Sunteți pe pagina 1din 9

Optimal queue length for orthopaedic surgery with

surgeon-specific queues and maximum waiting time

Antti Peltokorpi1, Juha-Matti Lehtonen2, Paulus Torkki1, Teemu Moilanen3

Helsinki University of Technology, Institute of Healthcare Engineering, Management and
Architecture, Espoo (Finland),,
Tampere University of Technology, Department of Industrial Management, Tampere (Finland),
Coxa Ltd, Hospital for Joint Replacement, Tampere (Finland),

Abstract: Lengthened waiting times to surgical operations lead to a diminished patient

satisfaction, deteriorated medical quality and increased costs. In order to avoid long queues,
authorities have enacted care guarantees that define the maximum waiting time to services. In this
study, the effect of queue length and care guarantee of maximum waiting time on OR productivity
is discussed from a case perspective of an orthopaedic surgery hospital. With 6 months care
guarantee the optimal average queue length was 3 months. This is only few weeks longer than
patient preferences. With shorter queues there existed problems in finding appropriate cases to
fulfill the OR session. With longer queues cases exceeding the care guarantee disturbed the
optimal scheduling. The study indicates that the introduction of care guarantee might deteriorate
the overall efficiency and increase the average waiting time. More sophisticated scheduling
algorithms are needed to maintain efficiency with care guarantees.

1 Introduction
Lengthened waiting times tend to be a problem in many public-funded health care systems. For
example in surgical operations long waiting times lead to a diminished patient satisfaction,
deteriorated medical quality and increased care costs (Kujala et al. 2006). Most of those
disadvantages burden primarily patients and her family and employer. Service provider might even
benefit from long waiting times if there is no real competition or financing is budget-based and not
fee for service.
In order to avoid long queues, authorities in countries with public-funded system have
implemented several policies to cut waiting times. Enacting care guarantee that defines the
maximum waiting time to services is one of the used policies (Hanning 1995). However, the effect
of those policies on the production efficiency is weakly understood.
From a service provider perspective, some queues for surgery are needed to maximize the use of
fixed operating room (OR) capacity. Based on basics of queuing theory, there exists always a
trade-off between short waiting times and high utilization rates. Optimization of OR efficiency
becomes more problematic when surgeons have own patients and queues. In this study, the effect
of queue length and care guarantee of maximum waiting time on OR productivity was analyzed
and discussed.

2 Managing waiting times

Deteriorating effects of waiting times have been considered in several studies. Coyte et al. (1994)
argued that from a patient perspective, optimal waiting time to elective surgery is between three to
eight weeks. Oudhoff et al. (2007) expanded the acceptable range between 2 and 25 weeks
depending on the type of disorder and the severity of physical and psychosocial problems of
patients. Anyway, those time frames are nowadays exceeded in many public-funded hospitals.
Long waiting time affects negatively on patient’s quality of life and recovery process (e.g.

Sampalis et al. 2001): Longer waits before surgery are associated with an increased likelihood of
not returning to work after surgery. Even if the waiting time has no effect on the final outcome,
during waiting time patients suffered most from restrictions in psychological well-being such as
depression, distress and reduced vitality (Hirvonen 2007).
While it is sometimes argued that long waiting times are maintained due to an efficient use of
resources, studies reveal that longer waits cause also additional care costs and administrative costs
for service provider (Kujala et al. 2006). E.g. in Finland hospitals have appointed particular queue
managers and nurses for patient prioritization and serving customers that ask about their care. The
role of waiting times as a limiting factor for demand in a free service is also questionable. The UK
study shows that cutting waiting lists did not generate more demand, and that consultant behaviour
is a major variable (Dawsin et al. 2006).
Public authorities’ responses to long waiting times can be divided to two categories: (1) Patient
choice -policy aims at increasing the competition among hospitals and improving quality and
access to care. E.g. in Norway patient can choose the service provider and waiting times for
surgery are seen online (Johnsen 2006). Waiting times for hip endoprosthesis vary between 3-42
weeks. UK has recently implemented quite similar policy where after the first referral patients can
choose where they have the treatment (NHS 2009).
Another policy for shortening waiting times is called (2) care guarantee. In that policy, the first
aim is not to increase competition, but enforce hospitals to care patients in a defined time frame. In
Finland, the maximum waiting time for surgery is 6 months after decision for surgery (FINLEX
2004). Similar maximum time in Sweden is 3 months (SKL 2005) and in Denmark 2 months (MIH
2002). In UK the care guarantee is about 4 months (DH 2009).
There exists also critique that fixed single number care guarantees are not optimal, because they do
not prioritize different patients and illnesses (Jacobson 2004). In Australia and New Zealand each
surgery type has several urgency rates with specific maximum waiting time (e.g. ACT Health
2007; MH 2009). Maximum waiting time for elective surgery is 12 months in Australia and 6
months in New Zealand. Norway has also modified the system toward more patient-specific care
guarantee (Hurst & Siciliani 2003).
It seems that care guarantee -policy is more widely used in public-funded systems than the patient
choice -policy. However, some countries, such as Canada and Denmark, have implemented a
mixed policy, where the citizens may choose among private hospitals or clinics if the waiting time
for treatment exceeds the maximum time (HE 2007; MIH 2002).
From a service provider point of view, care guarantee motivates them to keep the queue shorter
and care immediately patient that is in the risk to exceed the maximum time (Dawson et al. 2007).
In practice this means that sometimes patient, that is not optimal for a daily OR list, have to be
scheduled first. This might deteriorate the daily OR efficiency, measured by the number of
surgeries per OR day, but maximizes the overall efficiency in the hospital due to saved revenue.
The optimal queue length lies somewhere between zero and the care guarantee.

