department once.
Process in parae Athough a patient can ony be in one pace at one time, oth
er aspects of her care can be competed simutaneousy. For exampe, medication
preparation, physician review of tests, and chart documentation coud a be don
e at the same time. As more tasks are executed simutaneousy, the tota time a
patient spends in the process wi be reduced. Simiar to a chef who has a numbe
r of dishes on the stove synchronized to be competed at the same time, much of
the patient care process can be competed at the same time. Another eement of p
arae processing is the reationship of subprocesses to the main fow. For exa
mpe, a ab resut may need to be obtained before a patient enters the operating
suite. Many of these subprocesses can be synchronized through the anaysis and
use of takt time (Chapter 9). This synchronization wi enabe efficient process
fow and hep optimize the process. Load baancing If simiar workers perform t
he same task, a we tuned system can be designed to baance the work among them
. For exampe, a mass immunization cinic woud want to deveop its system so th
at a immunization stations are active at a times. This coud be accompished
by using a singe queue with mutipe immunization stations. Load baancing (or
oad eveing, heijunka) is difficut when empoyees can ony perform a imited
set of specific tasks (an unfortunate consequence of the super speciaization o
f the heathcare professions). When cross training of empoyees can be accompis
hed, oad baancing is easier. Aternative process fow paths and contingency p
ans The number and pacement of decision points in the process shoud be evauat
ed and optimized. A system with few decision points has few aternative paths an
d, therefore, does not respond we to unexpected events. Aternative paths (or
contingency pans) shoud be deveoped for these types of events. For exampe, a
standard cinic patient rooming system shoud have aternative paths if an emer
gency occurs, a patient is ate, a provider is deayed, or medica records are a
bsent. Critica path For compex pathways in a system, it is sometimes hepfu t
o identify the critica pathway using toos described in Chapter 5. If a critica
path can be identified, execution of processes on the pathway can be improved
(e.g., reduce average service time). In some cases, the process can be moved off
the criti
324
ca path and be processed in parae to it. Either technique wi decrease the
tota time on the critica pathway. In the case of patient fow, this wi decre
ase the patients tota time in the system.
Information feedback and rea time contro Some systems have a high eve of var
iabiity in their operations because of variabiity in the arriva of jobs or cu
stomers (patients) into the process and variabiity of the cyce time of each pr
ocess in the system. High variabiity in the system can ead to poor performance
. One too to reduce variabiity is the contro oop (Chapter 1). Information ca
n be obtained from one process and used to drive change in another. For exampe,
the number of patients in the emergency department waiting area can be continuo
usy monitored and if the number reaches a certain eve, contingency pans such
as foating in additiona staff from other portions of the hospita can be init
iated. Quaity at the source Many systems contain mutipe reviews, approvas, a
nd inspections. A system in which the task is done right the first time shoud n
ot require these redundancies. Deming first identified this probem in the proce
ss design of manufacturing ines that had inspectors throughout the assemby pro
cess (Deming 1998). This expensive and ineffective system was one of the factors
that gave rise to the quaity movement in Japan and, ater, in the United State
s. To eiminate inspections, it is important to design systems that embed quait
y at the source or beginning of a system. For exampe, a biing system that req
uires a cerk to inspect a bi before it is reeased does not have quaity bui
t into the process. A system that has many inspection steps and does not have qu
aity at the source shoud be redesigned. Let the patient do the work The Intern
et and other advanced information technoogies have aowed for more sef servic
e in service industries. Individuas are now comfortabe booking their own airi
ne reservations, buying goods onine, and checking themseves out at retaiers.
This trend can be expoited in heathcare with toos that enabe patients to be
part of the process. For exampe, onine toos are now avaiabe that aow pati
ents to make their own cinic appointments. Letting the patient do the work redu
ces the work of staff and provides an opportunity for quaity at the sourceif the
patients input the data, they are more ikey to be correct. Theory of constrai
nts Chapter 6 discussed the underying principes and appications of the theory
of constraints, which can be used as a powerfu process improvement too. First
, the botteneck in a system is identified, often through the observation
325
of queues forming in front of it. Once a botteneck is identified, it shoud be
expoited and everything ese in the system subordinated to it. This means that
other non botteneck resources (or steps in the process) shoud be synchronized
to match the output of the constraint. Ideness at a non botteneck resource cos
ts nothing and non bottenecks shoud never produce more than can be consumed by
the botteneck resource. Often, this wi cause the botteneck to shifta new bot
teneck wi be identified. However, if the origina botteneck remains, the pos
sibiity of eevating the botteneck needs to be evauated. Eevating botteneck
s requires additiona resources (e.g., staff, equipment), so a comprehensive out
comes and financia anaysis needs to be undertaken to determine the trade offs
among process improvement, quaity, and costs.
Process Improvement in Practice
Six Sigma
If the primary goa of a process improvement project is to improve quaity (redu
ce the variabiity in outcomes), the Six Sigma approach and toos described in C
hapter 8 wi yied the best resuts. As discussed previousy, Six Sigma uses se
ven basic toos: fishbone diagrams, check sheets, histograms, Pareto charts, fo
w charts, scatter pots, and run charts. It aso incudes statistica process co
ntro to provide an ongoing measurement of process output characteristics to ens
ure quaity and enabe the identification of a probem situation before an error
occurs. The Six Sigma approach aso incudes measuring process capabiitya measu
re of whether a process is actuay capabe of producing the desired outputand be
nchmarking it against other simiar processes in other organizations. Quaity fu
nction depoyment is used to match customer requirements (voice of the customer)
with process capabiities given that trade offs must be made. Poka yoke is emp
oyed seectivey to mistake proof parts of a process.
Lean
Process improvement projects focused on eiminating waste and improving fow in
the system or process can use many of the toos that are part of the Lean approa
ch (Chapter 9). The kaizen phiosophy, which is the basis for Lean, incudes the
foowing steps: Specify vaue: Identify activities that provide vaue from the
customers perspective. Map and improve the vaue stream: Determine the sequence
of activities or the current state of the process and the desired future state.
Eiminate nonvaue added steps and other waste.
326
Fow: Enabe the process to fow as smoothy and quicky as possibe. Pu: Enab
e the customer to pu products or services. Perfection: Repeat the cyce to en
sure a focus on continuous improvement. An important part of Lean is vaue strea
m mapping, which is used to define the process and determine where waste is occu
rring. Takt time is a measure of time needed for the process based on customer d
emand and can be used to synchronize fow in a process. Standardized work, an im
portant part of the Lean approach, is written documentation of the precise way i
n which every step in a process shoud be performed and is a way to ensure that
things are done the same way every time in an efficient manner. Other Lean toos
in incude the five Ss (a technique to organize the workpace) and spaghetti di
agrams (a mapping technique to show the movement of customers, patients, workers
, equipment, jobs, etc.). Leveing workoad (heijunka) so that the system or pro
cess can fow without interruption can be used to improve the vaue stream. Kaiz
en bitzes or events are Lean toos used to improve the process quicky, when pr
oject management is not needed (Chapter 9).
Simuation
Simuation is used to evauate what if situations. Usuay it is ess expensive or
speedier than changing the rea system and evauating the effects of those chan
ges. The process of simuation is: Mode deveopment: Deveop a mode of the pro
cess or situation of interest. Mode vaidation: Ensure that the mode accurate
y represents reaity. Simuation and anaysis of output: Run the simuation and
anayze the output to determine the answers to the questions asked, optimize the
process, or manage risk. Discrete event simuation (based on queueing theory) i
s used to mode system fows to improve the system. Chapter 10 provided an exten
sive description of the underying mathematics and the use of simuation to mode
operating systems. This chapter appies these concepts more broady to process
improvement, with a specific emphasis on patient fow.
Process Improvement Project: VVH Emergency Department
To demonstrate the power of many of the process improvement toos described prev
iousy, an extensive patient fow process improvement project at VVH wi be exa
mined. VVH identified patient fow in the emergency department as an important a
rea on which to focus process improvement efforts. The goa of
327
the project was to reduce tota patient time in the emergency department (both w
aiting and care deivery) whie maintaining or improving financia performance.
The first step for VVH eadership was to charter a mutidepartmenta team using
the project management methods described in Chapter 5. The head nurse for emerge
ncy services was appointed project eader. The team fet VVH shoud take a numbe
r of steps to improve patient fow in the emergency department and decided to sp
it the systems improvement project into three major phases. First, they woud p
erform simpe data coection and basic process improvement to identify ow hang
ing fruit and make obvious, straightforward changes. Once the team fet comforta
be understanding the basics of patient fow in the department, they woud work
to understand the eements of the system more fuy by coecting more detaied
data. Then, vaue stream mapping and the theory of constraints woud be used to
identify opportunities for improvement. Root cause anaysis woud be empoyed on
poory performing processes and tasks. These changes woud be made and their ef
fects measured. The third phase of the project was the use of simuation. Becaus
e the team woud have compete knowedge of patient fow in the system, they cou
d deveop and test a simuation mode with confidence. Once the simuation was
vaidated, the team woud continuousy test process improvements in the simuati
on mode and impement them in the emergency department. The team aso considere
d a fourth phase that woud impement reatime use of a simuation mode and mat
ch patient arrivas with emergency department resource needs, then effectivey d
epoy those resources. The specific high eve tasks in this project were: Phase
I 1. Observe patient fow and deveop a detaied process map. 2. Measure high
eve patient fow metrics for one week: Patients arriving per hour; Patients dep
arting per hour to inpatient; Patients departing per hour to home; and Number of
patients in the emergency department, incuding the waiting area and exam rooms
. 3. With the process map and data in hand, use simpe process improvement techn
iques to make changes in the process, then measure the resuts. Phase II 4. Set
up a measurement system for each individua process and take measurements over o
ne week.
328
5. Use vaue stream mapping and the theory of constraints to anayze patient fo
w and make improvements, then measure the effects of the changes. Phase III 6. C
oect data needed to buid a reaistic simuation mode. 7. Deveop the simuat
ion mode and vaidate it against rea data. 8. Use the simuation mode to cond
uct virtua experiments on process improvements. Impement promising improvement
s, and measure the resuts of the changes.
Phase I VVH process improvement project team members observed patient fow and r
ecorded the needed data. With the information coected, the team was abe to cr
eate a fairy detaied process map. They measured the foowing higheve operat
ing statistics reated to patient fow:
Patients arriving per hour = 10 Patients departing per hour to inpatient = 2 Pat
ients triaged to routine emergency care per hour = 8 Patients departing per hour
to home = 8 Average number of patients in various parts of the system (samped
every 10 minutes) = 20 Average number of patients in emergency department exam r
ooms = 4 Using Littes aw, the average time in the emergency department (through
put time) is cacuated as: Throughput time = T =I 24 Patients = 8 Patients Hour
= 3 Hours Hence, each patient spent an average of 3 hours, or 180 minutes, in t
he emergency department. However, Littes aw ony gives the average time in the
department at steady state. Therefore, the team measured tota time in the syste
m for a sampe of routine patients and found an average of 165 minutes. They as
o observed that the number of patients in the waiting room varied from 0 to 20,
and that the actua time it took to move through the process varied from 1 to mo
re than 5 hours. Initiay, the team focused on the emergency department admitti
ng subsystem as an opportunity for immediate improvement. Figure 11.2 shows the
compete emergency department system with the admitting subsystem highighted.
329
Triage cinica
Compexity
High
Intensive ED care
FIGURE 11.2 VVH Emergency Department (ED) Admitting Subsystem
Low
Waiting
Triage financia
Admitting Subsystem
Private insurance No
Yes
Admitting private insurance
Waiting
Routine ED care
End
Admitting Medicaid
NOTE: Created with Microsoft Visio.
The team deveoped the foowing description of the admitting process from its d
ocumentation of patient fow:
Patients who did not have an acute cinica probem were asked if they had heat
h insurance. If they did not have heath insurance, they were sent to the admitt
ing cerk who speciaized in Medicaid (to enro them in a Medicaid program). If
they had heath insurance, they were sent to the other cerk, who speciaized i
n private insurance. If a patient had been sent to the wrong cerk by triage, he
was sent to the other cerk.
The team determined that one process improvement change woud be to cross train
the admitting cerks on both private insurance and Medicaid eigibiity. This wo
ud provide for oad baancing, as patients woud automaticay
330
go to the free cerk and keep the workoad baanced. In addition, the system imp
rovement woud eiminate triage staff errors in sending patients to the wrong c
erk, hence providing quaity at the source.
Phase II Phase I produced some gains in reducing patient time in the emergency d
epartment. However, the team fet more detaied data were needed to improve even
further. As a first step in coecting these data, the team decided to measure
various parameters of the departments processes. Initiay, they decided to focus
on the time period from 2 p.m. to 2 a.m., Monday through Thursday, as this is t
he busy period in the emergency department and demand seemed reativey stabe d
uring these times. The team drew a more detaied process map (Figure 11.3) and p
erformed vaue stream mapping of this process (Figure 11.4). First, they evauat
ed each step in the process to determine if it was vaue added, nonvaue added, o
r nonvaue added but necessary. Then, they measured the time a patient spent at e
ach step in the process. The team found that after a patient had given his insur
ance information, he spent an average of 30 minutes of nonvaue added time in the
waiting room before a nurse was avaiabe to take his history and record the pr
esenting compaint, a process that took an average of 20 minutes. The percentage
of vaue added time for these two steps is:
(Vaue added time/Tota time)100 = (20 Minutes/[30 Minutes + 20 Minutes])100 = 4
0% The team beieved the waiting room process coud be improved through automati
on. Patients were given a tabet persona computer in the waiting area and asked
to enter their symptoms and history via a series of branched questions. The res
uts were sent via a wireess network to VVHs eectronic medica record (EMR). Th
is step usuay took patients about 20 minutes. Staff knew which patients had co
mpeted the eectronic interview by checking the EMR and coud then prioritize w
hich patient woud be seen next. This new procedure aso reduced the time needed
by the nurse to 10 minutes because it enabed the nurse to verify, rather than
actuay record, presenting symptoms and patient history. The percentage of vau
e added time for the new procedure is: (Vaue added time/Tota time)100 = ([Pati
ent history time + Nurse history time]/ [Patient history time + Wait time + Nurs
e history time])100 = ([20 Minutes + 10 Minutes]/[20 Minutes + 10 Minutes + 10 M
inutes])]100 = 75%
331
Triage cinica
Compexity
High
Intensive ED care
FIGURE 11.3 VVH Emergency Department (ED) Process Map: Focus on Waiting and Hist
ory
Low
Waiting
Admitting
Waiting Focus
Nurse history/ symptoms
Exam/ treatment
Discharge
End
NOTE: Created in Microsoft Visio.
The average throughput time for a patient in the emergency department was reduce
d by 10 minutes. The average time for patients to fow through the department (t
hroughput time) prior to this improvement was 155 minutes. Because this step was
on the critica path of the compete routine care emergency department process,
throughput time for noncompex patients was reduced to 145 minutes, a 7 percent
productivity gain. An anayst from the VVH finance department (a member of the
project team) was
332
FIGURE 11.4 VVH Emergency Department (ED): Vaue Steam Map: Focus on Waiting and
History
Intensive ED care
Patients
Arriva 12 rate #/ hr
Triage
Cyce time FTEs 5 m 1 #
Admitting (insurance)
Nurse (history)
Exam/ treatment
Cyce time FTEs Firsttime correct m # %
Discharge
Cyce time FTEs Firsttime correct m # %
Firsttime nm % correct
0 min 5 min 9 min
30 min 20 min
Hr Hr
Hr Hr
Created with eVSM software from GumshoeKI, Inc., a Microsoft Visio add on. NOTE:
NM = Number; FTE = fu time equivaent.
abe to ceary demonstrate that the capita/software costs for the tabet compu
ters woud be recovered within 12 months by the improvement in patient fow. Thi
s phase of the project used three of the basic process improvement toos discuss
ed in this chapter: Have the customer (patient) do it; Provide quaity at the so
urce; and Gain information feedback and rea time contro. Athough the process
improvements aready undertaken had a visibe effect on fow in the emergency de
partment, the team beieved more improvements were possibe. There were bottene
cks in the process, as evidenced by two waiting ines, or queues: (1) the waitin
g room queue, where patients waited before being moved to an exam room; and (2)
the most visibe queue for routine patients, the discharge area, where patients
occasionay had to stand because a of the areas chairs were occupied. In the d
ischarge area, patients waited a significant amount of time for fina instructio
ns and prescriptions. The theory of constraints suggests that the botteneck sho
ud be identified and optimized. However, aeviating or eiminating the patient
examination and treatment or discharge bottenecks woud require significant ch
anges in a ong standing process. Because this process improvement step seemed t
o have the probabiity of a high payoff but was a significant departure from exi
sting practice, it was decided to move to phase III of the project and use simu
ation to mode different options to improve patient fow in the examination/trea
tment and discharge processes.
333
Phase III: Simuation First, the team reviewed the basic terminoogy of simuati
on.
An entity is what fows through a system. Here, the entity is the patient. Howev
er, in other systems the entity coud be materias (e.g., bood sampe, drug) or
information (e.g., diagnosis, biing code). Entities usuay have attributes t
hat wi affect their fow through the system (e.g., mae/femae, acute/chronic
condition). Each individua process in the system transforms (adds vaue) to the
entity being processed. Each process takes time and consumes resources such as
staff, equipment, suppies, and information. Time and resource use can be define
d as an exact vaue (e.g., ten minutes) or a probabiity distribution (e.g., nor
mamean, standard deviation). Chapter 7 discusses probabiity distributions in de
tai. Most heathcare tasks and processes wi not require the same amount of ti
me each time they are performedthey wi require a variabe amount of time. These
variabe usage rates are best described as probabiity distributions. The geogr
aphic ocation of a process is caed a station. Entities fow from one process
to the next via routes. The routes can branched based on decision points in the
process map. Finay, because a process may not be abe to hande a incoming e
ntities in a timey fashion, queues wi occur at each process and can be measur
ed and modeed. The team next deveoped a process map and simuation mode for r
outine patient fow (Figure 11.5) in the emergency department using Arena simua
tion software from Rockwe Automation (2006) (see ache.org/books/OpsManagement
for inks to videos detaiing this mode and its operation). The team focused on
routine patients rather than those requiring intensive emergency care because o
f the high proportion of routine patients seen in the department. Routine patien
ts are checked in and their sef recorded history and presenting compaint(s) ar
e verified by a nurse. Then, patients move to an exam/treatment room and, fina
y, to the discharge area. Of the ten patients who come to the emergency departme
nt per hour, eight foow this process. Next, to buid a simuation mode that a
ccuratey refects this process, the team needed to determine the probabiity di
stributions of treatment time, admitting time, nurse history time, discharge tim
e, and arriva rate for routine patients. To determine these probabiity distrib
utions, they coected data on time of arriva in the department and time to per
form each step in the routine patient care process. Probabiity distributions we
re determined using the Input Anayzer function in Arena. Input Anayzer takes r
aw input data and finds the bestfitting probabiity distribution for them. Figur
e 11.6 shows the output of
334
FIGURE 11.5 VVH Emergency Department (ED) Initia State Simuation Mode
Patient arrives
Triage
Compexity
True
Intensive ED care
Fase Discharge area Admitting
Leave ED
Patient history Discharge Nurse history Waiting room Exam and treatment
NOTE: Mode created in Arena, Rockwe Automation.
Arena Input Anayzer for 500 observations of treatment time for emergency depart
ment patients requiring routine care. Input Anayzer suggested that the best fit
ting probabiity distribution for these data was trianguar, with a minimum of 9
minutes, mode of 33 minutes, and maximum of 51 minutes. The remaining data were
anayzed in the same manner, and the foowing best fitting probabiity distrib
utions were determined: Emergency routine patient arriva rate: exponentia (7.5
minutes between arrivas) FIGURE 11.6 Examination and Treatment Time Probabiit
y Distribution: Routine Emergency Department Patients
Number of occurrences
24 12
9
33 Treatment time (minutes)
51
335