3 Objectives and research process

3.1 Study objectives

The literature shows that care guarantees are mainly implemented due to equality objectives. This
study aims at considering the care guarantees from service provider point of view. The objectives
of this study are:
• Objective 1: What is the effect of a care guarantee length on OR efficiency?

• Objective 2: With a given care guarantee, what is the optimal queue length with respect to
OR efficiency?

3.2 Case description

The research is based on a situation of a Finnish orthopaedic specialist centre performing only
joint replacement surgery as a limited company. The hospital has five ORs, an outpatient clinic
and two wards, performing annually about 2600 surgical arthroplastic operations (Coxa 2008).
The nurses are allocated separately to specific units (outpatient clinic, ward and operating unit).
The anaesthesiologists are mainly allocated to the operating unit and post-anaesthesia care unit.
The orthopaedic surgeons are typically three days per week in the operating unit and two days per
week in the outpatient clinic making treatment decisions and carrying out follow-up visits.
The patient’s care path in hospital was modeled by collecting the existing process charts and
interviewing the personnel. Data for the analysis and simulation model included all actual
operations between 2.1.2006-18.2.2008 (5726 cases) from the patient information system
databases. Patient information included age, diagnosis and procedures. Resource information
included the surgeon, anaesthesiologists and nurses connected with an operation. Timestamps
collected were Admission Time, Patient in Operating Unit, Patient in OR, Surgery Start Time,
Surgery Finish Time, Patient out of OR, Patient out of PACU, Discharge Time (Donham et al.

3.3 Care path at case hospital

Figure 1: The main phases of the patient’s care path in hospital

Based on a referral, the patient is called to an orthopaedic surgeon examination. In the case of
surgery, with few exceptions the same surgeon that has met the patient will perform the operation
– so at this stage it is decided which surgeon will perform the operation. The orthopaedic surgeon
places the patient in the queue for operation and provides the case duration category and planned
operation type for the queue scheduler. Most patients will have a preoperative visit during which
the patient meets a nurse, a physiotherapist and an anaesthesiologist. If a long time (> 3 months)
has elapsed since the decision for surgery, there may also be a need to check again with the
orthopaedic surgeon. The queue scheduler schedules the preoperative examinations and the
operation. The queue scheduler uses case duration categories to construct the master surgery
schedule: i.e. allocates patients to ORs on specific days. At the time of the study all cases were
classified into four case duration categories by the surgeon. These categories refer to operation
time “120” (1/3 day), “180” (½ day), “240” (2/3 day) and a “360” (full day) operations. Fridays
are shorter (330) working minutes, so special arrangements are needed. Patients are admitted

mainly on the day of surgery and are discharged and transferred typically (80%) to a primary care
ward after 2-3 postoperative days. A control visit will take place routinely two months after the
operation with the orthopaedic surgeon.
Actual performance in the period 2.1.2006-18.2.2008 was 2.11 cases per day and OR. As there
were a small number of shorter days taken up by education and other issues, the figure is close to
but not exactly that achieved by regular operations. The actual waiting times are not
straightforward to calculate based on the data, as the data system did not have a care path ID nor a
date when the patient was assigned to the queue. There is a unique patient ID, but as there may be
several outpatient clinic visits by one patient as well one patient may have more than one
operation, it does not enable waiting time calculation. After filtering only cases with one operation
and filtering out those whose first visit in the data was after operation, there remained 1727 cases.
Their average waiting time was 98 days and there were some variation between surgeons, an
average of 121 days being the longest and 65 days the shortest of 97 and 142 cases, respectively.

3.4 Model description

The development process of the simulation model followed a Plan-Do-Check-Act-cycle (Deming
1986). The model consists of each surgeon’s patient queue, the surgeon schedule, planned
schedule (Table 1) according to case categories, the algorithm for searching cases to fit the
planned schedule for each day and OR (i.e. surgeon), with additional procedures for overdue cases
and for finding a substitute schedule when cases for carrying out the planned schedule are not
available in the queue and, finally, a calculation of actual case duration and day length in the OR.
Care guarantee applies to waiting time in surgeon’s patient queue. Calculation of the waiting time
for surgery begins from the day when patient was placed in the queue for operation by the
orthopaedic surgeon, and ends on the day of surgery by the same surgeon. Each week when the
weekly schedule is made, the model checks if there are cases that have been waiting for the
surgery longer than the guaranteed waiting time. Patients that had waited longer than a care
guarantee were scheduled to a first open slot regardless of the effect on the total productivity. In
order to study the effect of different care guarantees, the length of care guarantee of 3, 6 and 12
months were used in the model.
Actual queue lengths i.e. average waiting times are expected to affect the productivity and
therefore included as a variable in the simulations. Even though queue length in reality varies
when entry to the queue is random, in the model the number of patients in the queue was fixed in
order to achieve control over this variable. The values of queue length in the simulations were
limited by the care guarantee, because it makes no sense to use longer waiting times (i.e. queue
lengths) than the care guarantee.
In a previous study (Lehtonen et al. 2009) the objective was to increase productivity. As one such
method a new approach for making the master surgery schedule was developed. Table 1 shows the
planned surgery schedules in a recurring four-week-period used in the simulation and last column
shows schedule performance in cases per OR per day when schedule is adhered to perfectly. Two
new surgery duration classes were formed: (a) a new category “210” of fastest cases from the
“240” category with the idea of attempting to perform two such cases in a day and; (b) forming a
new category “90” based on average surgery time and surgery combinations from the “120 with
the idea of attempting to perform four such cases in a day and simultaneously increasing the day
length for by 30 min for such four-case days. This new approach to master surgery schedule
development increased productivity and therefore is included as a co-variable in studying the
productivity effects of care guarantee in the case.
Surgery 120+ 3x 180+ 120+ 360 180+ 120+ 210+ 4x 360+ 210+ ORs Cases/
Schedule 240 120 180 180 240 210 210 90 120 90 /week OR
“Base” 44 20 9 15 5 23.25 2.16

“All” 36 2 1 27 7 2 4 1 20 2.51

Table 1: Surgery Schedules in the simulation

The current (Base) case duration categorization correlates with surgery duration with r = 0.596 (n
= 5161). The categorization used in the scenario "All" is able to increase this to r = 0.626.
However, the unexplained variability remains considerable in both scenarios.
The performance measures for analyzing the effects of the schedules and queue length were the
daily number of operations at OR and planned surgery schedule adherence. Simulation run length
was 46 weeks (1 yr) with a warm-up time of 23 weeks. A correlated sampling approach was
applied to all sources of model input.

4 Results
Figure 2 shows the simulation results of a six-month care guarantee with different average waiting
times (queue length). The upper line denotes scenario “All” while the lower, steadier line shows
the results of scenario “Base”. The results shown are averages of five independent simulation runs
of 46 weeks (i.e. one year) with a preceding 23-week warm-up. The Figure 2 shows also a 95 %
confidence interval for each individual point calculated based on these five replications with a t-

y 2,3
t 2,1
p 2
1 2 3 4 5 6
Queue (in months)

Figure 2: Productivity as a function of queue length (i.e. average waiting time) with a 6-month
care guarantee. Upper line is scenario “All” and lower line scenario “Base”.
With a six-month care guarantee the optimal queue length in the scenario “Base” was 3 months
and productivity 2.22 cases per OR session. In scenario “All” the maximum productivity was 2.47
cases with optimal queue length also 3 months.
Scenario ”Base” Scenario ”All”
Queue Not from schedule Over CG Not from schedule Over CG
1 20 % 0% 32 % 1%
2 16 % 1% 15 % 12 %
3 10 % 4% 12 % 21 %
4 10 % 24 % 4% 61 %
5 4% 80 % 0% 99 %
6 0% 100 % 0% 100 %

Table 2: Surgery schedule non- adherence with the six-month care guarantee (CG)
Table 2 explains the causes of productivity differences in Figure 2 more clearly. With shorter
queues there existed problems in finding appropriate cases to fulfill the planned surgery schedule.
This can be seen in Table 2 in the “not from schedule” –columns: For example a one-month queue
in scenario “Base” does not have the cases available in the queue to meet the surgery schedule
20% of OR days. With increasing queues the proportion of cases exceeding the care guarantee
(CG) increases and disturbs the optimal daily scheduling, because such cases override the planned
surgery schedule. In Table 2 this can be seen as the increase in column Over CG, that shows the
percentage of OR days where the surgery schedule is overridden because of cases exceeding care
guarantee. Both these effects are more marked in the more productive scenario “All” than in the
less productive scenario “Base”. However, with pairwise t-test applicable to correlated sampling,
for both scenarios the three-month productivity is statistically significantly higher than either the
one-month or the six-month productivity of Figure 2.