Triage time: trianguar (2, 5, 7 minutes) Admitting time: trianguar (3, 8, 15 m
inutes) Patient history time: trianguar (15, 20, 25 minutes) Nurse history time
: trianguar (5, 11, 15 minutes) Exam/treatment time: trianguar (14, 36, 56 min
utes) Discharge time: trianguar (9, 19, 32 minutes)
The Arena mode simuation was based on 12 hour intervas (2 p.m. to 2 a.m.) and
repicated 100 times. Note that increasing the number of repications decreases
the haf width and, therefore, gives tighter confidence intervas. The number o
f repications needed depends on the desired confidence interva for the outcome
variabes. However, as the mode becomes more compicated, more repications wi
take more simuation time; this mode is fairy simpe, so 100 repications t
ake itte time and are more than sufficient. Most simuation software, incudin
g Arena, has the capabiity of using different arriva rate probabiity distribu
tions for different times of the day and days of the week, aowing for varying
demand patterns. However, the team beieved that this simpe mode using ony on
e arriva rate probabiity distribution represented the busiest time for the eme
rgency department, having observed that by 2 p.m. on weekdays there were usuay
no queues in either the waiting room or discharge area. The resuts of the simu
ation were reviewed by the team and compared with actua data and observations
to ensure that the mode was, in fact, simuating the reaity of the emergency d
epartment. The team was satisfied that the mode accuratey refected reaity. T
he focus of this simuation is the queueing occurring at both the waiting room a
nd in the discharge area and the tota time in the system. Figure 11.7 shows the
resuts of this base (current status) mode. On average, a patient spent 2.4 ho
urs in the emergency department. The team decided to examine the discharge proce
ss in depth because patient waiting time was greatest there. The emergency depar
tment had two rooms devoted to discharge and used two nurses to hande a disch
arge tasks, such as making sure prescriptions are given and home care instructio
ns are understood. However, because there were ony two nurses and four exam roo
ms, queuing was inevitabe. In addition, the patient treatment information neede
d to be handed off from the treatment team to the discharge nurse. The process i
mprovement team therefore decided to simuate having the discharge process carri
ed out by the examination and treatment team. Because the examination and treatm
ent team knew the patient information, the handoff task woud be eiminated. The
team estimated that this change woud save about five minutes and, therefore, d
ecided to simuate the new system by eiminating discharge as a separate process
. They estimated the probabiity distribution of the combined exam/treatment/dis
charge task by first estimating the probabiity distribution
336
FIGURE 11.7 VVH Emergency Department Initia State Simuation Mode Output
Repications: 100 Tota Time Routine patient
Time Unit: Hours Average 2.4207 HafWidth 0.08 Minimum Average 1.7953 Maximum Av
erage 3.4082 Minimum Vaue 1.2004 Maximum Vaue 5.2448
Queue Time Waiting Time Average Admitting queue 0.00526930 Discharge queue 0.397
2 Exam and treatment queue 0.3382 Nurse history queue 0.01764541 Triage queue 0.
06437939 Other Waiting Average Time Admitting queue 0.03458032 Discharge queue 2
.2481 Exam and treatment queue 2.1930 Nurse history queue 0.1136 Triage queue 0.
5394 HafWidth 0.00 0.26 0.38 0.01 0.05 Minimum Average 0.00267040 0.2888 0.2062
0.01298461 0.1145 Maximum Average 0.1001 5.1713 9.4408 0.4069 1.7216 Minimum Va
ue 0.00 0.00 0.00 0.00 0.00 Maximum Vaue 2.0000 13.0000 22.0000 5.0000 10.0000
HafWidth 0.00 0.26 0.38 0.01 0.05 Minimum Maximum Average Average 0.00048553 0
.01668610 0.06416692 0.8865 0.04167122 1.1956 0.00272715 0.05309733 0.01703829 0
.1402 Minimum Vaue 0.00 0.00 0.00 0.00 0.00 Maximum Vaue 0.2235 2.0531 2.5777
0.3694 0.6506
Resource Usage Instantaneous Utiization Discharge nurse 1 Discharge nurse 2 Exa
m room 1 Exam room 2 Exam room 3 Exam room 4 Financia cerk 1 Financia cerk 2
History nurse 1 History nurse 2 Triage nurse Average 0.8285 0.8360 0.8441 0.832
9 0.8182 0.8075 0.4615 0.4580 0.5294 0.5240 0.6267 HafWidth 0.01 0.01 0.01 0.01
0.02 0.02 0.01 0.01 0.01 0.01 0.01 Minimum Average 0.6715 0.6673 0.6253 0.6548
0.5358 0.6135 0.3320 0.3286 0.3886 0.3937 0.4861 Maximum Average 0.8972 0.9105 0
.9497 0.9297 0.9200 0.9156 0.5636 0.5823 0.6796 0.7107 0.8373 Minimum Vaue 0.00
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Maximum Vaue 1.0000 1.0000 1
.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
for handoff as trianguar (4, 5, 7 minutes). The team used Input Anayzer to sim
uate 1,000 observations of exam/treatment time, discharge time, and handoff tim
e using the previousy determined probabiity distributions for each. For each o
bservation, they added exam/treatment time to discharge time and subtracted hand
off time to find tota time. Input Anayzer found the best fitting probabiity d
istribution for the tota time for the new process as trianguar (18, 50, 82 min
utes). The team simuated the new process and found that patients woud spend an
average of 2.95 hours in the emergency departmentthis change
337
woud actuay increase time spent there. However, it woud aso eiminate the n
eed for discharge rooms. The team decided to investigate the effect of convertin
g the former discharge rooms to exam rooms and ran a new simuation incorporatin
g this change (Figure 11.8). The resut of this simuation is shown in Figure 11
.9: Both the number of patients in the waiting room (examine and treatment queue
) and the amount of time they waited were reduced substantiay. The staffing e
ves were not changed, as the discharge nurses were now treatment nurses. Physic
ian staffing was aso not increased, as some deay inside the treatment process
itsef had aways existed due to the need to wait for ab resuts; this deayed
a fina physician diagnosis. Having more patients avaiabe for treatment fied
this ab deay time for physicians with patient care. The most significant impr
ovement resuting from the process improvement initiative was that tota patient
throughput time now averaged 1.84 hours (110 minutes). This 33 percent reductio
n in throughput time exceeded the teams goa and was ceebrated by VVHs senior ea
dership. The summary of process improvement steps is dispayed in Tabe 11.1. FI
GURE 11.8 VVH Emergency Department (ED) Proposed Change Simuation Mode
Patient arrives
Triage
Compexity
True
Intensive ED care
Fase
Admitting
Patient history Nurse history Waiting room Exam and treatment
Leave ED
338
FIGURE 11.9 VVH Emergency Department Proposed Change Simuation Mode Output
Category Overview 4:11:35 PM Vaues Across A Repications VVH Emergency Repic
ations: 100 Entity Time Tota Time Routine patient Queue Time Waiting Time Admit
ting queue Exam and treatment and discharge queue Nurse history queue Triage que
ue Other Waiting Time Admitting queue Exam and treatment and discharge queue Nur
se history queue Triage queue Average 0.03400433 1.3571 0.1098 0.5629 Haf Width
0.00 0.31 0.01 0.08 Minimum Average 0.00218978 0.00838496 0.01120623 0.1227 Max
imum Average 0.0946 7.8288 0.5716 2.5046 Minimum Vaue 0.00 0.00 0.00 0.00 Maxim
um Vaue 3.0000 19.0000 4.0000 11.0000 Average 0.00519434 0.2039 0.01791752 0.06
635691 Haf Width 0.00 0.04 0.00 0.01 Minimum Maximum Average Average 0.00041085
0.01364095 0.00197293 1.1105 0.00244500 0.07537764 0.01863876 0.2547 Minimum Va
ue 0.00 0.00 0.00 0.00 Maximum Vaue 0.2235 2.2943 0.3417 0.8065 Average 1.8376
Haf Width 0.05 Minimum Average 1.5459 Maximum Average 2.8729 Minimum Vaue 1.0
063 Maximum Vaue 4.5989 Time Unit: Hours February 8, 2007
Resource Usage Instantaneous Utiization Exam room 1 Exam room 2 Exam room 3 Exa
m room 4 Exam room 5 Exam room 6 Financia cerk 1 Financia cerk 2 History nur
se 1 History nurse 2 Triage nurse Average 0.7827 0.7644 0.7626 0.7478 0.7859 0.8
030 0.4606 0.4529 0.5236 0.5154 0.6226 Haf Width 0.02 0.02 0.02 0.02 0.02 0.02
0.01 0.01 0.01 0.01 0.02 Minimum Average 0.5405 0.5468 0.5577 0.4984 0.5420 0.49
90 0.3250 0.2968 0.3642 0.3403 0.4742 Maximum Average 0.9303 0.9103 0.9052 0.899
3 0.9313 0.9472 0.5985 0.6119 0.6766 0.6982 0.8185 Minimum Vaue 0.00 0.00 0.00
0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Maximum Vaue 1.0000 1.0000 1.0000 1.000
0 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
339
Process Improvement Change Baseine, before any improvement Combine admitting fu
nctions Patients enter their own history into computer Combine discharge tasks i
nto examination and treatment process and convert discharge rooms to treatment r
ooms
Throughput Time, Routine Patients 165 minutes 155 minutes 145 minutes 110 minute
s
TABLE 11.1 Summary of VVH Emergency Department Throughput Improvement Project
The Arena software packaged with this book contains a demonstration simuation m
ode named Emergency Room. This mode is both more reaistic and more compicate
d than the one presented here and demonstrates a more powerfu use of the softwa
re, incuding submodes and a high eve of animation. Readers are encouraged to
expore this mode.
Concusion
The theory of swift, even fow (Schmenner 2001, 2004; Schmenner and Swink 1998)
provides a framework for process improvement and increased productivity. The eff
iciency and effectiveness of a process increase as the speed of fow through the
process increases and the variabiity associated with that process decreases. T
he movement of patients in a heathcare faciity is one of the most critica and
visibe processes in heathcare deivery. Reducing fow time and variation in p
rocesses has a number of benefits: Patient satisfaction increases. Quaity of c
inica care improves as patients have reduced waits for diagnosis and treatment.
Financia performance improves. This chapter demonstrates many approaches to th
e chaenges of reducing fow time and process variation. Starting with the stra
ightforward process map, many improvements can be found immediatey by inspectio
n. In other cases, the powerfu too of computer based discrete event simuation
can provide a roadmap to sophisticated process improvements. Ensuring quaity o
f care is another critica focus of heathcare organizations. The process improv
ement toos and approaches in this chapter coud aso be used to reduce process
variation and eiminate errors. Heathcare
340
organizations need to empoy the discipined approach described in this chapter
to achieve the needed improvements in fow and quaity.
1. How do you determine which process improvement toos shoud be used in a give
n situation? What is the cost and return of each approach? 2. Which process impr
ovement too can have the most powerfu effect and why? 3. How can barriers to p
rocess improvement, such as staff reuctance to change, ack of capita, techno
ogica barriers, or cinica practice guideines, be overcome? 4. How can the EM
R be used to make significant process improvements for both efficiency and quai
ty increases?
Chapter Exercises
1. Access the Nationa Guideine Cearinghouse (www.guideine.gov/) and transat
e one of the guideines described into a process map. In addition, add decision
points and aternative paths to dea with unusua issues that might occur in the
process. (Use of Microsoft Visio is encouraged here.) 2. Access the foowing p
rocess maps on the companion web site: Operating Suite Cancer Treatment Cinic U
se basic improvement toos, theory of constraints, Six Sigma, and/or Lean toos
to determine possibe process improvements (Sepveda et a. 1999). 3. Access the
foowing Arena modes on the companion web site: Pharmacy Distribution Cinic B
iing Deveop an improvement pan and use Arena to evauate it. Compare the res
uts from the base mode with your resuts.
References
Cark, J. J. 2005. Unocking Hospita Gridock. Heathcare Financia Management 59
(11): 94104.
341
Deming, W. E. 1998. Deming Web SiteThe Deming Phiosophy. [Onine information; retr
ieved 6/9/06.] deming.ces.cemson.edu/pub/den/deming_ phiosophy.htm. Institute
for Heathcare Improvement. 2006. Managing Patient Fow: Smoothing OR Schedue C
an Ease Capacity Crunch, Researchers Say. [Onine information; retrieved 7/12/06
.] www.ihi.org/IHI/Topics/Fow/PatientFow/ ImprovementStories/Managingpatientf
owSmoothingORscheduecaneasecapacitycrunchesresearcherssay.htm. . 2003. Optimizing
Patient Fow: Moving Patients Smoothy Through Acute Care Settings. Boston: Inst
itute for Heathcare Improvement. Institute of Medicine. 1999. To Err Is Human:
Buiding a Safer Heath System. Washington, D.C.: Nationa Academies Press. Rock
we Automation. 2006. Arena. [Onine information; retrieved 6/6/06.] www.arenasim
uation.com/products/basic_edition.asp. Rodi, S. W., M. V. Grau, and C. M. Orsin
i. 2006. Evauation of a Fast Track Unit: Aignment of Resources and Demand Resu
ts in Improved Satisfaction and Decreased Length of Stay for Emergency Departmen
t Patients. Quaity Management in Heathcare 15 (3): 16370. Schmenner, R. W. 2004.
Service Businesses and Productivity. Decision Sciences 35 (3): 33347. . 2001. Looking
Ahead by Looking Back: Swift, Even Fow in the History of Manufacturing. Producti
on and Operations Management 10 (1): 8796. Schmenner, R. W., and M. L. Swink. 199
8. On Theory in Operations Management. Journa of Operations Management 17 (1): 971
13. Schmidt, C., and M. Messer. 2005. A Pragmatic Approach to Improving Patient
Efficiency Throughput. [Onine information; retrieved 6/15/06.] www.ihi.org/ IHI
/Topics/Fow/PatientFow/ImprovementStories/APragmaticApproach toImprovingPatien
tEfficiencyThroughput.htm. Sepveda, J. A., W. J. Thompson, F. F. Baeser, M. I.
Avarez, and L. E. Cahoon III. 1999. The Use of Simuation for Process Improvemen
t in a Cancer Treatment Center. In Proceedings of the 31st Conference on Winter S
imuation: SimuationA Bridge to the Future, Vo. 2. New York: ACM Press. Waey,
P., K. Sivester, and R. Steyn. 2006. Managing Variation in Demand: Lessons from
the UK Nationa Heath Service. Journa of Heathcare Management 51 (5): 30922.
Recommended Reading
Anupindi, R., S. D. Deshmukh, and S. Chopra. 1999. Managing Business Process Fo
ws, 1st ed. Upper Sadde River, NJ: Prentice Ha.
CHAPTER
12
SCHEDULING AND CAPACITY MANAGEMENT
CHAPTER OUTLINE
Operations Management in Action Overview Staff Scheduing Riverview Cinic Urgen
t Care Staffing Using Linear Programming Job/Operation Scheduing and Sequencing
Rues Vincent Vaey Hospita and Heath System Laboratory Sequencing Rues Pat
ient Appointment Scheduing Modes Riverview Cinic Appointment Schedue Advance
d Access Patient Scheduing Advanced Access for an Operating and Market Advantag
e Impementing Advanced Access Obtain buy in Predict demand Predict capacity Ass
ess operations Work down the backog Going ive Metrics Advanced Access at Vince
nt Vaey East Arena advanced access simuation mode Resuts from base simuati
on Improved and optimized mode Concusion Discussion Questions Chapter Exercise
s References
342

advanced access appointment scheduing schemes appointment scheduing systems, s
ee appointment scheduing schemes backog, bad and good Baiey Wech rue bock
appointment schedue capacity contingency pan critica ratio (CR) eariest due
date (EDD) first come, first served (FCFS) individua appointment schedue kanba
n system inear programming, see mathematica programming Littes aw mathematica
programming mixed bockindividua appointment schedue Nationa Association of
Pubic Hospitas (NAPH) optima outpatient appointment scheduing too patient d
emand PCP coverage PCP match primary care physician (PCP) schedue function (Are
na) sequencing rues set function (Arena) shortest processing time (SPT) sack t
ime remaining (STR) staff scheduing tardiness utiization wait time for next ap
pointment
343

In 2001, the Nationa Association of Pubic Hospitas (NAPH) sponsored the pubi
cation of Advanced Access: A New Paradigm in the Deivery of Ambuatory Care Ser
vices (Singer 2001). When that initia report was written, the concept of advanc
ed access was new to NAPH members. However, the monograph generated strong inter
est among members, and use of the mode by safety net hospitas and heath syste
ms has become much more widespread. Speciaty services in which NAPH members hav
e begun using advanced access incude behaviora heath, dermatoogy, neuroogy,
ophthamoogy, orthopedics, and radioogy. The top two reasons NAPH members hav
e turned to advanced access appear to be a desire to improve patient satisfactio
n by reducing wait times for appointments and to reduce appointment no show rate
s. Given the extraordinary voume of ambuatory visits handed by NAPH cinics,
there is a genera desire among ambuatory care eaders to remove inefficiencies
and drive out waste. With no show rates as high as 50 percent, heath systems a
re eager to find ways to make demand more predictabe and provider workoads ste
adier whie increasing productivity. Other reasons for adopting an advanced acce
ss mode incude the desire to improve provider and staff satisfaction and conti
nuity of care. The resuts of impementing these advanced access scheduing syst
ems are remarkabe (Tabes 12.1 and 12.2) considering the high voume of patient
s and their compex medica conditions and socia situations.
Overview
Matching the suppy of goods or services to the demand for those goods or servic
es is a basic operationa probem. In a manufacturing environment, inventory can
be used to respond to fuctuations in demand. In the heathcare environment, sa
fety stock can be used to respond to fuctuations in demand for suppies (see Ch
apter 13), but it is not possibe to inventory heathcare services. Capacity mus
t, therefore, be matched to demand. If capacity is greater than demand, resource
s are underutiized and costs are high. Ide staff, equipment, or faciities inc
rease organizationa costs without increasing revenues. If capacity is ower tha
n demand, patients incur ong waits or find another provider. To match capacity
to demand, organizations can use demand infuencing strategies or capacity manag
ement strategies. Pricing and promotions are often used to infuence demand and
demand timing; however, this is often not 344
Chapter 12: Scheduing and Capacity Management

345
Organization Arrowhead Regiona Medica Center (Coton, CA) Boston Medica Cente
r (Boston) Cambridge Heath Aiance (Cambridge, MA) Contra Costa Regiona Medic
a Center (Martinez, CA) Denver Heath (Denver) NYCHHCBeevue Hospita Center (N
ew York) NYCHHCWoodhu Medica & Menta Heath Center (Brookyn, NY) Parkand He
ath & Hospita System (Daas) Riverside County Medica Center (Riverside, CA)
Santa Cara Vaey Heath & Hospita System (San Jose, CA) University of Coorad
o Hospita (Denver)
Preimpementation Postimpementation 28 2530 30 30 2535 39 50 40 40 3050 810 8 1215 2
0 15 717 32 25 20 18 14 810
TABLE 12.1 No Show Rates (%) Before and After Impementation of Advanced Access
Scheduing for Seected NAPH Members as of 2003
SOURCE: Singer and Regenstein (2003). Reprinted with permission from the Nationa
Association of Pubic Hospitas and Heath Systems.
Organization Arrowhead Regiona Medica Center (Coton, CA) Boston Medica Cente
r (Boston) Cambridge Heath Aiance (Cambridge, MA) Contra Costa Regiona Medic
a Center (Martinez, CA) Denver Heath (Denver) NYCHHCBeevue Hospita Center (N
ew York) NYCHHCWoodhu Medica & Menta Heath Center (Brookyn, NY) Parkand He
ath & Hospita System (Daas) Riverside County Medica Center (Riverside, CA)
Santa Cara Vaey Heath & Hospita System (San Jose, CA) University of Coorad
o Hospita (Denver)
Wait Time 0 02 2.5 01 01 3 12 714 07 10 34
TABLE 12.2 Wait Time (Days) for Advanced Access Appointments for Seected NAPH M
embers as of 2003
SOURCE: Singer and Regenstein (2003). Reprinted with permission from the Nationa
Association of Pubic Hospitas and Heath Systems.
346
a viabe strategy for heathcare organizations. In the past, many heathcare org
anizations used the demand eveing strategy of appointment scheduing; more rec
enty, many have begun moving to advanced access scheduing. Capacity management
strategies aow the organization to adjust capacity to meet fuctuating demand
and incude using part time empoyees, on ca empoyees, cross training, and o
vertime. Effective and efficient scheduing of patients, staff, equipment, faci
ities, or jobs can hep to match capacity to demand and ensure that scarce heat
hcare resources are utiized to their fuest extent. This chapter outines issu
es and probems faced in scheduing and discusses toos and techniques that can
be empoyed in scheduing patients, staff, equipment, faciities, or jobs. Topic
s covered incude: Staff scheduing; Job/operation scheduing and sequencing ru
es; Patient appointment scheduing modes; and Advanced access patient scheduin
g.
The scheduing of patients is a unique, but important, subprobem of patient fo
w. Since the mid twentieth century, much patient care deivery has moved from th
e inpatient setting to the ambuatory cinic. Because this trend is ikey to co
ntinue, matching cinic capacity to patient demand becomes a critica operating
ski. In addition, if this capacity can be depoyed to meet a patients desired s
chedue, marketpace advantage can be gained. Therefore, this chapter focuses on
advanced access (same day scheduing) for ambuatory patients. Topics covered i
ncude: Advantages of advanced access; Impementation steps; Metrics to measure
the operations of advanced access scheduing systems; and Arena simuation of an
advanced access cinic with aternative staffing and scheduing rues. Many of
the operations toos and approaches detaied in earier chapters are empoyed to
optimize the operations of an advanced access cinic.
Staff Scheduing
For fairy sma schedue optimization probems, where demand is reasonaby know
n and staffing requirements can therefore be estimated with certainty, mathemati
ca programming (Chapter 6) can be used to optimize staffing eves and schedue
s. A simpe exampe of this type of probem is iustrated beow. A more detaie
d probem is iustrated in Using Linear Programming to Minimize the Cost of Nurs
ing Personne

347
(Matthews 2005), and a summary of various optimization modes can be found in Nur
se Rostering ProbemsA Bibiographic Survey (Cheang et a. 2003). As the probems
become arger and more compex, deveoping and using a mathematica programming
mode becomes time and cost prohibitive. In those cases, simuation can be used
to answer what if scheduing questions: What if we added another nurse? or What if we
cross trained empoyees? See Chapter 11 and the advanced access section of this
chapter for exampes of these types of appications.
Riverview Cinic Urgent Care Staffing Using Linear Programming
Nurses who staff the Riverview Cinic after hours urgent care cinic have been c
ompaining about their schedues. They woud ike to work five consecutive days
and have two consecutive days off every seven days. Additionay, different nurs
es prefer different days off and beieve that their preferences shoud be assign
ed based on seniority; the most senior nurses shoud get their most desired days
off. Riverview Urgent Care Cinic (UCC) has coected patient demand data by da
y of the week and knows how many nurses shoud be on staff each day to service d
emand. Riverview UCC management wants to minimize nurse payro whie reducing t
he nurses compaints about their schedues. They decide to use inear programming
to hep determine a soution for this probem. Target staffing eves and saar
y expenses are shown in Tabe 12.3. First, Riverview needs to determine how many
nurses shoud be assigned to each possibe schedue. There are seven possibe s
chedues (Monday and Tuesday off, Tuesday and Wednesday off, etc.). The goa is
to minimize weeky saary expense, and the objective function is: Minimize: ($32
0 Sun. # of nurses) + ($240 Mon. # of nurses) + ($240 Tues. # of nurses) + ($240
Wed. # of nurses) + ($240 Thurs. # of nurses) + ($240 Fri. # of nurses) + ($320
Sat. # of nurses)
Sunday Monday Tuesday Wednesday Thursday Friday Saturday
Nurses needed/day Saary and benefits/ nurse day
5 320
4 240
3 240
3 240
3 240
4 240
6 320
TABLE 12.3 Riverview UCC Target Staffing Leve and Saary Expense
348
The constraints are: The number of nurses schedued each day must be greater tha
n the number of nurses needed each day. Sunday # of nurses 5 (1) Monday # of nur
ses 4 (2) Tuesday # of nurses 3 (3) Wednesday # of nurses 3 (4) Thursday # of nu
rses 3 (5) Friday # of nurses 4 (6) Saturday # of nurses 6 (7) The number of nur
ses assigned to each schedue must be greater than zero and an integer. # A (B,
C, D, E, F, or G) nurses 0 (814) # A (B, C, D, E, F, or G) nurses = integer (1521)
Figure 12.1 shows the Exce Sover setup of this probem. Sover (Figure 12.2)
finds that the Riverview UCC needs six nurses and shoud assign one nurse to sch
edues A, B, C, and D; two nurses to schedue E; and no nurses to schedues F an
d G. The tota saary expense with this optima schedue is: Minimize: ($320 5)
+ ($240 4) + ($240 4) + ($240 4) + ($240 3) + ($240 4) + ($320 6) = $8,080/Week
Next, Riverview needs to determine which nurses to assign to which schedue base
d on their preferences and seniority. Each nurse is asked to rank FIGURE 12.1 In
itia Exce Sover Setup of Riverview UCC Optimization

349
FIGURE 12.2 Riverview UCC Initia Sover Soution and Schedue Preference Setup
schedues A through E in order of preference for that particuar schedue. The n
urses preferences on a scae of one to five, with five being the most preferred s
chedue, are then weighted by a seniority factor. Riverview decides to use a sen
iority weighting factor based on the number of years a particuar nurse has work
ed at the UCC compared with the number of years the most senior nurse has worked
at the UCC. The goa is to maximize the nurses tota weighted preference scores
(WPS), and the objective function is: Maximize: Marys WPS + Annes WPS + Susans WPS
+ Toms WPS + Cathys WPS + Janes WPS The constraints are: Each nurse must either be
assigned to a particuar schedue or not assigned to that schedue. Mary assigne
d to schedue A (B, C, D, or E) = 0 or 1 (binary) (15) Anne assigned to schedue
A (B, C, D, or E) = 0 or 1 (binary) (610) Susan assigned to schedue A (B, C, D,
or E) = 0 or 1 (binary) (1115) Tom assigned to schedue A (B, C, D, or E) = 0 or
1 (binary) (1620) Cathy assigned to schedue A (B, C, D, or E) = 0 or 1 (binary)
(2125) Jane assigned to schedue A (B, C, D, or E) = 0 or 1 (binary) (2630) The nu
mber of nurses assigned to each schedue must be as required. # of nurses assign
ed to schedue A (B, C, or D) = 1 (3134) # of nurses assigned to schedue E = 2 (
35) Each nurse can ony be assigned to one schedue. Mary (Anne, Susan, Tom, Cat
hy, or Jane) A + B + D + C + E = 1 (3641) Figure 12.2 Shows the Exce setup of th
is probem.
350
FIGURE 12.3 Riverview UCC Fina Sover Soution for Individua Schedues
Sover (Figure 12.3) finds that Mary shoud be assigned to schedue D (her secon
d choice), Anne to schedue E (her first choice), Susan to schedue C (her first
choice), Tom to schedue E (his first choice), Cathy to schedue B (her second
choice), and Jane to schedue A (her first choice). A of the nurses now have 2
consecutive days off every 7 days and are assigned to either their first or sec
ond choice of schedue. Note that even this simpe probem has 20 decision varia
bes and 41 constraints.
Job/Operation Scheduing and Sequencing Rues
Job/operation scheduing ooks at the probem of how to sequence a poo of jobs
(or patients) through a particuar operation. For exampe, a cinic aboratory h
as many patient bood sampes that need to be testedin what order shoud the ab
work on the sampes? Or, a hospita has many surgeries waiting to be performedin
what order shoud those surgeries occur? The simpest sequencing probems consis
t of a poo of jobs waiting for ony one resource. Sequencing of those jobs is u
suay based on a desire to meet due dates (time at which the job is supposed to
be compete) by minimizing the number of jobs that are ate, minimizing the ave
rage amount of time by which jobs wi be ate, or minimizing the maximum ate t
ime of any job. It may aso be desirabe to minimize the time jobs spend in the
system or average competion time. Various sequencing rues can be used to sched
ue jobs through the system. Commony used rues incude: First come, first serv
ed (FCFS): Jobs are sequenced in the same order in which they arrive. Shortest p
rocessing time (SPT): The job that takes the east amount of time to compete is
first, foowed by the job that wi take the next east amount time, and so on
.

351
Eariest due date (EDD): The job with the eariest due date is first, foowed b
y the job with the next eariest due date, and so on. Sack time remaining (STR)
: The job with the east amount of sack (time unti due date processing time) i
s first, foowed by the job with the next east amount of sack time, and so on
. Critica ratio (CR): The job with the smaest critica ratio (time unti due
date/processing time) is first, foowed by the job with the next smaest criti
ca ratio, and so on. When there is ony one resource or operation for the jobs
to be processed through, the SPT rue wi minimize average competion time, and
the EDD rue wi minimize average ateness and maximum ateness; however, no s
inge rue wi accompish both objectives. When jobs (or patients) must be proc
essed through a series of resources or operations with different possibe sequen
cing at each, the situation becomes more compex, and a particuar rue wi not
resut in the same outcome for the entire system as for the singe resource. Of
ten, simuation is used to evauate these more compex systems and hep determin
e optimum sequencing. For a busy resource, the SPT rue is often used. This rue
wi compete a greater number of jobs in a shorter amount of time, but some jo
bs with ong competion times may never be finished. To aeviate this probem,
the SPT rue is often used in combination with other rues. For exampe, in some
emergency departments, ess severe cases (those with a shorter processing time)
are separated and fast tracked to free up examination rooms quicky. For time s
ensitive operations, where ateness is not toerated, the EDD rue is appropriat
e. Because it is the easiest to appy, the FCFS rue is typicay used when the
resource has excess capacity and no jobs wi be ate. In a Lean environment, se
quencing rues become irreevant because the idea size of the poo of jobs is r
educed to one and a kanban system (a form of FCFS) can be used to pu jobs thro
ugh the system (Chapter 9).
Vincent Vaey Hospita and Heath System Laboratory Sequencing Rues
The aboratory at Vincent Vaey Hospita and Heath System (VVH) recenty ost
one of its ab technicians, and the ab manager, Jessica Simmons, does not beie
ve she wi be abe to find a quaified repacement for at east one month. This
has greaty increased the workoad in the ab, and physicians have been compai
ning that their requested bood work is not being competed in a timey manner.
In the past, Jessica has divided the bood testing among the technicians and had
them compete the tests on an FCFS basis. She is considering using a different
sequencing rue to try to satisfy more of the physicians. In anticipation of thi
s change, she requested that each physician put a desired competion time on eac
h request for bood testing. To investigate the effects of changing the sequenci
ng rues, she decides to take the first five requests for boodwork from
352
one of the technicians and anayze job competion under various scheduing rues
. Note that, for five jobs, there are 120 possibe sequences in which the jobs c
oud be competed. Tabe 12.4 shows the time to compete bood work on each samp
e and the time when the physician has requested that it be compete. Tabe 12.5
shows the order in which jobs wi be processed and resuts under different seq
uencing rues. Tabe 12.6 compares the various sequencing rues. The FCFS rue p
erforms poory on a measures. The SPT rue minimizes average competion time,
and the EDD rue minimizes average tardiness. Under both rues, three jobs are t
ardy and the maximum tardiness is 150 minutes. After ooking at these resuts, J
essica decides to impement the EDD rue for aboratory bood tests to minimize
the number of tardy jobs and the average tardiness of jobs. She hopes this wi
reduce physician compaints unti a new technician can be hired.
Patient Appointment Scheduing Modes
Appointment scheduing modes attempt to minimize patient waiting time whie max
imizing utiization of the resource (cinician, machine, etc.) the patients are
waiting to access. Soriano (1966) cassifies appointment scheduing systems into
four basic types: bock appointment, individua appointment, mixed bockindividu
a appointment, and other. A bock appointment scheme schedues the arriva of a
patients at the start of a cinic session. Patients are usuay seen FCFS, bu
t other sequencing rues can be used. This type of scheduing system maximizes u
tiization of the cinician, but patients can experience ong wait times. An ind
ividua appointment scheme assigns different, equay spaced appointment times t
o each individua patient. In a common modification of this type of system, diff
erent appointment engths are avaiabe and assigned based on the type of patien
t. This system reduces patient waiting time but decreases utiization of the ci
nician. Increasing the interva between arrivas resuts in a reduction of both
waiting time and utiization. A mixed bockindividua appointment scheme schedue
s a group of
TABLE 12.4 VVH Laboratory Bood Test Information
Sampe A B C D E
Processing Time Due Time (Minutes) (Minutes from Now) 50 100 20 80 60 100 160 50
120 80
Sack
CR
100 50 = 50 100/50 = 2.00 160 100 = 60 160/100 = 1.60 50 20 = 30 50/20 = 2.50 12
0 80 = 40 120/80 = 1.50 80 60 = 20 80/60 = 1.33

353
Sequence FCFS A B C D E
Start Time
Processing Time
Competion Time
Due Time
Tardiness
0 50 150 170 250
50 100 20 80 60 Average
50 150 170 250 310 186
100 160 50 120 80
TABLE 12.5 VVH Laboratory Bood Test Sequencing Rues
170 50 = 120 250 120 = 130 310 80 = 230 (120 + 130 + 230)/5 = 96
SPT C A E D B
0 20 70 130 210
20 50 60 80 100 Average
20 70 130 210 310 148
50 100 80 120 160
130 80 = 50 210 120 = 90 310 160 = 150 (50 + 90 + 150)/5 = 58
EDD C E A D B
0 20 80 130 210
20 60 50 80 100 Average
20 80 130 210 310 150
50 80 100 120 160
130 100 = 30 210 120 = 90 310 160 = 150 (30 + 90 + 150)/5 = 54
STR E C D A B
0 60 80 160 210
60 20 80 50 100 Average
60 80 160 210 310 164
80 50 120 100 160

80 50 = 30 160 120 = 40 210 100 = 110 310 160 = 150 (30 + 40 + 110 + 150)/5 = 66
CR E D B A C
0 60 140 240 290
60 80 100 50 20 Average
60 140 240 290 310 208
80 120 160 100 50
140 120 = 20 240 160 = 80 290 100 = 190 310 50 = 260 (20 + 80 + 190 + 260)/5 = 1
10
A times in tabe are in minutes.
354
TABLE 12.6 Comparison of VVH Bood Test Sequencing Rues
Sequencing Rue FCFS SPT EDD STR CR
Average Competion Time 186 148* 150 164 208
Average Tardiness 96 58 54* 66 110
Number of Tardy Jobs 3* 3* 3* 4 4
Maximum Tardiness 230 150* 150* 150* 260
*Best vaues. Times are in minutes.
patients to arrive at the start of the cinic session, foowed by equay space
d appointment times for the remainder of the session. This type of system can be
used to baance the competing goas of increased utiization and decreased wait
ing time. Other appointment schemes are modifications of the first three types.
Simuation has been used to study the performance of various appointment schedu
ing modes and rues. See Cayiri and Vera (2003) for a review of these studies
. More recenty, Hutzschenreuter (2004) and Kaandorp and Kooe (2007a) evauated
some of these rues, taking into account reaistic environmenta factors such a
s no shows and variabe service time. Athough no scheduing rue or scheme was
found to be universay superior, the BaieyWech rue (Baiey and Wech 1952) p
erformed we under most conditions. This rue schedues two patients at the beg
inning of a cinic session, foowed by equay spaced appointment times for the
remainder of the session. Kaandorp and Kooe (2007b) deveoped a mathematic mod
e to determine an optima schedue using a weighted average of expected waiting
times of patients, ide time of the cinician, and tardiness (the probabiity t
hat the cinician has to work ater than schedued mutipied by the average amo
unt of added time) as the objective. This too uses simuation to compare a user
defined schedue to the optima schedue found using the mode and is avaiabe
onine at http://obp.math.vu.n/heathcare/software/ges/.
Riverview Cinic Appointment Schedue
Physicians at VVHs Riverview Cinic typicay see patients for six consecutive ho
urs each day. Each appointment takes an average of 20 minutes; therefore, each c
inician is schedued to see 18 patients/day. Two percent of patients are no sho
ws. Currenty, Riverview is using an individua appointment scheme with appointm
ents schedued every 20 minutes. However, cinicians have been compaining that
they often have to work ate but are ide at various points during the course of
the day. Riverview decides to use the Optima Outpatient Appointment Scheduing
Too (Kaandorp and Kooe 2007b) to see if a better scheduing mode coud be us
ed to aeviate these compaints without increasing patient waiting time to an u
nacceptabe eve.

355
Figure 12.4 shows the resuts of this anaysis when waiting time weight is 1.5,
ide time weight is 0.2, and tardiness weight is 1.0. The optima schedue foo
ws the Baiey Wech rue. Under this rue, patient waiting is increased by 5 min
utes, but both ideness and tardiness are decreased. Riverview does not beieve
that the additiona waiting time FIGURE 12.4 Riverview Cinic Appointment Schedu
ing
NOTE: Copyright 2007 Guido Kaandorp and Ger Kooe.
356
is unacceptabe and decides to impement this new appointment scheduing scheme.
Advanced Access Patient Scheduing
Advanced Access for an Operating and Market Advantage
In the eary 1990s, Mark Murray, M.D., M.P.A., and Catherine Tantau, R.N., deve
oped the concept of advanced access at Kaiser Permanente in Northern Caifornia.
Their goa was to eiminate ong patient waits for appointments and bottenecks
in cinic operations (Singer 2001). The principes they deveoped and refined h
ave now been impemented by many eading heathcare organizations gobay. Beca
use most cinics today use traditiona scheduing systems, ong wait times are p
revaent, and appointments are ony avaiabe weeks, or even months, in advance.
The further in advance visits are schedued, the greater the fai (noshow) rate
becomes. To compensate, providers doube or even tripe book appointment sots
. Long deays and queues occur when a patients schedued actuay appear for t
he same appointment sot. This probem is compounded by patients who have urgent
needs requiring that they be seen immediatey. These patients are either worked
into the schedue or sent to an emergency department, decreasing both continuit
y of care and revenue to the cinic. At the emergency department, patients are f
requenty tod to see their primary care physicians (PCPs) in one to three days,
further compicating the probem. Advanced access is impemented by beginning e
ach day with a arge portion of each providers schedue open for urgent, routine,
and foow up appointments. Patients are seen when they want to be seen. This d
ramaticay reduces the fai rate, as patients do not have to remember cinic vi
sits they booked ong ago. Because there is no doube or tripe booking, patien
ts are seen on time and schedues run smoothy inside the cinic. Cinics using
advanced access can provide patients with the convenience of wak in or urgent c
are, with the added advantage of maintaining continuity of care with their own d
octors and cinics. Parente, Pinto, and Barber (2005) studied the impementation
of advanced access in a arge Midwestern cinic with a patient pane of 10,000.
After impementation of this system, the average number of days between caing
for an appointment and actuay being seen by a doctor decreased from 18.7 to 1
1.8. However, the most significant finding was that 91.4 percent of patients now
saw their own PCP, as opposed to 69.8 percent before the impementation of adva
nced access.
Impementing Advanced Access
Impementing advanced access is difficut. Changing from a ong standing system,
abeit a fawed one, is chaenging. However, foowing a few we

357
prescribed steps increases the probabiity of success. In a study of arge urban
pubic hospitas, Singer (2001) deveoped the foowing methodoogy to impemen
t advanced access.
Obtain buy in Leadership is key to making this major change, and the advanced ac
cess system must be supported by senior eaders and providers themseves. It may
be hepfu to tour other cinics that have impemented advanced access to get a
sense of how this system can work successfuy. For arge systems, it is best t
o start sma in one or two cinica settings. Once initia operating probems a
re resoved and cinic staff is positive about the change, advanced access can b
e carefuy impemented in a number of additiona cinics in the system. Predict
demand The first quantitative step in impementation is to measure and predict
demand from patients. For each day during a study period, demand is cacuated a
s the voume of patients requesting appointments (today or in the future), wak
in patients, patients referred from urgent care cinics or emergency departments
, and number of cas defected to other providers. After initia demand cacua
tions are performed, they can be made more sophisticated by considering day of t
he week, seasonaity, demand for same day versus schedued appointments, and eve
n cinica characteristics of patients. Predict capacity The capacity of the ci
nic needs to be determined once demand is cacuated. In genera, this is the su
m of appointment sots avaiabe each day. Capacity can vary dramaticay from d
ay to day, as providers usuay have obigations for their time other than seein
g patients in the cinic. Determining whether a cinics capacity can meet expecte
d demand is reativey easy using Littes aw (described in detai in Chapter 11)
. Singer (2001) reported that many pubic hospita cinics initiay fet that d
emand exceeded capacity in their operations. However, many of these cinics foun
d hidden capacity in their systems by more effectivey using providers (e.g., mi
nimizing paperwork) and using for exams space that had been used for storage. An
other opportunity to improve the capacity of a cinic is to standardize and mini
mize the ength of visit times. A cinic with high variabiity in appointment ti
mes wi find that it has many sma bocks of unused time (Sanderson 2007). Ass
ess operations The impementation of advanced access provides the opportunity to
review and improve the core patient fow and operations in a cinic. The toos
and techniques of process mapping and process improvement, particuary vaue
358
stream mapping and the theory of constraints, shoud be appied before advanced
access is impemented.
Work down the backog Working down the backog is one of the most chaenging ta
sks in impementing advanced access, as providers must see more patients per day
unti they have caught up to same day access. For exampe, each provider coud
work one extra hour per day and see three additiona patients unti the backog
is eiminated. The number of days to work off a backog can be cacuated by:
Days to work down backog = Current backog/Increase in capacity where Current b
ackog = Appointments on the books/ Average number of patients seen per day and
Increase in capacity = (New service rate [Patients/Day]/ Od service rate [Patie
nts/Day]) 1
Going ive Once a cinic is ready to go ive, it must first determine how many a
ppointment sots to reserve for same day access. Singer and Regenstein (2003) re
port that pubic hospita cinics are eaving 40 percent to 60 percent of their
sots avaiabe for same day access, but other cinics have reported eaving up
to 75 percent of sots avaiabe. It is important to educate patients, as they w
i be surprised by the abiity to see a provider the day they request an appoin
tment. Many edery patients may not choose this option, as they may need more t
ime for preparation or to arrange transportation. No cinic operates in a compe
tey stabe environment, so it is usefu to prospectivey deveop contingency p
ans in case a provider is i or caed away on an emergency. Contingencies can
aso be predictabe increases in demand such as routine physicas in the weeks p
receding the start of schoo. Good contingency panning wi ensure the smooth a
nd efficient operation of an advanced access system.
Metrics
Gupta et a. (2006) deveoped a set of key indicators that can be used to evaua
te the performance of advanced access scheduing systems:

359

PCP match: percentage of same day patients who see their own PCP PCP coverage: p
ercentage of same day patients seen by any physician Wait time for next appointm
ent (or third next avaiabe appointment) Good backog: appointments schedued i
n advance because of patient preference Bad backog: appointments waiting becaus
e of ack of sots Most we functioning advanced access systems have high PCP m
atch and PCP coverage. Depending on patient mix and preferences the good backog
can be reativey arge, but a arge or growing bad backog can be a signa tha
t capacity or operating systems in the cinic need to be improved.
Advanced Access at Vincent Vaey East
VVH operates a sma primary care cinicVVH Easton the citys east side. The cinic
has four famiy physicians. Dr. Smith is the senior provider, and he has a high
number of edery patients. Because he is aso VVHs medica director for quaity,
he works a imited schedue each week and is not accepting any new patients. Dr
s. Anderson and Biings have both worked at the cinic for a number of years an
d have reativey fu patient panes. Dr. Kumarin was recenty added to the ci
nic to accommodate growth. The staff at VVH East has decided to impement advanc
ed access as a mode for the rest of the VVH primary care system. However, befor
e impementation of this new scheduing system, VVH East decided to deveop a si
muation mode using Arena to determine the optima mix of providers and schedu
ing rues. Dr. Smith ed this simuation work, as he had been impressed with res
uts achieved in the VVH emergency department. He was concerned that advanced ac
cess might hinder his ong standing reationships with many of his edery patie
nts.
Arena advanced access simuation mode Before constructing the mode, the advanc
ed access team identified the goas of the simuation.
In consideration of Dr. Smiths ong standing work at the cinic, the simuation w
oud focus on how we the new advanced access mode aowed him to maintain pri
mary care continuity (PCP match) with his patients and ensure that most coud be
seen on the same day as requested. The simuation resuts woud hep to determi
ne the best poicies for baancing work among the four physicians and ensure tha
t as many patients as possibe coud be seen during the providers work schedues.
The unit of simuation was one week, and a appointments were assumed to be re
quested as same day. Athough approximatey 25 percent of
360
appointments woud ikey be schedued in advance (good backog), assuming 100 p
ercent same day requests was a way to test the fexibiity and performance of th
e system. This simuation is an approximate, but not exact, mode of the operati
ons of the cinic. If two patients appear at the cinic simutaneousy and want
to see the same doctor, this mode assumes that one is seen and the other is put
into a queue. In actuaity, because most same day scheduing is done on the pho
ne, the cinic scheduing staff woud ask the patient if a sighty different ti
me woud work; most patients woud wiingy accept this sight change. Therefor
e, the queue statistics in this mode are not stricty representative of the det
aied cinic operations but give a good fee for how much difficuty the cinic
scheduing staff wi have in finding a sot for patients on the day they reques
t. Modeing every sot in a same day schedue is possibe but requires a much mo
re compex mode. The interested reader is directed to Chapter 5 of Simuation w
ith Arena (Keton, Sadowski, and Sturrock 2004). The eve of simuation present
ed here tracks actua operations cosey enough to aow the VVH East team to mo
de different staffing and coverage scenarios to make reasonabe decisions. The
mode contains five patient fowsone for each doctors existing patients and one fo
r new patients. The physicians have determined a backup schedue in case they ar
e busy. Most patients are wiing to be treated by the backup doctor. The Arena
mode is shown in Figure 12.5. The construction and operation of this mode are
avaiabe on the companion web site.
FIGURE 12.5 VVH East Base Simuation of Advanced Access
Visit ength Anderson
Anderson patients treated
Tota Anderson patients
Visit ength Biings
Biings patients treated
Tota Biings patients Dispose 1
Visit ength Kumarin
Kumarin patients treated
Tota Kumarin patients
Visit ength New
New patients treated
Tota New patients
NOTE: Mode created in Arena, Rockwe Automation, Inc.

361
Arena aows the user to provide mutipe resources for each process using the se
t function. The initia backup rues for the doctors are dispayed in Tabe 12.7.
In addition, Arena provides the capabiity to specify when a resource is avaia
be with the schedue function. The modeing team decided to mode one week, with
each day containing eight one hour sots (9 a.m. to noon and 1 p.m. to 6 p.m.).
The initia physician schedues are shown in Tabe 12.8.
Primary Backup Smith Biings Anderson Kumarin Biings Secondary Backup Anderso
n Kumarin Biings Kumarin Tertiary Backup Biings
Patients Primary Physician Smith Biings Anderson Kumarin New patient
TABLE 12.7 Initia Physician Backup Rues
Physician Anderson Biings Kumarin Smith Physician Anderson Biings Kumarin Sm
ith
Monday Hours 1 2 3 4
5
6 1
7 1
8 1
Tuesday Hours 1 2 3 4 5 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
6 1 1 1 6 1 1 1
7
8
TABLE 12.8 Initia Physician Schedue
1 1
1 1
1 1
1 1
1 1
1 1 6 1 1 1
1 7 1
1 8
1 1 7 1 1
1 1 8 1 1
Wednesday Hours 1 2 3 4 5 1 1 1 1 1 1 1 1 1 1 1
Thursday Hours 1 2 3 4 5 1 1 1 1 1 1 1 1 1 1 1
1
1
Friday Hours Physician Anderson Biings Kumarin Smith A 1 1 1 2 1 1 3 1 1 1 4
1 1 1 5 6 7 8 Tota Weeky Hours 20 21 32 15 88
1
1
1
1
362
Data on arriva rate and service time were coected by physician and patient ty
pe. Probabiity distributions for each doctors patient arriva rate and service t
ime were determined using Arenas Input Anayzer (Tabe 12.9). Dr. Smiths patients
had onger treatment times, as they were predominanty edery. New patients as
o had onger treatment times, as initia cinica histories needed to be gathere
d. Because the VVH East team wanted to track specificay what happened to Dr. S
miths patients during the simuation runs, a submode for the fow of his patient
s was created. This submode, which routed patients to backup physicians based o
n avaiabiity (Figure 12.6), simuates the set funcTABLE 12.9 Initia Arriva R
ate and Service Time Distributions
Arriva Rate Exponentia Distribution Mean (Hours) 1.0 1.2 1.0 2.0 0.8 Triangua
r Distribution of Service Time: Minimum, Mode, Maximum (Hours) 0.15, 0.50, 0.80
0.10, 0.20, 0.60 0.10, 0.20, 0.60 0.15, 0.50, 0.80 0.25, 0.50, 1.00
Physician Smith Biings Anderson Kumarin New patient
FIGURE 12.6 VVH East Advanced Access Submode for Dr. Smiths Patients
Smith avaiabe? Ese
STATE(Smith)= 1
Smith sees Smith patient
Smith patient seen by Smith
Anderson avaiabe? Ese
STATE(Anderson)= 1
Anderson sees Smith patient
Smith patient seen by Anderson
Biings avaiabe? Ese
STATE(Biings)= 1
Biings sees Smith patient
Smith patient seen by Biings
Smith Anderson Biings Kumarin
NOTE: Mode created in Arena, Rockwe Automation, Inc.

363
tion in Arena but aso records the needed data. First running this submode at s
ow speed, a usefu feature of the software, aowed the team to observe that pa
tient fow and physician coverage were operating as predicted. The submode is d
esigned to check in turn whether each doctor is busy and, if so, move the patien
t to the next doctor in the backup schedue. If a physicians are busy, the pat
ient is returned to Dr. Smiths queue. This aows Dr. Smith to see as many of his
own patients as possibe. Another advantage of the Arena animation feature is t
hat the entire mode can be run incrementay, one step at time, with a focus on
queues that may occur. The timing of arge queues can then be matched to provid
er staffing schedues to suggest modifications and improvements.
Resuts from base simuation The simuation was repicated 100 times with these
parameters; resuts are shown in Figure 12.7. The modeing team made a number of
important observations regarding this base case. First, Dr. Smith was seeing an
average of 33.4 patients each week (Figure 12.7). Because he was ony working 1
5 hours per week and his patients averaged 0.5 hour per visit, he coud not keep
up; some of these patients were being moved to future weeks for appointments. T
his was aso indicated by Dr. Smiths ong average queue time of 3.5 hours. Athou
gh the scheduing staff probaby coud work hard to minimize this effect, Dr. Sm
iths bad backog was ikey increasing by at east 3.4 patients/week.
33.4 Patient 15 Hours 2 Patients = 3.4 Patients Week Week Week Hour
Smith patient were being een by hi colleague. Of the 44.8 patient who wante
d to ee Dr. Smith, only 33.4 actually aw hima PCP coverage ratio of 74.5 percen
t. Another intereting obervation wa that although three of the four doctor w
ere quite buy, Dr. Anderon wa buy for only 56.5 percent of her cheduled hou
r.
Improved and optimized model The VVH-Eat taff decided to imulate the followin
g change to thi bae model:
1. Dr. Anderon wa moved to the firt choice for new patient, a thi wa wher
e the highet volume wa and he had room in her chedule. Dr. Billing wa remo
ved from new patient. 2. A conequence of adding new patient to Dr. Anderon,
Dr. Billing wa now allocated a the firt backup to Dr. Smith. The new covera
ge rule are diplayed in Table 12.10.
364
Application to Contemporary Healthcare Operation Iue
FIGURE 12.7 VVH-Eat Advanced Acce Simulation: Arena Output of Bae Cae, Pati
ent Seen by Each Phyician
Counter Count Smith patient een by Smith Smith patient een by Anderon Smith p
atient een by Billing Total Anderon patient Total Billing patient Total Ku
marin patient Total New patient Total Smith patient Average 33.4500 10.4500 1
.9400 33.3400 37.4700 20.5400 47.6300 45.8400 Half Width 0.77 0.56 0.22 1.20 1.2
2 0.82 1.16 1.09 Minimum Average 26.0000 5.0000 1.0000 21.0000 23.0000 11.0000 3
6.0000 36.0000 Maximum Average 43.0000 18.0000 6.0000 47.0000 53.0000 34.0000 62
.0000 58.0000
Scheduled Phyician Utilization Anderon Billing Kumarin Smith
Average 0.5653 0.9911 0.9695 0.9001
Half Width 0.02 0.01 0.01 0.02
Minimum Average 0.3381 0.8545 0.8031 0.6758
Maximum Average 0.8118 1.0659 1.0505 1.0989
Queue Watiting Time Anderon patient treated.Queue Anderon ee Smith patient.
Queue Billing patient treated.Queue Billing ee Smith patient.Queue Kumarin
patient treated.Queue New patient treated.Queue Smith ee Smith patient.Queue
Half Width 0.06 0.00 0.20 0.00 0.20 0.19 0.27 Minimum Average 0.3526 0.00 0.501
2 0.00 0.5130 0.6180 1.2700 Maximum Average 1.9739 0.00 4.7427 0.00 5.2605 4.708
0 8.5842 Minimum Maximum Value Value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 5.2846 0
.00 8.2147 0.00 9.0715 8.2618 17.3846
Average 0.8739 0.00 2.2280 0.00 2.4131 2.1809 3.5750
NOTE: Model created in Arena, Rockwell Automation, Inc.
TABLE 12.10 New Phyician Backup Rule
Patient Primary Phyician Smith Billing Anderon Kumarin New patient
Primary Backup Smith Billing Anderon Kumarin Anderon
Secondary Backup Billing Kumarin Billing Kumarin
Tertiary Backup Anderon
Chapter 12: Scheduling and Capacity Management

365
3. Becaue the entire ytem eemed undertaffed baed on the queue time, it wa
agreed that Dr. Anderon and Billing would increae their hour of work in t
he clinic (Table 12.11). The revied model wa replicated 100 time; reult are
hown in Figure 12.8. The team oberved that thi taffing ytem improved key
metric in the VVH-Eat advanced acce ytem. Dr. Smith wa now eeing an aver
age of only 28.6 patient per week, which hould help to eliminate hi bad backl
og. Thi wa alo indicated by a reduction in hi effective queue time from 3.5
to 2.4 hour. The utilization of all taff wa now more balanced, and no ingle
phyician had a utilization rate of more than 91 percent. Becaue the cheduled
utilization of Dr. Billing dropped from 99 percent to 75 percent, he aked the
imulation team to rerun the model with hi hour reduced to the bae cae, allo
wing him to have Friday afternoon off. TABLE 12.11 Optimized Cae Phyician Sch
edule*
Phyician Anderon Billing Kumarin Smith Phyician Anderon Billing Kumarin Sm
ith
Monday Hour 1 2 3 4 1 1 1 1 1 1 1 1
5 1 1
6
7
8
Tueday Hour 1 2 3 4 5 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
6 1 1 1 6 1 1 1
7
8
1 1 1 6 1 1 1
1 1 7 1
1 1 8
1 1 7 1 1
1 1 8 1 1
Wedneday Hour 1 2 3 4 5 1 1 1 1 1 1 1 1 1 1 1
Thurday Hour 1 2 3 4 5 N 1 1 N 1 1 1 1 1
1 1 1 1
1 1
Friday Hour Phyician Anderon Billing Kumarin Smith All 1 1 1 2 1 1 3 1 1 1 4
1 1 1 5 6 7 8 Total Weekly Hour 24 24 32 15 95
1
N 1
N 1
N 1
*New hour indicated by N.
366
FIGURE 12.8 VVH-Eat Advanced Acce Simulation Arena Output: New Model Patient
Seen by Each Phyician
Counter Count Smith patient een by Smith Smith patient een by Anderon Smith p
atient een by Billing Total Anderon patient Total Billing patient Total Ku
marin patient Total New patient Total Smith patient Scheduled Phyician Utili
zation Anderon Billing Kumarin Smith Average 28.6300 2.9500 6.6600 33.4100 40.
4200 20.8700 48.9400 38.2400 Half Width 0.94 0.38 0.47 1.04 1.37 0.95 1.36 1.14
Minimum Average 18.0000 0.00 2.0000 19.0000 23.0000 9.0000 34.0000 25.0000 Maxim
um Average 43.0000 10.0000 14.0000 47.0000 56.0000 33.0000 66.0000 52.0000
Average 0.8274 0.7526 0.9008 0.7658
Half Width 0.02 0.02 0.01 0.03
Minimum Average 0.5471 0.5731 0.6737 0.4841
Maximum Average 1.0260 0.9364 1.0235 1.0841
Queue Watiting Time Anderon patient treated.Queue Anderon ee Smith patient.
Queue Billing patient treated.Queue Billing ee Smith patient.Queue Kumarin
patient treated.Queue New patient treated.Queue Smith ee Smith patient.Queue
Half Width 0.06 0.00 0.60 0.00 0.11 0.08 0.19 Minimum Average 0.2012 0.00 0.2287
0.00 0.1771 0.2487 0.8435 Maximum Average 1.8313 0.00 1.7219 0.00 3.0453 2.4176
5.1536 Minimum Maximum Value Value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 4.0748 0.
00 4.5518 0.00 7.3276 5.3610 13.0790
Average 0.5937 0.00 0.6397 0.00 1.3949 0.8708 2.4854
NOTE: Model created in Arena, Rockwell Automation, Inc.
Concluion
Advanced acce i a more efficient and patient-friendly method to deliver ambul
atory care. However, it i difficult to implement and maintain unle leaderhip
and taff are committed to it ucce. The ue of imulation tool uch a Are
na can help clinic leader explore variou option uch a cheduling rule to o
ptimize their advanced acce ytem.
Dicuion Quetion
1. What job equencing rule do you ee mot often in healthcare? Why? Can you th
ink of any additional job equencing rule not decribed in thi text?

367
2. How could advanced acce technique be ued for: a. An ambulatory urgery ce
nter? b. A freetanding imaging center? 3. What are the conequence of uing ad
vanced acce in a multipecialty clinic? How could thee tool be applied to pr
ovide ame-day cheduling? 4. Can advanced acce technique be ued with appoin
tment cheduling cheme? Why or why not?
Chapter Exercie
1. Two of the nure (Mary and Joe) at Riverview UCC have decided to work part-t
ime rather than full-time. They would like to work only two (conecutive) day p
er week. Becaue they would be part-time employee, alary and benefit per nur
e-day for thee nure would be reduced to $160 on weekday and $220 on weekend
day. Riverview could hire an additional full-time nure if needed. Should River
view UCC agree to thi requet? If the clinic doe agree, will additional nure
need to be hired? Auming that part-time nure and any new hire will accept
any chedule and preference for the remainder of the nure are the ame, what
new chedule would you recommend for each nure? 2. The VVH radiology department
currently ue FCFS to determine how to equence patient x-ray. On a typical d
ay, the department collect data related to patient x-ray. Ue thee data to co
mpare variou equencing rule. Auming thee data are repreentative, what rul
e hould the radiology department be uing and why? (An Excel preadheet contai
ning thee data can be found on ache.org/book/OpManagement.) Ue thee data to
compare variou equencing rule. Auming thee data are repreentative, what
rule hould the radiology department be uing and why?
Proceing Time (Minute) 35 15 35 25 30 25 35 30 20 Due Time (Minute from Now)
140 180 360 290 420 20 180 290 110 (Continued)
Patient A B C D E F G H I
368
Patient (Continued) K L M N O P Q R S
Proceing Time (Minute)
Due Time (Minute from Now)
25 15 30 20 20 10 10 15 20
150 270 390 220 400 330 80 230 370
3. Go to ache.org/book/OpManagement to ue the Optimal Outpatient Appointment
Scheduling Tool (Kaandorp and Koole 2007b) to compare variou appointment chedu
ling cheme. Firt, aume an eight-hour day that can be divided into 10-minute
time block (48 time interval), 15-minute ervice time for patient, 24 patien
t een according to the individual appointment cheme (a patient i cheduled t
o be een every 20 minute), and 5 percent no-how. Compare thi to the mall n
eighborhood optimal chedule with waiting time weight of one, idle time weight o
f one, and tardine weight of one. What are the difference in the two chedule
? Which would you chooe? Why? Increae the waiting time weight to three and co
mpute mall neighborhood optimal chedule. How i thi optimal chedule differen
t from the previou one? Finally, change the ervice time to 20 minute and comp
are the individual appointment chedule cheme to the mall neighborhood optimal
chedule with a waiting time weight of one and three. Which chedule would you
chooe and why? 4. A clinic want to work down it backlog to implement advanced
acce. The clinic currently ha 1,200 booked appointment and ee 100 patient
a day. The taff ha agreed to extend their chedule and can now ee 110 pati
ent per day. What i their current backlog, and how many day will it take to r
educe thi to 0? 5. Ue the VVH-Eat Arena model and remove all phyician backup
. How would you evaluate the reult baed on the advanced acce metric? 6. U
e the VVH-Eat Arena model and modify it by increaing patient demand for Dr.
Anderon and Billing by 20 percent. Modify provider chedule and rule to acco
mmodate thi increae, and evaluate the reult baed on the advanced acce met
ric.

369
Reference
Bailey, N. T. J., and J. D. Welch. 1952. Appointment Sytem in Hopital Outpatie
nt Department. Lancet 259: 11058. Cayirli, T., and E. Veral. 2003. Outpatient Sche
duling in Health Care: A Review of the Literature. Production and Operation Mana
gement 12: 51949. Cheang, B., H. Li, A. Lim, and B. Rodrigue. 2003. Nure Roteri
ng Problem A Bibliographic Survey. European Journal of Operational Reearch 151 (
1): 44760. Gupta, D., S. Potthoff, D. Blower, and J. Corlett. 2006. Performance M
etric for Advanced Acce. Journal of Healthcare Management 51 (4): 24659. Hutzc
henreuter, A. 2004. Waiting Patiently: An Analyi of the Performance Apect of
Outpatient Scheduling in Healthcare Intitute. Mater thei, Virje Univeriteit
, Amterdam. Kaandorp, G. C., and G. Koole. 2007a. Optimal Outpatient Appointment
Scheduling. Health Care Management Science 10 (3): 21729. . 2007b. Optimal Outpatient
Appointment Scheduling Tool. [Online information; retrieved 6/24/07.] http://obp
.math.vu.nl/healthcare/oftware/ ge/. Kelton, D. W., R. P. Sadowki, and D. P.
Sturrock. 2004. Simulation with Arena, 3rd ed. New York: McGraw Hill. Matthew,
C. H. 2005. Uing Linear Programming to Minimize the Cot of Nure Peronnel. Jour
nal of Healthcare Finance 32 (1): 3749. Parente, D. H., M. B. Pinto, and J. C. Ba
rber. 2005. A Pre-Pot Comparion of Service Operational Efficiency and Patient S
atifaction Under Open Acce Scheduling. Health Care Management Review 30 (3): 2
2028. Sanderon, P. 2007. Peronal correpondence with author. Singer, I. A. 2001
. Advanced Acce: A New Paradigm in the Delivery of Ambulatory Care Service. W
ahington, D.C.: National Aociation of Public Hopital and Health Sytem. Si
nger, I. A., and M. Regentein. 2003. Advanced Acce: Ambulatory Care Redeign
and the Nation Safety Net. Wahington, D.C.: National Aociation of Public Hop
ital and Health Sytem. Soriano, A. 1966. Comparion of Two Scheduling Sytem.
Operation Reearch 14 (3): 38897.
CHAPTER
13
SUPPLY CHAIN MANAGEMENT
CHAPTER OUTLINE
Operation Management in Action Overview Supply Chain Management Tracking and Ma
naging Inventory Inventory Claification Sytem Inventory Tracking Sytem Radi
o frequency identification Warehoue Management Demand Forecating Averaging Met
hod Simple moving average Weighted moving average Exponential moothing Trend,
Seaonal, and Cyclical Model Holt trend-adjuted exponential moothing techniqu
e Linear regreion Winter triple exponential moothed model Autoregreive inte
grated moving average model Model Development and Evaluation VVH Diaper Demand
Forecating Order Amount and Timing Economic Order Quantity Model VVH Diaper Ord
er Quantity Fixed Order Quantity with Safety Stock Model VVH Diaper Order Quanti
ty Additional Inventory Model Inventory Sytem Two-Bin Sytem Jut-in-Time Mat
erial Requirement Planning and Enterprie Reource Planning Procurement and Ve
ndor Relationhip Management Strategic View Concluion Dicuion Quetion Chap
ter Exercie Reference
370

ABC inventory claification ytem autoregreive integrated moving average (AR
IMA) bar coding caual forecating economic order quantity (EOQ) enterprie reo
urce planning (ERP) e-procurement forecating group purchaing organization (GP
O) jut-in-time (JIT) kanban linear regreion, linear trend material requiremen
t planning (MRP) mean abolute deviation (MAD) mean quared error (MSE) Pareto
principle point of ue, point of ervice (POS) radio frequency identification (R
FID) reorder point (R) afety tock (SS) ervice level (SL) imple moving averag
e (SMA) ingle exponential moothing (SES) tockout upply chain management (SCM
) time erie analyi, ee forecating trend-adjuted exponential moothing two
-bin ytem vendor-managed inventory (VMI) weighted moving average (WMA) Winter
triple exponential moothing
371

Chief financial officer earching for way to cut cot while jump-tarting er
vice might want to pend an afternoon with Allen Caudle, vice preident of uppl
y chain at Swedih Medical Center in Seattle, Wahington. Caudle upply-chain t
rategie aved Swedih more than $30 million in 2005. A hi peer increaingly
rely on group purchaing organization (GPO) to get better price from vendor,
Caudle and hi team manage Swedih pending the old-fahioned way, through tri
ct internal buying policie, central purchaing, elf-contracting, and tight ven
dor control. It want alway that way. Swedih witched from the GPO model to el
f-contracting even year ago, bucking the trong indutry trend that fueled the
growth of purchaing powerhoue like Novation, Premier, and Neoforma (now Glob
al Health Exchange). The homegrown ytem work94 percent of Swedih pending i
done on contractpartly becaue internal department head and vendor alike mut a
bide by the golden rule: Nobody get paid without an approved purchae order. Ne
w-manager orientation include an hour long upply-chain primer from Caudle to d
rive the leon home. After that, It only take one time for them to learn. If a
manager ha a plauible excue, I give them one mulligan. But I alo ay, Dont ak f
or another chance. New vendor are adept at finding way around the rule, but oon
realize their reward i nonpayment, he add. So that there are no urprie, al
l vendor firt have to pa a 10-quetion quiz on Swedih contracting arrangeme
nt. The firt entence of the document read, I undertand that I will not be pa
id if I dont have a purchae order for thi project, Caudle ay. A devotee of Toy
ota lean management approach, he alo eye everything from anetheiology technici
an work habit to the location of infuion pump in patient room with wate eli
mination in mind. Hi logan i: Ue our mind before we pend our buck. And thou
gh he work for a healthcare provider that encompae four hopital, more than
a dozen clinic, a homecare ervice, and affiliated phyician office, hi appr
oach could be adapted to mot indutrie, Caudle ay. For intance, an in-houe
time-and-efficiency tudy found that Swedih anetheiology technician pent 34
percent of their time looking for and tocking upplie rather than giving care
to patient under edation . . . alo, 30 percent of the anethetic drug were
out of date, Caudle report. Outpatient urgery wa another problem child. Clinic
ian had to leave the operating room area an average of eight time per cae to
find and retrieve equipment, tool, or upplie. To remedy both the anetheiolo
gy and outpatient urgery dilemma, Swedih overhauled the phyical environment.
372
Chapter 13: Supply Chain Management

373
We cleaned up the area and got rid of wated pace. Moved room of equipment that
were rarely or never ued. Put inventorie of item that were ued right outid
e the operating room, at the point of ue. Created cae-cart a needed, intea
d of ahead of time, he explain. It wa like cleaning out your garage. Mot of it
wa pretty baic. Through the reorganization, Caudle ay they focued on a gener
al guideline of le variability and greater tandardization. No work proce e
caped crutiny. Even the clinical procedure litort of like recipe naming tool,
poitioning device, and the like required to perform a gallbladder removal or
hernia repairwere examined for unneceary item. Now, go-get in outpatient urger
y are down from eight to one per cae. My goal i to get mundane upply-chain dut
ie out of their way o they can pend more time with patient, he add. Swedih
i jut one example of how upply chain management can operationally tranform h
ealthcare, according to Loui Fieren, enior vice preident of upply chain and
capital project management for Trinity Health. Baed in Novi, Michigan, Trinity
i the fourth-larget Catholic health ytem in the United State, with jut un
der 45,000 employee, 44 hopital (29 owned, 15 managed), and 379 outpatient cl
inic. Revenue were $5.7 billion in 2005. Fieren cite a much-quoted tatitic
that 31 percent of hopital expene, on average, are attributed to upplying t
he facility, which he aptly decribe a an argument for a higher level of manage
ment. But intead of going the Swedih route, many hopital are turning to compa
nie like Wetminter, Coloradobaed Global Health Exchange, provider of an elect
ronic trading exchange. Memberbuyer and ellerare healthcare player and includ
e manufacturer, ditributor, facilitie uch a hopital, nuring home, clin
ic and hopice, GPO, and upply-chain firm. Global let hopital connect to
u once, and connect to provider, kind of like Amazon.com, explain pokewoman
Karen Conway. Hopital member ave time and money by handling upply chain func
tion with multiple trading partner through the exchange automated environment,
he add. Bottom line, healthcare utilization of companie like Global can impac
t the national economy. That a big claim, but Conway ha number to back it up. F
or intance, according to a tudy by the Healthcare Ditribution Management Ao
ciation, many hopital overpay medical/urgical upplier by 27 percent. Another
indutry tudy found that 40 percent of a hopital buyer time and 68 percent of
an account-payable worker time i pent on manual proceing and rework, he con
tinue. Another common problem i data ynchronization, in which the hopital da
ta and the upplier data dont match. Conway ay there i an average 35 percent i
nconitency between hopital and upplier data that,
374
obviouly, caue error and lead to more rework. The average cot hared by a h
opital and upplier to reearch and correct a ingle order dicrepancy i $15 t
o $50, he add. Then there the human factor to conider. Many hopital buy expe
nive material management oftware, but dont get the expected return on invetme
nt due to inadequate uer trainingtoo little intruction or too hort a learning
curveor high turnover among trained peronnel. We forget that it take time to cha
nge, Conway comment, adding that failed IT implementation are often followed by
a trip through the trough of diilluionment. Going Global offer tunning effici
encie through it proce automation, he ay: Requiition are filled three t
ime fater; purchae order completed even to 15 time fater; invoice proce
ed twice a fat. Error go down, too. For intance, Conway cite a 50 percent
reduction in purchae-order dicrepancy between order initiation, acknowledgment
, and invoice. Other cutomer report pending half a much time reearching inv
oice dicrepancie thank to automation. The goal, he ay, i to boot automati
on between financial, material and clinical, and to increae data accuracy. Thi
i not a revolution, Conway conclude. It definitely an evolution.
SOURCE: Knowledge@W.P. Carey. Tranforming U.S. Health Care: Supply-Chain Makeove
r Rejuvenate Medical Center. W. P. Carey School of Buine, Arizona State Unive
rity. knowledge.wpcarey.au.edu/index.cfm?fa= viewArticle&id=1223&pecialId=42.
Reprinted with permiion.
Overview
Ecalating cot are driving many healthcare organization to examine and optimi
ze their upply chain. Efficient and effective upply chain management (SCM) i
becoming increaingly important in the healthcare arena. Thi chapter introduce
the concept of SCM and the variou tool, technique, and theorie that can en
able upply chain optimization. The major topic covered include: SCM baic; T
ol for tracking and managing inventory; Forecating; Inventory model; Inventor
y ytem; Procurement and vendor relationhip management; and Strategic SCM.
After completing thi chapter, reader hould have a baic undertanding of SCM,
which will help them undertand how SCM could be ued in

375
their organization and enable them to begin to employ SCM-related tool, techni
que, and theorie to optimize upply chain.
Supply Chain Management
The upply chain include all of the procee involved in getting upplie and
equipment from the manufacturer to ue in patient care area. SCM i the managem
ent of all activitie and procee related to both uptream vendor and downtr
eam cutomer in the value chain. Becaue SCM require managing relationhip ou
tide a well a inide an organization, SCM i a broad field of thought. SCM i
becoming increaingly important in the healthcare arena. Supply cot in hopit
al account for between 15 percent and 25 percent (Healthcare Financial Manageme
nt Aociation 2002, 2005) of operating cot. Tranaction cot are etimated a
t $150 per order for the buyer and eller (Healthcare Financial Management Aoc
iation 2001). SCM i aimed at reducing cot and increaing efficiencie aocia
ted with the upply chain. Effective SCM i enabled by new technologie and old me
thodologie ued to not only reduce upply-aociated cot and effort but alo
to improve the efficiency of upply procee. Many of the technique ued to im
prove upply chain performance in other indutrie are applicable to healthcare.
Technology-enabled olution include e-procurement, radio frequency identificat
ion (RFID), bar coding, point-of-ue data entry and retrieval, and data warehou
ing and management. Thee technologie have been ued in other indutrie, and h
ealthcare i increaingly finding that they can not only reduce cot, but alo i
ncreae afety. A ytem view of the upply chain can lead to a better underta
nding of procee and how bet to improve and optimize them. SCM i really abou
t managing relationhip with vendor and cutomer to enable the optimization o
f the entire chain (rather than jut piece of it) and reult in benefit for a
ll member of the chain. For SCM to be effective, information i needed. Reliabl
e and accurate data are needed to determine where the greatet improvement and
gain can be made by improving the upply chain.
Tracking and Managing Inventory
Inventory i the tock of item held by the organization either for ale or to
upport the delivery of a ervice. In healthcare organization, inventory typical
ly include upplie and pharmaceutical. Thi tock allow organization
376
to cope with variation in upply and demand while making cot-effective orderin
g deciion. Inventory management i concerned with determining how much invento
ry to hold, when to order, and how much to order. Effective and efficient invent
ory management require a claification ytem, inventory tracking ytem, reli
able forecat of demand, knowledge of lead time, and reaonable etimate of ho
lding, ordering, and hortage cot.
Inventory Claification Sytem
Not all inventory i equal: Some item may be critical for the organization oper
ation, ome may be cotly or relatively inexpenive, and ome may be ued in la
rge volume while other are eldom needed. A claification ytem can enable o
rganization to manage inventory more effectively by allowing them to focu on t
he mot important inventory item and place le emphai on le important item
. The ABC claification ytem divide inventory item into three categorie b
aed on the Pareto principle. Vilfredo Pareto tudied the ditribution of wealth
in 19th-century Milan and found that 80 percent of the wealth wa controlled by
20 percent of the people (Reh 2007). Thi ame idea of the vital few and the tr
ivial many i found in quality management (Chapter 8) and ale (80 percent of
ale come from 20 percent of cutomer) a well a inventory management. The A i
tem have a high dollar volume (70 percent to 80 percent) but account for only 5
percent to 20 percent of item, B item are moderate dollar (30 percent) and it
em (15 percent) volume, and C item are low dollar (5 percent to 15 percent) and
high item (50 percent to 65 percent) volume. The claification of item i not
related to their unit cot; an A item may have high dollar volume becaue of hi
gh uage and low cot or high cot and low uage. Item vital to the organizatio
n hould be aigned to the A category even if their dollar volume i low to mod
erate. The A item are the mot important and, therefore, the mot cloely manag
ed. The B and C item are le important and le cloely managed. In a hopital
etting, pacemaker are an example of A item and facial tiue might be a C it
em. The A item are likely ordered more often, and inventory accuracy i checked
more often. Thee item are good candidate for bar coding and point-of-ue y
tem. The C item do not need to be a cloely managed and, often, a two-bin y
tem i ued for management and control.
Inventory Tracking Sytem
An effective inventory management ytem require a mean of determining how muc
h of a particular item i available. In the pat, inventory record were updated
manually and were typically not very accurate. Bar coding and pointof-ue yte
m have eliminated much of the data input inaccuracy, but inven-

377
tory record are till imperfect. A phyical count mut uually be performed to
enure that the actual and recorded amount are the ame. Although many organiza
tion perform inventory count on a periodic bai (e.g., once a month), cycle c
ounting ha been found to be more helpful in enuring accuracy and eliminating e
rror. More accurate inventory record not only enable efficient inventory manag
ement, but they can alo help to eliminate the hoarding that occur when provide
r are concerned the item will be unavailable when needed. In a typical cycle-co
unting ytem, a phyical inventory i performed on a rotating chedule baed on
item claification. The A item might be counted every day and C item only on
ce a month. Electronic medication order and matching can not only allow an orga
nization to track demand, but it can alo reult in greater patient afety the pa
tient and order are matched at the time of adminitration. Rule can be input in
to the ytem to alert provider to advere drug interaction and, thu, help to
eliminate error. Sytem are being developed that allow for complete, current
patient record available bedide. The availability of patient and drug hitory
can make care better and afer.
Radio frequency identification RFID i a tool for identifying object, collectin
g data about them, and toring that data in a computer ytem with little human
involvement. RFID tag are imilar to bar code, but they emit a data ignal tha
t can be read without actually canning the tag. RFID tag can alo be ued to d
etermine the location of the object to which they are attached. However, RFID ta
g are more expenive than bar coding. The tudy RFID in Healthcare (BearingPoint
and National Alliance for Healthcare Information Technology 2006) found that RFI
D technology i being ued in a variety of application within the healthcare in
dutry, including patient flow management, acce control and ecurity, upply c
hain ytem, and mart helving. Real-time medical equipment tracking ytem a
nd patient afety ytem, uch a thoe for identification and medication admin
itration, were found to be the major area where RFID i expected to be ued in
the future. PinnacleHealth hopital in Harriburg, Pennylvania, ha ucceful
ly implemented RFID technology to track and locate expenive medical equipment (
Wright 2007). The ytem can be queried to determine where a particular piece of
equipment i currently located, enabling employee to quickly find the needed e
quipment rather than earch the hopital for it. PinnacleHealth believe the y
tem will pay for itelf in two year by enabling the hopital to determine real
uage of purchaed and rented equipment and eliminating theft.
378
Warehoue Management
Warehoue management ytem can enable healthcare organization to manage and d
ecreae their torage and facility cot. Bar coding and point-of-ue ytem ca
n reduce the labor needed by automating data entry in receiving. Automated data
entry will alo reduce error and allow for more accurate determination of actua
l inventory held. Information about demand to the warehoue or torage facility
can be ued to organize that facility o that more heavily demanded item are mo
re acceible and le frequently demanded item are not a acceible. Thi can
ignificantly reduce labor cot aociated with the torage facility.
Demand Forecating
Knowledge of demand and demand variation within the ytem can enable improved d
emand forecating, which, in turn, can allow inventory reduction and greater a
urance that omething will be available when needed. Bar coding and point-of-u
e ytem allow organization to track when and how many upplie are being con
umed, ue that information to forecat demand organization wide, and plan how to
meet that demand in the mot effective manner. Forecating, or time erie anal
yi, i ued to predict what will happen in the future, baed on data obtained
at et interval in the pat. For example, forecating could be ued to predict
the number of patient who will be een in the emergency department in the next
year (month, day) baed on the number of patient who have been een there in th
e pat. Time erie analyi account for the fact that data taken over time may
be related to one another and, therefore, violate the aumption of linear reg
reion. Forecating method range from imple to complex. Here, the impler met
hod are decribed; only a brief dicuion of the more complicated method i p
rovided.
Averaging Method
All averaging method aume that the variable of interet i table or tationa
rynot growing or declining over time and not ubject to eaonal or cyclical vari
ation.
Simple moving average A imple moving average (SMA) take the lat p-value and
average them to forecat the value in the next period.

379
Ft =
Dt 1 + Dt 2 + ... + Dt
where Ft = forecast for eriod t (or the coming eriod), Dt 1 = value in the re
vious time eriod, and = # of time eriods.
Weighted moving average In contrast to the SMA, where all values from the ast a
re given equal weight, a weighted moving average (WMA) weights each of the revi
ous time eriods. Tyically, the more recent eriods are assumed to be more rele
vant and are given higher weight.
Ft = w 1Dt 1 + w 2Dt 2 + ... + w Dt
vious time eriod, w = weight for time eriod , and w1 + w2 + ... + w = 1.
Exonential smoothing The roblem with the revious two methods, SMA and WMA, is
that a large amount of historical data are required. With single exonential sm
oothing (SES), the oldest data are eliminated once new data have been added. The
forecast is calculated by using the revious forecast as well as the revious a
ctual value with a weighting or smoothing factor, alha. Alha can never be grea
ter than one, and higher values of alha ut more weight on the most recent eri
ods.
Ft = Dt 1 + ( 1 )Ft 1
vious time eriod, and = smoothing constnt 1.
Trend, Sesonl, nd Cyclicl Models
Holts trend-djusted exponentil smoothing technique SES ssumes tht the dt fl
uctute round resonly stle men (no trend or consistent pttern of growt
h or decline). If the dt contin trend, Holts trend-djusted exponentil smoo
thing model cn e used. Trend-djusted exponentil smoothing works much like si
mple smoothing except tht two componentslevel nd trendmust e updted ech perio
d. The level is smoothed estimte of the vlue of the dt t the end of ech
period, nd the trend is smoothed estimte of verge growth t the end of ec
h period. Agin, the weighting or smoothing fctors, lph nd delt, cn never
exceed one, nd higher vlues put more weight on more recent time periods.
380
Applictions to Contemporry Helthcre Opertions Issues
FITt = Ft + Tt
nd
Ft = Dt 1
where FITt =
or perio t,
perio, 0
+ ( 1 )FITt 1 Tt = Tt 1 + ( Ft 1 FITt 1 )
forecast for perio t incluing the tren, Ft = smoothe forecast f
Tt = smoothe tren for perio t, Dt 1 = value in the previous time
=smoothing constnt 1, nd 0 = smoothing constant 1.
Linear regression Alternatively, when a tren exists in the ata, regression ana
lysis (Chapter 7) is often use for forecasting. Deman is the epenent, or Y,
variable, an the time perio is the preictor, or X, variable. The regression e
quation
Y = b (X ) + a
can be restate using forecasting notation:
Ft = b (t ) + a
where Ft = forecast for perio t, b = slope of the regression line, an a = Y in
tercept. To fin b an a, D = actual eman, D = average of all actual emans,
t = time perio, an t = average of time perios: b = (t t ) (D D )/(t t )2 an a
= D bt . In time series forecasting, the preictor variable is time. Regression
analysis is also use in forecasting when a causal relationship exists between a
preictor variable (not time) an the eman variable of interest. For example,
if the number of surgeries to be performe at some future ate is known, that i
nformation can be use to forecast the number of surgical supplies neee.
Winters triple exponential smoothe moel In aition to ajusting for a tren, W
inters triple exponential smoothe moel ajusts for a cycle or seasonality. Auto
regressive integrate moving average moels Autoregressive integrate moving ave
rage (ARIMA) moels, evelope by Box an Jenkins (1976), are able to moel a wi
e variety of time series
Chapter 13:
upply Chain Management
381
behavior. However, ARIMA is a complex technique; although it often prouces appr
opriate moels, it requires a great eal of expertise to use.
Moel Development an Evaluation
Forecasting moels are evelope base on historical time series ata using the
previously escribe techniques. Typically, the best moel is the simplest moel t
hat minimizes the forecast error associate with that moel. Mean absolute evia
tion (MAD) an/or mean square error (M
E) can be use to etermine error levels
.
MAD =
Dt Ft
t =1
n
n
M
E =
(D
t =1
n
t
Ft )
2
n
t = Perio number D = Actual eman for the perio
F = Forecast eman for the perio n = Total number of perios
(Many of these forecasting moels are available as ownloas on the companion we
b site, www.ache.org/books/OpsManagement).
VVH Diaper Deman Forecasting
Jessie Jones, purchasing agent for Vincent Valley Hospital an Health
ystem (VV
H), wants to forecast eman for iapers.
he gathers information relate to pas
t eman for iapers (Table 13.1) an plots it (Figure 13.1). The plot of weekly
eman shows no cycles or trens, so Jessie believes that an averaging metho w
oul be most appropriate.
he compares the forecasts obtaine with a five perio
MA; WMA with weights of 0.5, 0.3, an 0.2; an exponentially smoothe forecast
with alpha of 0.25.
MA forecast:
p A + A12 + A11 + A10 + A 9 60 + 43 + 53 + 54 + 45 F14 = 13 = = 51 5 5
WMA forecast:
Ft =
At 1 + At 2 + ... + At p
Ft = w 1 At 1 + w 2 At 2 + ... + w p At p F14 = w 1 A13 + w 2 A12 + .w 3 A11 = 0.5

60 + 0.3 43 + 0.2 53 = 53.5
382
TABLE 13.1 VVH Weekly Diaper Deman
Perio 1 2 3 4 5 6 7 8 9 10 11 12 13
Week of 1 Jan 8 Jan 15 Jan 22 Jan 29 Jan 5 Feb 12 Feb 19 Feb 26 Feb 5 Mar 12 Mar
19 Mar 26 Mar
Cases of Diapers 70 42 63 52 56 53 66 61 45 54 53 43 60
FIGURE 13.1 Plot of VVH Weekly Diaper Deman
80 70
Weekly eman
60 50 40 30 20 10 0 1 2 3 4 5 6 7 Perio 8 9 10 11 12 13
Chapter 13:
383
Exponentially smoothe forecast:
Ft = At 1 + ( 1 )Ft 1 F14 = 0.25 A13 + 0.75 F13 = 0.25 60 + 0.75 52 = 54
Because each metho gives a ifferent forecast, Jessie ecies to compare the me
thos to try to etermine which is best.
he uses the Excel Forecasting Template
(foun on the companion web site) to perform the calculations (Figure 13.2).
h
e fins that both MAD an M
E are lowest with the WMA metho an ecies to use
that metho for forecasting. Therefore, she forecasts that 53.5 cases of iapers
will be emane the week of April 2, perio 14.
Orer Amount an Timing
Inventory management is concerne with the following questions: How much invento
ry shoul the organization hol? When shoul an orer be place? How much shoul
be orere? To answer these questions, organizations nee to have reasonable es
timates of holing, orering, an shortage costs. Knowlege of lea times an
imple Moving Average

Weighte Moving Average (3 perios)
Weight 3 Weight Weight 2 1
ingle Exponential
moothing
Perios MAD M
E
5 7 86
Least Recent MAD M
E
0.2 6 75
0.3
0.5
Most Recent MAD M
E
0.25 8 135
FIGURE 13.2 Excel Forecasting Template Output: VVH Diaper Deman
Perio Actual Forecast Error
1 2 3 4 5 6 7 8 9 10 11 12 13 14 70 42 63 52 56 53 66 61 45 54 53 43 60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 70 42 63 52 56 53 66 61 45 54 53 43 60
1 2 3 4 5 6 7 8 9 10 11 12 13 14 70 42 63 52 56 53 66 61 45 54 53 43 60 70 63 63
60 59 58 60 60 56 56 55 52 54 28 0 11 4 6 8 1 15 2 3 12 8
57 53 58 58 56 56 56 51 51
4 13 3 13 2 3 13 9
58 53 56 54 60 61 54 53 52 48 53.5
6 3 3 12 1 16 0 0 9 12
384
eman forecasts is also essential to etermining the best answers to inventory

questions.
Economic Orer Quantity Moel
In 1915, F. W. Harris evelope the economic orer quantity (EOQ) moel to answe
r inventory questions. Although the assumptions of this moel limit its usefulne
ss in real situations, it provies important insights into effective an efficie
nt inventory management. To ai in unerstaning the moel, efinitions for some
key inventory terms are provie. Lea time is the interval between placing an
orer an receiving it. Holing (or carrying) costs are associate with keeping
goos in storage for a perio of time, usually one year. The most obvious of the
se costs is the cost of the space an the cost of the labor an equipment neee
to operate the space. Less obvious costs inclue the opportunity cost of capita
l an those costs associate with obsolescence, amage, an theft of the goos.
These costs are often ifficult to measure an are commonly estimate as one thi
r to one half the value of the store goos per year. Orering (or setup) costs
are the costs of orering an receiving goos. They may also be the costs assoc
iate with changing or setting up to prouce another prouct.
hortage costs are
the costs of not having something in inventory when it is neee. Inepenent
eman is generate by the customer an is not a result of eman for another goo
or service. Depenent eman results from another eman. For example, the em
an for hernia surgical kits (epenent) is relate to the eman for hernia sur
geries (inepenent). Back orers cannot be fille when receive, but the custom
er is willing to continue waiting for the orer to be fille.
tockouts occur wh
en the esire goo is not available. The basic EOQ moel is base on the follow
ing assumptions: Deman for the item in question is inepenent; Deman is known
an constant; Lea time is known an constant; Orering costs are known an con
stant; an Back orers, stockouts, an quantity iscounts are not allowe.
The EOQ inventory orer cycle (Figure 13.3) consists of stock or inventory being
receive at a point in time. An orer is place when the amount of stock on han
is just enough to cover the eman that will be experience uring lea time.
The new orer arrives at the exact point when
Chapter 13:
385
the stock is completely eplete. The point at which new stock shoul be orere
, the reorer point (R), is the quantity of stock emane uring lea time. Reo
rer point = R = L where R = reorer point, = average eman per time perio,
an L = lea time (in the same units as above). The EOQ inventory orer cycle s
hows that the average amount of inventory hel will be: Orer quantity Q = 2 2 T
he number of orers place in one year will be: Yearly eman D = Orer quantity
Q Total costs are the sum of holing an orering costs. Yearly holing costs a
re: Cost to hol one item one year Average inventory = h Q 2
Orer qty, Q Inventory level
Deman rate
FIGURE 13.3 EOQ Inventory Orer Cycle
Reorer point, R
0
Orer place
Orer receive
Orer place Lea time
Orer receive
Lea time
386
Yearly orering costs are: Cost to place one orer Yearly number of orers = o D
Q
Total yearly costs are then:
h Q 2 + o D Q
Figure 13.4 illustrates these relationships. An inspection of this graph shows t
hat total cost will be minimize when holing costs equal orering costs. (This
can also be proven using calculus.) The orer quantity that will minimize total
costs is foun when:
h Q * 2 = o D Q *
Rearranging this equation, the optimal orer quantity is:
Q 2 = (2 o D) h an Q* =
2o D h
FIGURE 13.4 EOQ Moel Cost Curves
Annual cost ($)
Minimum total cost
Total cost
Carrying cost = h Q/2
Orering cost = o D/Q
Orer quantity (Q) Optimal orer quantity Q*
Chapter 13:
387
everal important insights into inventory management can be gaine from an exami
nation of this simple moel. There are trae offs between holing costs an ore
ring costs: As holing costs increase, optimal orer quantity ecreases, an as
orering costs increase, optimal orer quantity increases. Many organizations, i
ncluing those in the healthcare inustry, believe that the costs of holing inv
entory are much higher than was previously thought. As a consequence, these orga
nizations are ecreasing orer quantities an working to ecrease orer costs by
streamlining procurement processes.
VVH Diaper Orer Quantity
Jessie Jones, VVHs purchasing agent, wants to etermine the optimal orer quantit
y for iapers. From her forecasting work, she knows that annual eman for iape
rs, D, is:
Time Perio =
53.5 Cases 52 Weeks 2, 782 Cases = Year Week Year
Each case of iapers costs $5, an Jessie estimates holing costs at 33 percent.
It costs $100 to place an orer. Lea time for iapers is 1 week.
he calculate
s the EOQ, Q*, as: 2o D = h 2 $100 2, 782 Cases $1.67 Case
= 333, 174 Cases 2 = 577 Cases
he calculates the reorer point, R, as:
L =
53.5 Cases 1 Week = 53.5 Cases Week
Jessie will nee to place an orer for 577 cases of iapers when 53.5 cases rema
in in stock.
Fixe Orer Quantity with
afety
tock Moel
The basic EOQ moel assumes that eman is constant an known. This means that t
he amount of stock carrie in inventory only nees to match eman. In reality,
eman is selom constant, an excess inventory must be hel to meet variations
in eman an avoi stockouts. This excess inventory is calle safety stock (
)
an is efine as the amount of inventory carrie over an above expecte eman
. Figure 13.5 illustrates this moel.
388
Inventory level
FIGURE 13.5 Variable Deman Inventory Orer Cycle with
afety
tock
Orer quantity (Q)
Reorer point (R)
afety stock (
) 0 Lea time Lea time

Time
This
moel assumes that eman varies an is normally istribute (Chapter 7)

. The moel also assumes that a fixe quantity equal to EOQ will always be orer
e. The EOQ will remain the same as in the basic moel, but the reorer point wi
ll be ifferent because of the nee for
.
R = L +
The amount of
to carry is etermine by variation in eman an esire servi

ce level.
ervice level is efine as the probability of having an item on han
when neee. For example, suppose that orers are place at the beginning of a t
ime perio an receive at the en of that perio. If eman is expecte to be 1
00 units in the next time perio with a stanar eviation of 20 units an 100 u
nits on han at the start of the perio, the probability of stocking out is 50 p
ercent an the service level is 50 percent. If eman is normally istribute, t
here is a 50 percent probability of its being higher than the mean an a 50 perc
ent probability of its being lower than the mean. Deman woul be greater than t
he stock on han in half of the time perios. To increase the service level,
is neee. For example, if the stock on han at the start of the time perio is
120 units (20 units of
), the service level woul increase to 84 percent an t

he probability of a stockout woul be reuce to 16 percent. Because eman is a
ssume to follow a normal is
Chapter 13:
389
tribution an 120 units is exactly one stanar eviation higher than the mean o
f 100 units, the probability of being less than one stanar eviation above the
mean is 84 percent. There is a 16 percent probability of being more than one st
anar eviation above the mean (Figure 13.6). A service level of 95 percent is
typically use in inustry. However, if one stockout every 20 time perios is un
acceptable, a higher service level target is neee.
is the z value associate

with the esire service level (number of stanar eviations above the mean)
times the stanar eviation of eman uring lea time.
= z L
Note that with thi model, the only time demand variability can be a problem i
during lead time. Becaue an order i triggered when a certain level of tock i
reached, any variation in demand prior to that time doe not affect the reorder
point. Thi model alo provide ome important inight into inventory manageme
nt. There are trade-off between the amount of SS held and ervice level. A the
deired ervice level increae, the amount of SS neededand therefore the amount
of inventory heldincreae. A the variation in demand during lead time increae
, the amount of SS increae. If demand variation or lead time can be decreaed
, the amount of SS needed to reach a deired ervice level will alo decreae. M
any healthcare organization are working with their upplier to reduce lead tim
e and, therefore, SS level. FIGURE 13.6 Service Level and Safety Stock
Probability of meeting demand during lead time = ervice level = 84%
R = Reorder point
Probability of a tockout = 16%
Example unit
100
120
Average demand during lead time = dL Z 0 1
390
VVH Diaper Order Quantity
After learning more about inventory model, Jeie Jone realized that the reord
er point that he had decided on uing the baic EOQ model would caue the hopi
tal to be out of diaper during 50 percent of the order cycle. Becaue diaper
would be ordered five time per year, thi meant that the hopital would be out
of diaper at leat twice a year. Jeie believed that thi wa an unacceptable
amount of tockout and decided that SS wa needed to avoid them. She decided th
at a ervice level of 95 percent, or one tockout every 4 year, would be accept
able. Jeie gathered more information related to demand for diaper over the pa
t year and determined that the tandard deviation of demand during lead time wa
11.5 cae of diaper. She calculated the amount of SS needed a:
z L = 1.64 11.5 = 18.9 Cae
Her new reorder point i: 53.5 Cae dL + SS
Week She will need to place an order for 577

remain in tock. The forecating template found on
ued to perform thee calculation, and the output
hown in Figure 13.7.
=
1 Week
+ 18.9 Cae = 72.4 Cae
cae of diaper when 72.4 cae
the companion webite can be
related to Jeie problem i
Additional Inventory Model

Many inventory model that addre ome of the limiting aumption of the EOQ m
odel have been developed. One that may be of interet i the fixed time period w
ith SS model. In the fixed order quantity with SS model, the order quantity i f
ixed and the time when the order i placed varie. In the fixed time period with
SS model, the order quantity varie and the time when FIGURE 13.7 VVH 95 Percen
t Service Level Reorder Point
Reorder Point (ROP) with EOQ Ordering Average daily demand Average lead time Std
dev demand during lead time Service level increment Stockout rik z aociated
with ervice level Average demand during lead time Safety tock Reorder point d
= 7.64 Unit 7 Day L= L= 11.5 Unit SL = 0.95 SL = 0.05 1.64 dL = 53.48 Unit S
S = 18.9 Unit ROP = 72.4 Unit Reorder Point Probability 0.0
20.0
40.0
60.0
80.0
100.0
Daily demand Daily demand ROP

391
the order i placed i fixed. Thi type of model i applicable when vendor deli
ver on a et chedule or if one upplier i ued for many different product and
order are bundled and delivered together on a et chedule. Generally, thi i
tuation require more SS becaue tockout are poible during the entire time b
etween order, not jut the lead time for the order. Model that account for qua
ntity dicount and price break have alo been developed. More information on t
hee higher-level model can be found in mot inventory management textbook.
Inventory Sytem
In practice, variou type of ytem are employed for management and control of
inventory. They range from imple to complex, and organization typically emplo
y a mixture of thee ytem.
Two-Bin Sytem
The two-bin ytem i a imple, eaily managed ytem often ued for Band/or C-t
ype item. In thi ytem, inventory i eparated into two bin. Thee do not ne
cearily have to be actual bin or container; a mean of identifying the item
a being in the firt or econd bin i imply needed. Inventory i taken from t
he firt bin. When that bin i emptied, an order i placed. Inventory from the
econd bin i ued during the lead time for the order to be received. The amount
of inventory held in each bin can be determined from the fixed order quantity wi
th SS model. The amount of inventory held in the firt bin would ideally be the
EOQ minu the reorder point. The amount of inventory in the econd bin would equ
al the reorder point.
Jut-in-Time
Jut-in-time (JIT) inventory ytem are baed on Lean concept and employ a typ
e of two-bin ytem called the kanban ytem. (See Chapter 9 for a decription o
f thi type of ytem.) Becaue inventory level are controlled by the number of
kanban in the ytem and inventory i wate in a Lean ytem, organization try
to decreae the number of kanban a much a poible.
Material Requirement Planning and Enterprie Reource Planning
Material requirement planning (MRP) ytem were firt employed by manufacturin
g organization in the 1960 when computer became commercially available. Thee
ytem were ued to manage and control the purchae and production of dependen
t-demand item. A imple example illutrate the logic of MRP (Figure 13.8). A t
able manufacturer know (or forecat) that 50 table, coniting of a top and f
our
392
leg, will be demanded five week in the future. The manufacturer know that it
take one week to produce a table if both the leg and top are available. The c
ompany alo know that there i a two-week lead time for table leg and a threeweek lead time for table top. From thi information, MRP determine that for th
e organization to have 50 table in week 5, it need to have 50 table top and 2
00 table leg in week 4. The company alo need to order 200 table leg in week
2 and 50 table top in week 1. The ame type of logic can be employed in healthc
are for dependentdemand item. For example, if the demand for a particular type
of urgery i known or can be forecat, upplie related to thi type of urgery
can be ordered baed on MRP-type logic. Enterprie reource planning (ERP) y
tem evolved from thee relatively imple ytem a computing power grew and o
ftware application became more ophiticated. ERP-type ytem found in healthc
are today encompa the entire organization and include finance, accounting, hum
an reource, patient record, and o on, in addition to inventory management an
d control function. Technology can enable improvement in many apect of health
care organization.
Procurement and Vendor Relationhip Management
Analyzing the procee ued for procurement and improving them can reult in i
gnificant aving to an organization. Technology can be ued to not only treaml
ine procee but alo to improve data reliability, accuracy, and viibility. St
reamlining procurement procee can reduce aociated labor cot. Eprocurement
i one example of how technology can be employed to make procurement more effic
ient. The eae of obtaining product information,
FIGURE 13.8 MRP Logic
Order table top
Order table leg
Week
1
2
3
4
5

393
reduced time aociated with the actual procurement proce, and increaed ue o
f a limited number of upplier can ignificantly reduce cot. Information abou
t upplier reliability can be maintained in thee ytem to allow organization
to make informed choice about vendor. For example, one vendor may be much le
expenive but extremely unreliable, wherea another may be lightly more expen
ive but more reliable and fater. An analyi may how that it i le cotly t
o ue the lightly more expenive vendor becaue the amount of SS held or the ne
ed to expedite hipment may be reduced. Tracking and reviewing upplier perform
ance can aid in enuring that quality upplier are being ued. Value-baed tan
dardization can be employed to reduce both the number and quantity of item held
. Focuing on high-ue or high-cot item can leverage the benefit of tandardi
zation and reduce the number of upplier to the organization. Fewer upplie an
d upplier can reult in both labor and material cot aving. Outourcing can
be an effective mean of enuring upply availability and reducing internal labo
r. Ditributor can break order down by point of ue and deliver directly to th
at point a needed rather than having the organization peronnel perform that fu
nction. Prepackaged upply pack or urgical cart can reduce the amount of in-h
oue labor needed to organize thee upplie and enure that the correct upplie
are available when needed. Vendormanaged inventory i another way to outource
ome of the work involved with procurement. Automated upply cart or cabinet
and point-of-ue ytem can enable vendor-managed inventory. GPO can be employ
ed to increae order quantitie and reduce cot. Diintermediation i another p
oible mean of upply chain improvement. Reducing the number of organization
in the chain can reult in lower cot and fater, more reliable delivery.
Strategic View
Mot importantly, effective SCM require a trategic ytem analyi and deign
. Thi trategic view enable ytem olution rather than individual olution
. A trategic deign will enable ytem integration, allowing for improved deci
ion making acro the organization. Succeful SCM initiative require the ame
thing a Six Sigma, Lean, and the Baldrige criteria: Top management upport an
d collaboration, including time and money. Employee buy-in, including clinician
upport and front-line empowerment. Evaluation of the tructure and taffing of
the upply chain to enure that it will upport the deired improvement and tha
t all relevant function
394
are repreented in a meaningful way. Cro-functional team may be the bet way
to enure thi. Proce analyi and improvement, including a thorough and compl
ete undertanding of exiting ytem, procee, and protocol (through proce
mapping) and their improvement. Collection and analyi of relevant, accurate d
ata and metric to determine area of improvement, mean of improvement, and whe
ther improvement wa achieved. Evaluation of technology-enabled olution in ter
m of both cot and benefit. Training in the ue of new technologie and techn
ique. Thi i eential for broad application and ue in the organization and f
or ucce. Internal awarene program to highlight both the need for and benef
it of trategic SCM. Improved inventory management through better undertanding
of the ytem conequence of unofficial inventory, JIT ytem, and improved
inventory tracking ytem. Enhancement of vendor partnerhip through informati
on haring and the invetigation and determination of mutually beneficial oluti
on. Performance tracking hould be employed to determine the bet vendor to in
volve in thi proce. Finally, the organization mut continually educate and u
pport a ytemwide view of the upply chain and eek improvement for the ytem
rather than individual department or organization in that ytem.
Concluion
In the pat, healthcare organization did not really focu on SCM iue; today,
increaing cot preure are driving them to examine and optimize their upply
chain. The idea and tool preented in thi chapter will help the healthcare
upply chain profeional to achieve thee goal.
Dicuion Quetion
1. Why i SCM important to healthcare organization? 2. Lit ome inventory item
found in your organization. Which of thee might be claified a A, B, or C i
tem? Why? How would you manage thee item differently depending on their cla
ification? 3. Think of an item for which your organization carrie SS. Why i SS
needed for thi item? Can the amount of SS needed be reduced? How?

395
4. Decribe the ERP ytem() found in your organization. How could they be impr
oved?
Chapter Exercie
1. Ue the web to invetigate and ummarize commercially available oftware olu
tion for healthcare organization. 2. Thi problem ue information from a data
et available on the companion web ite. The data et contain the raw data a
well a reduced/reorganized data for eae of analyi. Ue the forecating templ
ate found on the companion web ite or Minitab to forecat total U.S. healthcare
expenditure for 2010 uing SMA, WMA, SES, trend-adjuted (or double) exponenti
al moothing, and linear trend. a. Which model do you believe give the bet for
ecat? b. Do you ee any problem with your model? c. Repeat the above for hopi
tal care, phyician ervice, other profeional ervice, dental ervice, home
health care, precription drug, and other. I any one of thee driving the inc
reae in healthcare expene? 3. The Excel inventory template found on the compan
ion web ite may be helpful for thi problem. An Excel preadheet with data for
the problem can alo be downloaded from ache.org/book/ OpManagement. Hopital
purchaing agent Abby Smith need to order examination glove. Currently, he o
rder 1,000 boxe of glove whenever he think there i a need. Abby ha heard
that there i a better way to do thi and want to ue EOQ to determine how much
to order and when. She collect the following information. Cot of glove: Carr
ying cot: Cot of ordering: Lead time: Annual demand: $4.00/box 33%, or $_____
_____/box $150/order 10 day 10,000 boxe/year
a. What
r than
1,500,
y will
quantity hould Abby be ordering? Prove that your order quantity i bette
Abby by graphing ordering cot, holding cot, and total cot for 1,000,
and 2,000 boxe. b. How often will Abby need to order? About how many da
there be between order?
396
c. Auming that Abby i not worried about SS, when hould he place her order?
Draw another graph to illutrate why he need to place her order at that partic
ular point. d. Abby i concerned that the reorder point he determined i wrong
becaue demand for glove varie. She gather the following uage information:
Period (10 Day Each) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Average
Demand 274 274 284 274 254 264 264 284 274 294 274 284 264 274 274
e. Abby decide he will be happy if the probability of a tockout i 5 percent.
How much SS hould Abby carry? f. If Abby were to et up a two-bin ytem for g
love, how many boxe of glove would be in each bin?
Reference
BearingPoint and National Alliance for Healthcare Information Technology. 2006.
RFID in Healthcare: Poied for Growth. [Online information; retrieved 2/15/07.]
www.bearingpoint.com. Box, G. E. P., and G. M. Jenkin. 1976. Time Serie Analy
i: Forecating and Control, 2nd ed. San Francico: Holden-Day. Harri, F. W. 19
13. How Many Part to Make at Once. Factory, The Magazine of Management 10 (2): 13
5136, 152.

397
Healthcare Financial Management Aociation. 2005. HFMA 2005 Supply Chain Benchm
arking Survey. [Online information; retrieved 8/9/06.] www.hfma.org/ library/acc
ounting/cotcontrol/2005_Supply_Chain_Benchmk.htm. . 2002. Reource Management: The
Healthcare Supply Chain 2002 Survey Reult. [Online information; retrieved 2/1
9/07.] www.hfma.org/library/ accounting/cotcontrol/Supply_Chain_2002_Survey.htm
. . 2001. Reource Management Update: Healthcare Supply Chain. [Online information;
retrieved 2/19/07.] www.hfma.org/library/accounting/ cotcontrol/Update_Healthc
are_Supply_Chain.htm. Knowledge@W.P.Carey. 2006. Tranforming U.S. Health Care: S
upply-Chain Makeover Rejuvenate Medical Center. W. P. Carey School of Buine,
Arizona State Univerity, April 6. [Online information; retrieved 11/5/07.] http
://knowledge.wpcarey.au.edu/index.cfm?fa=viewArticle&id=1223& pecialId=42. Reh
, F. J. 2007. Pareto PrincipleThe 80-20 Rule. [Online information; retrieved 2/15/0
7.] management.about.com/c/generalmanagement/a/ Pareto081202.htm. Wright, C. M.
2007. Where My Defibrillator? More Effectively Tracking Hopital Aet. APICS 17
(1): 2833.
PART
V
putting it all together for operational excellence
CHAPTER
14
HOLDING THE GAINS
CHAPTER OUTLINE
Overview Human Reource Planning Managerial Accounting Control Which Tool to U
e: A General Algorithm A. Iue Formulation B. Strategic or Operational Iue C
. The Balanced Scorecard for Strategic Iue D. Project Management E. Baic Per
formance Improvement Tool F. Quality and Six Sigma G. Lean H. Simulation I. Sup
ply Chain Management J. Holding the Gain Data and Statitic Operational Excell
ence Level 1 Level 2 Level 3 Level 4 Level 5 VVH Strive for Operational Excelle
nce The Healthcare Organization of the Future Concluion Dicuion Quetion Ca
e Study: Vincent Valley Hopital and Health Sytem Reference
400

cot-volume-profit analyi (CVP) diagnoi-related group human reource (HR) p
lanning human reource department (HR) Intitute for Clinical Sytem Improveme
nt managerial accounting
401
Putting It All Together for Operational E xcellence

Overview
Thi chapter conclude and integrate thi book. It include: Three trategie t
o be ued to maintain the gain in operational improvement project: human reou
rce (HR) planning, managerial accounting, and control ytem. An algorithm tha
t ait practitioner in chooing and applying the tool, technique, and meth
od decribed in thi book. An examination of how Vincent Valley Hopital and He
alth Sytem (VVH) ued the tool for operational excellence. An optimized health
care delivery ytem for the future. The preceding chapter preent an integrate
d approach to achieving operational excellence. Firt, trategy execution and ch
ange management ytem need to be well developed. The balanced corecard and fo
rmal project management technique are effective method to employ in thee key
organizational challenge. Quantitative tool can now be applied, uch a tateof-the-art data collection and analyi tool and problem-olving and deciion-m
aking technique. Procee and cheduling ytem can be improved with Six Sigm
a, Lean, and imulation. Supply chain technique will maximize value and minimiz
e cot in operation. The final challenge in achieving healthcare operation ex
cellence i to hold the gain. High energy i uually preent when a new initiat
ive i introduced, at the tart of a large project, or at the beginning of an ef
fort to olve a problem. However, a time pae, new prioritie emerge, team me
mber change, and operation can drift back to unatifactory level. A trategy
for holding the gain mut be developed at the beginning of any operation impr
ovement effort. HR planning, managerial accounting, and control ytem are the
key to maintaining the gain. Although thi book i not focued primarily on HR
or finance, thee function are eential to utaining the improvement achiev
ed. Staff from thee upport department hould be engaged at the beginning of o
peration improvement activitie and invited to be part of project team if po
ible. More extenive information related to thee functional area can be found
in Human Reource in HealthcareManaging for Succe (Fried, Fottler, and Johnon
2005) and Healthcare Finance: An Introduction to Accounting and Financial Manag
ement (Gapenki 2005).
402
Chapter 14: Holding the Gain

403
Human Reource Planning
The effect of many of the project management and proce improvement tool decr
ibed in thi book can be a major change in the work live of a healthcare organi
zation employee. Many of thee change will be in the procee of an employee
work, hopefully making it more productive and fulfilling. Some of the more power
ful tool, uch a Lean, Six Sigma, and imulation, can provide major productivi
ty gainin ome cae, 30 percent to 60 percent increae can be achieved. A clin
ical proce improvement project may ignificantly change the tak that fill an
employee workday. In thi environment, a diciplined plan for employee redeploy
ment or retraining i eential. Many healthcare organization fail at thi crit
ical tep, a they lack procee to capture and maintain gain in productivity
and quality improvement. A part of the executive function of a healthcare organ
ization, the human reource (HR) department erve a a trategic partner in ma
king effective and long-lating change. During each annual planning cycle, trat
egic project to further the goal of the organization are identified. Many of t
hee initiative become part of the balanced corecard. At thi point, the HR de
partment hould be included to undertake planning, placing the right peron in th
e right job at the right time (Jackon and Schuler 1990, 223). Thi proce i h
own in Figure 14.1. The HR taff need to etimate the effect of each project or
initiative that will be undertaken during the year. If the project ha a goal o
f providing more ervice with the ame taff, the HR tak in the future will be
to maintain thi taffing level. Broader HR planning can now occur, uch a trac
king the availability of worker for thee poition in external labor pool or
identifying and training exiting employee to fill thee role if turnover occu
r. If, on the other hand, the likely outcome of a project will be to reduce ta
ff in a department, it i important to be clear to the organization about the ne
xt tep. If unfilled poition are no longer needed, the mot prudent tep i t
o eliminate them. However, if the poition i currently filled, exiting employe
e need to be tranferred to different department in need of fulltime equivalen
t (FTE). If there are no opening in other department, thee employee may be
come part of a pool of employee ued to fill temporary hortage inide the org
anization. Retraining for other open poition i alo an option if the diplace
d employee ha related kill. Becaue they have jut participated in proce im
provement project, thee taff member might alo receive more training in proc
e improvement tool and be aigned to other department to aid in their proje
ct. If none of thee option i poible, the lat action available to the mana
ger i to lay off the employee. It i difficult to execute project that will cl
early reult in job loit i almot impoible to get employee to redeign
404
FIGURE 14.1 Proce for HR Planning
Project identified
Decreae taffing? Ye
No
Plan for maintaining taff
Vacant poition? No FTE needed in other department?
Ye
Eliminate vacant poition
Ye
Retrain and pool or redeploy
No Lay off
themelve out of a job. However, layoff can generally be avoided in healthcare
, a labor hortage are widepread. In addition, mot project identified houl
d be of the firt type, thoe that will increae throughput with exiting taff,
a thee tend to be the mot critical for improved patient acce and increae
in the quality of clinical care. The HR planning function hould be ongoing and
comprehenive. A well-communicated plan for employee reaignment and replaceme
nt hould be in place. By identifying all potential project during the annual p
lanning cycle, the HR department can develop an organization-wide taffing plan.
Without thi critical function, many of the gain in operating improvement wil
l be lot.
Managerial Accounting
The econd key tool for holding the gain i the ue of managerial accounting (G
apenki 2005, Chapter 5). In contrat to financial accounting, which i ued to
prepare financial tatement (the pat), managerial accounting focue on the fu
ture. Managerial accounting can be ued to project the profitability of a partic
ular project that improve patient flow or model the

405
revenue gain from a clinical pay-for-performance (P4P) contract. Even project
that appear to have no financial effect can benefit from managerial accounting.
For example, a project to reduce hopital-acquired infection may not only provi
de higher quality care, but it may alo reduce the length of tay for thee pati
ent and therefore improve the hopital profitability. It i ueful to have a me
mber of the finance taff engaged with operation improvement effort. Thi team
member hould perform an initial analyi of the expected financial reult for
a project and monitor the financial model throughout the project. She hould al
o enure that the financial effect of an individual project flow through to th
e financial reult for the entire organization. Monte Carlo imulation (Chapter
10) can be extremely helpful in evaluating the rik and reward aociated wit
h variou project or deciion. The firt tep in managerial accounting i to u
ndertand an operating unit revenue ource and how it change with a change in o
peration. For example, capitation revenue may flow to a primary care clinic; in
thi cae, a reduction in the volume of ervice will reult in a profitability
gain. However, if the revenue ource for the clinic i fee-for-ervice payment
, the reduction in volume will reult in a revenue lo. Evaluating many revenue
ource in healthcare can be complex. For example, undertanding inpatient hop
ital reimburement via diagnoirelated group can be complicated, a ome diagno
e pay ubtantially more than other. In addition, many other rule affect net
reimburement to the hopital, o a comprehenive analyi need to be undertak
en. The trend toward conumer-directed healthcare and healthcare aving account
mean that the retail price of ome ervice will alo affect net revenue. If
a market-enitive outpatient ervice i priced too high, net revenue may declin
e a conumer demand will be lower. Next, the cot for the operation mut be id
entified and egmented into three categorie: variable, fixed, and overhead. Var
iable cot are thoe that vary with the volume of the ervice; a good example i
upplie ued with a procedure. Fixed cot are thoe that do not vary with vo
lume and include uch item a pace cot and equipment depreciation. Employee
are uually deignated a fixed cot, although they may be variable if the vol
ume of ervice change ubtantially and taffing level are adjuted baed on
volume. The final cot category i overhead, which i allocated to each departme
nt or unit in an organization that generate revenue. Thi allocation pay for c
ot of department that do not generate revenue. Overhead formula are critical
to undertanding the effect of making operational change. For example, an over
head rate baed on a percentage of revenue will have a ubtantially different e
ffect than one baed on the quare footage a department occupie.
406
TABLE 14.1 Managerial Accounting: CVP Analyi
Bae Tet volume Revenue/tet Total revenue Cot Variable cot/unit Fixed cot
Overhead Total cot Profit
Backlogged Proce Improvement Project 1,500 $150 $225,000
Financial Lo Proce Improvement Project 1,050 $150 $157,500
Bae 1,000 $150 $150,000
1,000 $150 $150,000
$38 $85,000 $20,000 $143,000 $7,000
$38 $85,000 $20,000 $162,000 $63,000
$38 $120,00 $20,000 $178,000 ($28,000)
$38 $80,000 $20,000 $139,900 $17,600
The next tep in the managerial accounting proce i to conduct a cot-volume-p
rofit (CVP) analyi. Table 14.1 illutrate a CVP analyi of two outpatient e
rvice at VVH. In the firt cae, the ervice i backlogged and current profit (
bae cae) i $7,000 per year. However, if a proce improvement project i unde
rtaken, the volume can be increaed from 1,000 to 1,500 tet per year. If taff
ing and other fixed cot remain contant, the net profit i increaed to $63,00
0 per year. The econd example how a ituation where the ervice i operating
at an annual lo of $28,000. In thi cae, the proce improvement goal i to r
educe fixed cot (taffing) with a light increae in volume. The reult i a $
40,000 reduction in fixed cot, which yield a profit margin of $17,600. HR plan
ning i critical in a project uch a thi to enure a comfortable tranition fo
r diplaced employee.
Control
The final key to holding the gain i a control ytem. Control ytem have two
major component: meaurement/reporting and monitoring/repone. Chapter 6 dic
ue many tool for data capture and analyi, with an objective of finding and
fixing problem. However, many of the ame tool hould be deployed for continu
ou reporting of the reult of operation

407
improvement project. It i important that data collection ytem for monitorin
g outcome be built into any operation improvement project from the beginning.
Once data collection i underway, reult hould be diplayed both numerically a
nd graphically. The run chart (Chapter 8) i till one of the mot effective too
l to monitor the performance of a proce. Figure 14.2 illutrate a imple run
chart for birthing center patient atifaction, where a goal of greater than 90
percent atified patient ha been agreed upon. Thi type of chart can how pr
ogre over time to enure that the organization i moving toward it goal. In
addition to a robut data capture and reporting ytem, a plan for monitoring an
d repone i critical. Thi plan hould include identification of the individua
l or team reponible for the operation and a method for communicating the repor
t to them. In ome cae, thee operation improvement activitie are of uch
trategic importance that they may become part of a departmental or organizationwide balanced corecard. A repone plan hould alo be part of the ongoing cont
rol ytem. A procedure or plan hould be developed to addre ituation in whi
ch a proce fail to perform a it hould. Jidoka and andon ytem (Chapter 9)
can help organization to dicover and correct intance where there i a probl
em with ytem performance. Control chart (Chapter 8) can be ued to identify o
ut-of control ituation. Once an out-of-control ituation i identified,
100
Percentage atified or better
90
FIGURE 14.2 Run Chart for Birthing Center Patient Satifaction
80
70 Facilitie 60 Clinical quality High touch 50
t
o No ber ve m be De r ce m be Ja r nu a Fe ry br ua ry M ar ch
ril
ly
r
ay M
m be
Au
Se p
Oc t
te
Month
Ap
Ju n
Ju
gu
e
408
action hould be taken to determine the pecial or aignable caue and eliminat
e it.
Which Tool to Ue: A General Algorithm
Thi book preent an array of technique, tool, and method to achieve operati
onal excellence. How doe the practitioner chooe from thi broad array? A in c
linical care, there i a mix of art and cience in chooing the bet approach. A
general algorithm for electing tool i preented below. (ache.org/book/OpMa
nagement contain an automated and more detailed verion of thi algorithm.) The
general logic dicuion at the beginning of thi chapter ha been expanded to
provide a more finely detailed path through the logic (Figure 14.3).
A. Iue Formulation
Firt, formulate the iue you wih to addre. Determine the current tate and
a deired tate (e.g., competitor have taken 5 percent of our market hare in o
btetric and we want to recapture the market, the pediatric clinic lot $100,00
0 lat year and we want to break even next year, public ranking for our diabete
care put our clinic below the median and we want to be in the top quartile).
It i important to frame the problem correctly to enure that the outcome i the
right anwer to the right quetion rather than the right anwer to the wrong qu
etion. In particular, all relevant takeholder hould be conulted. A number o
f effective deciion-making and problem-olving tool can be ued to: Frame the
quetion or problem; Analyze the problem and variou olution to the problem; a
nd Implement thoe olution. The tool and technique identified next provide a
bai for tackling difficult, complicated problem. The deciion-making proce
: a generic deciion proce ued for any type of proce improvement or problem
olving. (Plan-do-check-act [PDCA], define-meaure-analyze-improve-control [DMA
IC], and project management all follow thi ame baic outline.) -Framing: ued
to enure that the correct problem or iue i actually being addreed. -Gatherin
g intelligence: find and organize the information needed to addre the iue (d
ata collection).
FIGURE 14.3
Develop and ue a balanced corecard (4) Ue formal project management approach
(5) Ye Data collection and analyi (7)
Algorithm for Ue of the Tool, Technique, and Methodologie in Thi Book
Start
Iue formulation (6)
Ye
Ye
Data collection and analyi i ued in conjuction with a number of operation i
mprovement tool, and their ue i indicated by thi box
I the iue trategic? No No Focu on what if iue?
Focu on improving quality or reducing variation? Focu on reducing wate or imp
roving flow? Ye Ye Focu on cheduling? Data Ue imulation (10) Data No Hold
the gain (14) No Ue Lean tool (9) Data Ye Scheduling (12) Ye No
No
Focu on upply chain
Large project? Ue Six Sigma tool (8) Data
Ye
No
Ue SCM (13)
Apply baic proce improvement tool (11)
Data
End
Model created in Microoft Viio. NOTE: Chapter number are in parenthee.
409
410
-Coming to concluion: determine the olution to the problem (data analye). -Le
arning from feedback: enure that learning i not lot and that the olution act
ually work (holding the gain). Mapping tool. -Mind mapping: ued to help form
ulate and undertand the problem or iue. -Proce mapping, activity mapping, a
nd ervice blueprinting: ued to picture the ytem and proce tep. Root-caue
analyi (RCA) tool. -Five why technique and fihbone diagram: ued to identi
fy caue and root caue of problem to determine how to eliminate thoe proble
m. -Failure mode and effect analyi (FMEA): a more detailed root cauetype ana
lyi that can be ued to identify and plan for both poible and actual failure
.
B. Strategic or Operational Iue
Next decide whether the iue i trategic (e.g., major reource and highlevel
taff will be involved) or part of ongoing operation. If the iue i trategic
, go to tep C; if it i operational, go to tep D or E, depending on the ize a
nd cope of poible olution.
C. The Balanced Scorecard for Strategic Iue
To effectively implement a major trategy, develop a balanced corecard to link
initiative and meaure progre. Element of the balanced corecard will includ
e: Strategy map: ued to link initiative or project to achieve the deired ta
te Four perpective: initiative and project that pan the four main perpecti
ve of the balanced corecardfinancial, cutomer/patient, operation, and employe
e learning and growth Metric: ued to meaure progre; include both leading (
predictive) and lagging (reult) meaure If the balanced corecard contain a
major initiative, go to tep D, project management; otherwie, go to tep E, ba
ic performance improvement tool.
D. Project Management
The formal project management methodology hould be ued for initiative that ty
pically lat longer than ix month and involve a project team. Project manageme
nt include thee tool:

411
Project charter: a document that outline takeholder, project ponor, project
miion and cope, a change proce, expected reult, and etimated reource
required. Work breakdown tructure: a liting of tak that will be undertaken t
o accomplih the project goal, with aigned reponibilitie and etimated dur
ation and cot. Schedule: linking of tak in precedence and relationhip and
identification of the critical path that determine the overall duration of the
project. Change control: a method to formally monitor progre and make change
during the execution of a project. Rik management: an identification of project
rik and plan to mitigate each rik. If the project i primarily concerned wi
th improving quality or reducing variation, ue the project management technique
and tool decribed in tep G, quality and Six Sigma. If the operating iue i
large enough for project management and primarily concerned with eliminating wa
te or improving flow, go to tep H, Lean. If the iue i related to evaluating
and managing rik or analyzing and improving flow, go to tep I, imulation. If
the project i focued on upply chain iue, go to tep J, upply chain manag
ement (SCM). If the project focu i not encompaed by Six Sigma, Lean, imulat
ion, or SCM, return to tep E and ue the baic performance improvement tool wi
thin the larger project management ytem.
E. Baic Performance Improvement Tool
Baic performance improvement tool are ued to improve and optimize a proce.
In addition to RCA, the following tool can be helpful in moving toward more eff
ective and efficient procee and ytem: Optimization uing linear programmin
g: ued to determine the optimal allocation of carce reource. Theory of cont
raint (TOC): ued to identify and manage contraint in the ytem. The TOC tec
hnique conit of five tep: 1. Identify the contraint (or bottleneck). 2. Ex
ploit the contraint: determine how to get the maximum performance out of the co
ntraint without major ytem change or capital improvement. 3. Subordinate ev
erything ele to the contraint: other nonbottleneck reource (or tep in the p
roce) hould be ynchronized to match the output of the contraint. 4. Elevate
the contraint: do omething (capital expenditure, taffing increae, etc.) to
increae the capacity of the contraining reource until it i no longer the con
traint. Something ele will be the new contraint. 5. Repeat the proce for th
e new contraint.
412
Force field analyi: ued to identify and manage the force working for and aga
int change (applicable to any change initiative, including TOC, Six Sigma, and
Lean). If thee tool provide an optimal olution, go tep J, holding the gain.
However, ometime the operating iue are o large that they will benefit fro
m the formal project management dicipline. In thi cae, go to tep D. If the p
roject i relatively mall and focued on eliminating wate, go to tep G, Lean,
where the kaizen event tool can be ued for quick improvement.
F. Quality and Six Sigma
The focu of quality initiative and the Six Sigma methodology i on improving q
uality, eliminating error, and reducing variation. DMAIC: the proce improveme
nt or problem-olving technique ued in Six Sigma. The DMAIC technique conit
of five tep: 1. Define: define the problem or proce (ee tep A, iue formu
lation). 2. Meaure: determine the current tate of the proce (ee the ection
Data and Statitic). 3. Analyze: analyze the collected data to determine how t
o fix the problem or improve the proce. 4. Improve: make change to improve th
e proce or olve the problem. 5. Control: enure that change are embedded in
the ytem (ee tep J, holding the gain). Note that at any point in the proce
it may be neceary to loop back to a previou tep. Once the proce i compl
ete, tart around the loop again. Seven baic quality tool: ued in the DMAIC p
roce to improve the proce or olve the problem. The baic quality tool are:
1. Fihbone diagram: ued for analyzing and illutrating the root caue of an
effect. 2. Check heet: a imple form ued to collect data. 3. Hitogram: a grap
h ued to how frequency ditribution. 4. Pareto chart: a orted hitogram. 5.
Flowchart: a proce map. 6. Scatter plot: a graphic technique to analyze the re
lationhip between two variable. 7. Run chart: a plot of a proce characterit
ic in chronological equence. Statitical proce control: an ongoing meauremen
t of proce output characteritic to enure quality. Enable the identificatio
n of a problem ituation before an error occur.

413
Proce capability: a meaure of whether a proce i actually capable of produc
ing the deired output. Benchmarking: the determination of what i poible bae
d on what other are doing. Ued for comparion purpoe and goal etting. Quali
ty function deployment: ued to match cutomer requirement (voice of the cutom
er) with proce capabilitie, given that trade-off mut be made. Poka-yoke: mi
take-proofing. Once thee tool have produced atifactory reult, proceed to
tep J, holding the gain.
G. Lean
Lean initiative are typically focued on eliminating wate and improving flow i
n the ytem or proce. Kaizen philoophy: the proce improvement technique u
ed in Lean. The kaizen technique conit of the following tep: 1. Specify val
ue: identify activitie that provide value from the cutomer perpective. 2. Map
and improve the value tream: determine the equence of activitie or the curre
nt tate of the proce and the deired future tate. Eliminate nonvalue-added t
ep and other wate. 3. Flow: enable the proce to flow a moothly and quickly
a poible. 4. Pull: enable the cutomer to pull product or ervice. 5. Perf
ection: repeat the cycle to enure a focu on continuou improvement. Value tre
am mapping: ued to define the proce and determine where wate i occurring. T
akt time: a meaure of time needed for the proce baed on cutomer demand. Thr
oughput time: a meaure of the actual time needed in the proce. Five S: a tec
hnique to organize the workplace. Spaghetti diagram: a mapping technique to how
the movement of cutomer (patient), worker, equipment, and o on. Kaizen bli
tz or event: ued to improve the proce quickly, when project management i not
needed. Standardized work: written documentation of the precie way in which ev
ery tep in a proce hould be performed. A way to enure that thing are done
the ame way every time in an efficient manner. Jidoka and andon: technique or
tool ued to enure that thing are done right the firt time to catch and correc
t error. Kanban: cheduling tool ued to pull rather than puh work.
414
Single-minute exchange of die: a technique to increae the peed of changeover.
Heijunka: leveling production (or workload) o that the ytem or proce can fl
ow without interruption. Once thee tool have produced atifactory reult, pr
oceed to tep J, holding the gain.
H. Simulation
Simulation i ued to evaluate what if ituation. Uually it i le expenive or
peedier than changing the real ytem and evaluating the effect of thoe chan
ge. The imulation proce conit of the following tep: 1. Model developmen
t: develop a model of the proce or ituation of interet. 2. Model validation:
enure that the model accurately repreent reality. 3. Simulate and analyze ou
tput: run the imulation and analyze the output to determine the anwer to the
quetion aked, optimize the proce, or manage rik. Monte Carlo imulation: t
ypically ued to evaluate and manage the rik aociated with variou deciion
baed on random variable. Dicrete event imulation: baed on queueing theory.
Ued to model ytem flow to improve the ytem. Once thee tool have produce
d atifactory reult, proceed to tep J, holding the gain.
I. Supply Chain Management
SCM focue on all of the procee involved in getting upplie and equipment f
rom the manufacturer to ue in patient care area. SCM i the management of all
activitie and procee related to both uptream vendor and downtream cutome
r in the value chain. Effective and efficient management of the upply chain re
quire an undertanding of all of the following: Tool for tracking and managing
inventory; Forecating; Inventory model; Inventory ytem; Procurement and ve
ndor relationhip management; and Strategic SCM.
Once thee tool have produced atifactory reult, proceed to tep J, holding
the gain.

415
J. Holding the Gain
Once ucceful operational improvement have been completed, three tool can be
ued to enure that thee change will endure: 1. HR planning: a plan to ue em
ployee in new way after an improvement project i completed. 2. Managerial acc
ounting: a tudy of the expected financial conequence and gain after an opera
tion improvement project ha been implemented. 3. Control ytem: a et of too
l to monitor the performance of a new proce and method to take corrective ac
tion if deired reult are not achieved.
Data and Statitic
All of the aforementioned tool, technique, and methodologie require data and
data analyi. Tool and technique aociated with data collection and analyi
include the following: Data collection technique: ued to enure that valid da
ta are collected for further analyi. Graphic diplay of data: ued to ee the da
ta. Mathematic decription of data: ued to compare et of data and for imula
tion. Statitical tet: ued to determine whether difference in data exit. Re
greion analye: ued to invetigate and define relationhip among variable.
Forecating: ued to predict future value of random variable.
Operational Excellence
Many leading hopital, medical group, and health plan are uing the tool and
technique contained in thi book. Unfortunately, thee tool have not een wid
epread ue in healthcare, nor have they been a comprehenively applied a in o
ther ector of the economy. The author have developed a cale for the applicat
ion of thee tool to gauge progre toward comprehenive operational excellence
in healthcare.
Level 1
There are no organized operation monitoring or improvement effort at level 1.
Quality effort are aimed at compliance and the ubmiion of data to regulating
agencie.
416
Level 2
At level 2, the organization ha begun to ue operation data for deciion makin
g. There are pocket of proce improvement activitie where proce mapping and
PDCA or rapid prototyping are employed. Evidence-baed medicine (EBM) guideline
are ued in ome clinical activitie.
Level 3
Senior management ha identified operation improvement effort a a priority by
level 3. The organization conduct operation improvement experiment, ue a d
iciplined project management methodology, and maintain a comprehenive balance
d corecard. Some P4P bonue are received, and the organization obtain above-a
verage core on publicly reported quality meaure.
Level 4
A level-4 organization engage in multiple proce improvement effort uing a c
ombination of project management, Six Sigma, Lean, and imulation tool. It ha
trained a ignificant number of employee in the advanced ue of thee tool, an
d thee individual lead proce improvement project. EBM guideline are compre
henively ued, and all P4P bonue are achieved.
Level 5
Operational excellence i the primary trategic objective of an organization at
level 5. Operation improvement effort are underway in all department, led by
departmental taff who have been trained in advanced tool. The organization ue
real-time imulation to control patient flow and operation. New EBM guideline
and bet practice for adminitrative operation are developed and publihed b
y thi organization, which core in the top 5 percent of any national ranking o
n quality and operational excellence. A few leading organization currently are
at level 4, but mot reide between level 2 and 3. Our friend at VVH are at th
e top of 3 and moving toward 4.
VVH Strive for Operational Excellence
A can be recalled from Chapter 3, VVH leaderhip wa encouraged
participate in the Bridge to Excellence (BTE) program. Becaue
team believed that P4P wa an important new trend, they added an
heir corporate balanced corecard: Conduct project to optimize
at leat a 5 percent increae in revenue. VVH
by it board to
the leaderhip
initiative to t
P4P and generate

417
reorganized it tructure to combine a number of operation and quality activiti
e into a new organization-wide department known a operation management and qu
ality. The BTE program i employer ponored and provide participating medical
group with up to $80 a year per patient with diabete if they are top performer
in diabete care. The pecific goal VVH choe to achieve were: Hemoglobin A1c
of le than 7 percent Low-denity lipoprotein choleterol lower than 100 mg/dL
Blood preure lower than 130/80 mm Hg Nonmoking tatu Apirin for patient o
ver age 40
The firt tep in the project wa to identify a team and develop a project chart
er and chedule (Chapter 5). Both the HR and finance department were included i
n the project team to model financial conequence (new revenue, poible new c
ot, capital requirement) and the potential effect on taffing level. The pro
ject team began by collecting data on current performance and ummarizing them u
ing viual and mathematic technique to determine where performance wa not mee
ting goal (Chapter 7). A proce map wa contructed and compared with Intitut
e for Clinical Sytem Improvement guideline to determine where procee could
be improved to achieve the deired reult. Variou Six Sigma tool (fihbone d
iagram, check heet, Pareto chart, and catter plot) were employed to furthe
r analyze and improve the proce (Chapter 8). The clinician on the project tea
m performed a careful analyi to determine which area of the treatment of pati
ent with diabete could be tandard and which needed cutomization. The tandar
d module were then examined for both effectivene and efficiency uing value
tream mapping (Chapter 9). Patient flow for the tandard care module wa alo m
odeled and entered into a imulation model to tet variou patient movement opti
on (Chapter 10). Change were identified, many of them requiring either a taff
ing change or a change in VVH electronic medical record. Becaue many option we
re available and it wa not clear which would achieve the deired reult, a dec
iion tree (Chapter 6) wa contructed to identify the optimal proce improveme
nt. Finally, once the project team began to implement thee proce improvement
, the reult were monitored with control chart (Chapter 7). After ix month,
VVH had realized a gain of 7 percent in revenue from thi project.
418
The Healthcare Organization of the Future
A future healthcare organization operating at level 5 i illutrated in Figure 1
4.4. Thi care delivery ytem would ue many of the tool and technique contai
ned in thi text. A demand prediction model (Chapter 13) would generate predicti
on of demand for inpatient and ambulatory care ervice. Becaue much of the ca
re delivered in thee ite would be through the ue of EBM guideline (Chapter
3) that have optimized procee (Chapter 7 to 9), the reource requirement co
uld be predicted a well; thee prediction would drive cheduling and upply ch
ain ytem. A key component of thi future ytem i a real-time operation mon
itoring and control ytem. Thi ytem would ue the imulation and modeling te
chnique decribed in Chapter 10 to monitor, control, and optimize patient flow
and diagnotic and treatment reource. Macro-level control ytem uch a the
balanced corecard (Chapter 4) would enure that thi ytem meet the organizat
ion trategic objective. The reult would be a finely tuned healthcare delivery
ytem providing high-quality clinical care in the mot efficient manner poib
le.
FIGURE 14.4 An Optimized Healthcare Delivery Sytem of the Future
Demand prediction ytem: Volumeclinical condition
Ambulatory care model EBM baed
Emergency and inpatient care modelEBM baed
Predicted reource need: Facilitie Staff Supplie
Staff cheduling ytem Supply chain ytem
Real-time operation monitoring and control
Clinical operation
Real-time control Real-time data

419
Concluion
We hope that thi text i helpful to you and your organization on your journey t
oward level 5. We are intereted in your progre whether you are a new member o
f the health adminitration team, a eaoned department head, or a phyician lea
derpleae ue the e-mail addree on the companion web ite to inform u of your
uccee and let u know what we could do to make thi a better text. Becaue
many of the tool dicued in thi text are evolving, we will continuouly upda
te the companion web ite at ache.org/book/OpManagement with reviion and add
ition; check there frequently. We too are triving to be level 5.
Dicuion Quetion
1. Identify method to reduce employee reitance to change during an operation
improvement project. 2. What hould be the key financial performance indicator
ued to analyze performance change for hopital? Clinic? Health plan? Public
health agencie? 3. Decribe tool (other than control chart) that can be ued
to enure that procee achieve their deired reult. 4. Decribe the tool,
method, and technique in thi book that would be ued to addre the following
operating iue: a. A hopital laboratory department provide reult late and
frequently erroneouly. b. A clinic web-baed patient information ytem i not
being ued by the expected number of patient. c. An ambulatory clinic i finan
cially challenged but ha a low taffing ratio compared with imilar clinic.
Cae Study: VVH
VVH ha a eriou problem: A major trategic objective of the health ytem wa
to grow it ambulatory care network, but the organization i facing a number of
challenge. Although a new billing ytem wa intalled and variou reimburemen
t-maximization trategie were executed, total cot in the ytem exceed revenu
e. Thi wa occurring even though the clinic taff felt buy and backlog appoint
ment were growing. Analyi of clinic data
420
alo indicated an increaing number of patient were canceling appointment or w
ere no how. In addition, a new group of multipecialty and primary care phyicia
n had been created from the merger of three eparate group; thi clinic i agg
reively competing with VVH for privately inured patient. The new large clini
c i making ame-day clinic appointment available and heavily advertiing them.
The board of VVH aked the chief executive officer (CEO) to develop a plan to a
ddre thi growing concern. The CEO firt formed a mall trategy team to lead
improvement effort; it firt tep wa to aign the chief operating officer, c
hief financial officer, and medical director to direct the planning and finance
taff on the improvement team. VVH ultimately decided that it needed to increae
the number of patient een by clinician and begin to implement advanced acce
in it clinic. Becaue VVH believe in knowledge-baed management and haring
in improved method of delivering health ervice, the organization ha made it
data and information available on the companion web ite. VVH invited tudent
and practitioner to help them improve thi ytem. 1. Frame the original iue
for VVH. Mind map and RCA may be ueful here. 2. How would you addre the nohow and cancellation iue? 3. Develop a project charter for one project ao
ciated with VVH problem. 4. Develop a balanced corecard for VVH clinic. 5. If
VVH decided that it hould focu on increaing throughput in the ytem, how wo
uld you go about doing thi? Be pecific. 6. Would imulation be a ueful tool i
n VVH quet to implement advanced acce? Why or why not? How would you go about
implementing advanced acce? Be pecific.

Health Maxhxchxhnagement

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Health Maxhxchxhnagement

Încărcat de

Drepturi de autor:

Formate disponibile

Preface

3. Performance improvement tools, techniques, and programs; 4. Applications to c

KEY TERMS AND ACRONYMS

Introduction to Healthcare Operations

Chapter 1: The Challenge and the Opportunity

Chapter 1: The Challenge and the Opportunity

Chapter 1: The Challenge and the Opportunity

Chapter 1: The Challenge and the Opportunity

Compare actual to needed staff based on patient demand

Chapter 1: The Challenge and the Opportunity

Chapter 1: The Challenge and the Opportunity

Chapter 1: The Challenge and the Opportunity

KEY TERMS AND ACRONYMS

Introduction to Healthcare Operations

KEY TERMS AND ACRONYMS

Introduction to Healthcare Operations

an examination of how payers are using the

Conclusion Discussion Questions Chapter Exercise

Setting Goals and E xecuting Strategy

In a study on the use of the balanced scorecard in healthcare delivery organizat

Employee motivation A progressive culture and motivated employees are clearly co

organization, those that do not

Financial To financially sustain our mission, on what must we focus?

KEY TERMS AND ACRONYMS

Setting Goals and E xecuting Strategy

KEY TERMS AND ACRONYMS

Performance Improvement Tools, Techniques, and Programs

Design of Experiments Conclusion

Performance Improvement Tool, Technique, and Program

Hitogram and Pareto Chart

hape occurred o frequently that people began to expect it and it came to be ca

Triangular ditribution The triangular ditribution i often ued in modeling an

1.0 1.2 1.6 1.4 0.2 3.0

0.6 1.7 1.0 0.7 0.6 0

2 0.36 2.89 1.00 0.49 0.36 5.10

it would e importnt to rememer tht the lower the vlue for th

KEY TERMS AND ACRONYMS

Performance Improvement Tool, Technique, and Program

USL LSL USL LSL and is estimated by C  = 6s 6 x LSL USL x = min or 3 3

Riverview Clinic Proce Capability

KEY TERMS AND ACRONYMS

Performance Improvement Tool, Technique, and Program

# % % hr % hr hr 1 5 hr 20 80 hr

upplie and equipment wa reduced to approximately 20 minute in an eight-hour

KEY TERMS AND ACRONYMS

Performance Improvement Toos, Techniques, and Programs

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simulation

$ (17,032.68) $ (15,443.75) $ (26,655.61) $ (65,788.80) $ (46,067.85)

Chapte 10: Simulation

Chapter 10: Simuation

Chapter 10: Simuation

3 = = 3 Patients or 43 Ls = Ws = Arriva rate Time in the system = 3 Patients/Hour

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simuation

KEY TERMS AND ACRONYMS

Appications to Contemporary Heathcare Operations Issues

KEY TERMS AND ACRONYMS

USL LSL USL LSL and is estimated by C = 6s 6 x LSL USL x = min or 3 3

Performance Improvement Toos, Techniques, and Programs

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simuation

3 = = 3 Patients or 43 Ls = Ws = Arriva rate Time in the system = 3 Patients/Hour

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simuation

Chapter 10: Simuation

Appications to Contemporary Heathcare Operations Issues

Appications to Contemporary Heathcare Operations Issues

Chapter 12: Scheduing and Capacity Management

Chapter 12: Scheduing and Capacity Management

Chapter 12: Scheduing and Capacity Management

Chapter 12: Scheduing and Capacity Management

Chapter 12: Scheduing and Capacity Management

Chapter 12: Scheduing and Capacity Management

Chapter 12: Scheduing and Capacity Management

Chapter 12: Scheduing and Capacity Management

Chapter 12: Scheduing and Capacity Management

Chapter 12: Scheduing and Capacity Management

moel assumes that eman varies an is normally istribute (Chapter 7)