Figure3: Productivity as a function of queue length (i.e. average waiting time) with different care
guarantee length in scenario “Base”.
Figure 3 shows the productivity of care guarantees of three, six and twelve months with different
queue lengths in the scenario “Base”. As in Figure 2, the data points shown are averages of five
independent replications of a 46-week simulation with a 23-week warm-up. For a short queue (one
month), the performance is equal regardless of the care guarantee: the difficulty of finding cases
matching the surgery schedule dominates. As queue length approaches the care guarantee, the
productivity is reduced with all care guarantees as already shown in Figure 2 for the six-month
care guarantee. A productivity peak is found when waiting times are somewhere in the middle of
care guarantee and no queue at all. The productivity maximum is statistically significantly higher
than one-month productivity for each care guarantee length at p=0.05 level. Overall, the absolute
differences are quite small with the scenario “Base”, due to the rather simple and robust surgery
Figure 4 shows the productivity of different care guarantees of three, six and twelve months with
different queue lengths in the scenario “All” with same simulation procedure as in Figure 3.

Figure 4: Productivity as a function of queue length (i.e. average waiting time) with different care
guarantee length in scenario “All”.
For a short queue (one month), the performance is again almost equal regardless of the care
guarantee: the difficulty of finding cases matching the surgery schedule dominates. As queue
length approaches the care guarantee, the productivity is reduced very clearly with all care
guarantees. Out of the queue lengths simulated, a productivity peak is found with six and twelve
months care guarantees at the waiting time of three months. The six-month care guarantee has a
higher productivity maximum (2.47 cases per OR and day) than the three-month care guarantee
(2.39), and the difference as calculated by pairwise t-test is statistically significant at p=0.05. The
12-month care guarantee has still slightly higher productivity maximum (2.51) than the six-month
care guarantee, but this difference is not statistically significant at p=0,05 although it is at p=0.1.
The absolute differences are more remarkable with the scenario “All”, due to the increased number
of planning classes and more complex and ambitious surgery schedule.

5 Discussion and conclusions

The simulation results lead to several conclusions regarding the length of care guarantee in the
case environment of orthopedic surgery: Firstly, productivity is decreased when the average
waiting times approach the care guarantee limit, because of disrupting effects of forcing the
overdue cases to override the planned surgery schedule. In such a case, either increase of care
guarantee or shortening of the queue would increase productivity.
Secondly, longer queue lengths and waiting times increase productivity as long as the case waiting
times remain well below care guarantee. Shorter care guarantees in turn mean that the optimum
queue length is shorter and therefore overall maximum attainable productivity lower.
Thirdly, the exact surgery scheduling approach has a clear influence on the sizes and details of the
productivity effect. Therefore, the particulars of each situation make any generalizations hard
regarding the actual effect size from our case situation research.
The results generate also lot of discussion about the further policies and studies: The study showed
that in surgical services the optimal waiting time from service provider perspective might be only
few weeks longer than patient preferences. This means that patient preferences for shorter waiting

times are not out of keeping with efficient service production. If current surgery queues are longer
than the optimal 3 months, there exists basically no lack of resources but incomplete practices to
adapt resources to actual demand.
The study also indicates that in a situation of very long queues, the introduction of care guarantee
might first deteriorate the overall efficiency and increase the average waiting time. Dawson et al.
(2007) argued that in patient choice -policy hospitals which lose patients to competitors have a
bigger incentive to reduce their waiting times or improve quality. Contrary to care guarantee, this
leads to shortened average waiting times. There is some evidence that both from patient
satisfaction and efficiency point of views the patient choice -policy leads to better results than care
guarantees. However, more research is needed about advantages and disadvantages of those two
More sophisticated scheduling algorithms to find appropriate cases with expiring cases might
partially solve the problem of diminished efficiency. In practice, queue managers and nurses could
anticipate cases that will exceed the limit and schedule that group before the date of expiration if
the effect on efficiency remains limited. Developing more intelligent scheduling algorithms is an
area of further research.
The study also highlights the question of maintaining a constant queue length. Capacity
flexibilities or demand management are needed in order to keep the waiting time near optimum. In
surgical services surgeons can be allocated between outpatient clinics and ORs based on the
queues in different phases in the process. On the other hand, scheduling should take more into
account the different priorities and preferences of patient (Oudhoff et al. 2007). Therefore, there is
need for further research to develop scheduling algorithms that best fit the varying patient
preferences with OR efficiency.

ACT Health (2007) Waiting Time and Elective Patient Management Policy. Chief Executive, ACT
Health, Australia. Available June 3rd, 2009 at:
Coxa (2009), homepage, (accessed 14.1.2009)
Coyte, P.; Wright, J.; Hawker, G.; Bombardier, C.; Dittus, R.; Paul, J.; Freund, D.; Ho, E. (1994)
Waiting Times for Knee-Replacement Surgery in the United States and Ontario, The New
England Journal of Medicine, Vol. 331, pp. 1068-1071.
Dawsin, D.; Jacobs, R.; Martin, S.; Smith, P. (2006) The impact of patient choice and waiting time
on the demand for health care: results from the London Patient Choice project. Applied
Economics (UK), Vol. 38, No. 12.
Dawson, D.; Gravelle, H.; Jacobs, R.; Martin, S.; Smith P. (2007) The effects of expanding patient
choice of provider on waiting times: evidence from a policy experiment. Health Economics,
Vol. 16: pp. 113–128.
Deming, W.E. (1986) Out of the Crisis. Cambridge, MA: Massachusetts Institute of Technology,
Centre for Advanced Engineering Study.
DH (2009).18 Weeks Referral to Treatment Statistics. Department of Health. Available June 3rd,
2009 at:
Donham, R.T.; Mazzei, W.J.; Jones, R.L. (1996) Glossary of times used for scheduling and
monitoring of diagnostic and therapeutic procedures. American Journal Anesthesiology 23:

FINLEX (2004). Act on Care Guarantee. Available June 3rd, 2009 in Finnish at:
Hanning, M. (1995) Waiting for care: maximum waiting-time guarantee in the Nordic Countries:
background, design, and effects, Spri-rapport, 0586-1691; 412, Stockholm.
HE (2007) Wait-time conference avoids care guarantee. Health Edition, Vol. 11 No. 7. 2007.
Hirvonen, J. (2007) Effect of waiting time on health outcomes and service utilization: A
prospective randomized study on patients admitted to hospital for hip or knee replacement.
Doctoral dissertation (article-based), University of Helsinki, Faculty of Medicine,
Department of Public Health National Research and Development Centre for Welfare and
Hurst, J.; Siciliani, L. (2003) Tackling Excessive Waiting Times for Elective Surgery: A
Comparison of Policies in Twelve OECD Countries. OECD Health working papers.
Jacobson, P. (2004) Health care markets and the health care guarantee: baking a better loaf, or
baking enough bread? Policy Options, August 2004.
Jonhsen, J.R. (2006) Health Systems in Transition, Norway. European Observatory on Health
Systems and Policies, Vol. 8 No. 1 2006.
Kujala, J.; Lillrank, P.; Kronström, V.; Peltokorpi, A. (2006): Time based management of patient
processes. Journal of Health, Organisation and Management, Vol. 20 Issue 6, 2006.
Lehtonen, J-M.; Torkki, P.; Peltokorpi, A.; Moilanen, T. (2009) How to increase productivity by
improving scheduling in operating rooms? 16th International Annual EurOMA Conference,
Göteborg, June 14-17, 2009.
MH (2009) Elective services guidelines. Ministry of Health, New Zealand. Available June 3rd,
2009 at:
MIH (2002) Health care in Denmark. Ministry of the Interior and Health. 5. edition, August 2002.
Available June 3rd, 2009 at:
NHS (2009). The right to choice and information: NHS Constitution. Available June 3rd, 2009 at:
Oudhoff, J.; Timmermans, D.; Rietberg, M. (2007) The acceptability of waiting times for elective
general surgery and the appropriateness of prioritising patients. BMC Health Services
Research, Vol 27, No. 32, 2007.
Sampalis, J.; Boukas, S.; Liberman, M.; Reid, T.; Dupuis, G. (2001) Impact of waiting time on the
quality of life of patients awaiting coronary artery bypass grafting. CMAJ. 2001 August 21;
165(4): 429–433.
SKL (2005) Vårdgaranti / Vård i rimlig tid. Svenska Kommuner och Landsting. Available June
3rd, 2009 in Swedish at: