Sunteți pe pagina 1din 7

A NATIONAL TRAUMATIC COMA DATA BANK

Selma C. Kunitz, Cynthia Gross,


Barbara Nichols and Sylvia Edelstein
Computer Applications Section, Office of Biometry and Field Studies, NINCDS/NIH
Bethesda, Maryland 20014

The medical record, which contains data routinely


collected as a part of the care of patients, has
ABSTRACT long been recognized as a potential resource for
research.4-
However, the data contained in
A number of university hospital centers are par- these records have frequently not been useful for
ticipating in the pilot of a national computeriz- research purposes for the following reasons. The
ed clinical data bank for traumatic coma. Pa- data elements are not consistently defined
tient data including history, symptoms, treat- among hospital centers or even among physicians
ment, and outcome are collected prospectively at the same center; they are not uniformly col-
during the patient care and follow up process lected on each patient; and they are generally
according to common definitions. The data bank difficult to retrieve from the medical record.
was implemented to demonstrate its usefulness
In the 1970's, clinical data banks for lung,
for clinical research and patient management. To cardiovascular,8 and rheumatic diseases were
achieve these goals, a clinical data bank requires introduced.0' , 8These data banks store
research methodology and computer technology information on the course of illness for large
linked with the needs of the patient environment. numbers of patients and provide rapid access to
a dynamic "textbook of chronic disease" to ait
This paper describes the research design, the the decision-making process of the physician.
implementation, and the computer system of the In addition, these data banks provide the computer
Traumatic Coma Data Bank, which began in 1979. systems technology required to store the temporal
relationships among items of information which
collectively constitute the patients' clinical
course. To be useful for research, however, a
clinical data bank requires both the appropriate
computer technology and a workable research
strategy.
INTRODUCTION
The National Institute of Neurological and Commu-
Traumatic head injury is a major health problem nicative Disorders and Stroke (NINCDS) of the
in the United States. Accidental head injuries National Institutes of Health (NIH) has developed
affect over four hundred thousand Americans each a national collaborative pilot data bank program*,
year, and many of these persons will experience which combines the clinical data bank concept
periods of unconsciousness. Motor vehicles are with the formal structure of research methodology.
the major cause of these accidents, and the Within the context of the pilot data bank program,
majority of these victims are young.40 A number a clinical data bank is defined as a collection
of therapies have recently been introduced to im- of data bases from several hospital centers,
prove the management of traumatic coma victims, each containing patients' medical history, des-
and further study is needed to assess the impact cription of the onset of the disease (or precip-
of these therapies on survival, recovery of itating accident), the therapy, and the outcome,
function, and quality of life. recorded throughout the patients' clinical course
according to an established set of definitions.
As with many other neurologic disorders, clinical These data are entered, stored and retrieved in
research in traumatic coma is hampered by a lack a rapid, flexible manner through the use of a
of useful data. Data from large numbers of pa- computer system which retains information on
tients, carefully and prospectively collected, the data's temporal relationships.
according to precise definitions, are needed to
identify and elucidate the course of traumatic Many academic hospital centers are actively par-
coma and its sequelae. Data are needed to des- ticipating in the collaborative data banks, whose
cribe the complexities of management and the two major goals are to provide a data resource
patterns of survival and recovery. Multidi- for clinical research and to aid in patient
mensional assessment batteries are needed to
characterize short and long term patient outcome *An NINCDS Stroke Data Bank Network is also part
during follow-up periods. of this program,

1333
U.S. Government work not protected
by U.S. copyright.
management. If the pilot data banks are success- systems have provided a formatted medical record
ful, they will provide models for similar data but have not attempted to develop and adhere to
banks for other diseases and other purposes. uniform definitions which are a requirement of
clinical research. They have also not provided
The Traumatic Coma Data Bank is discussed in the the ability to access or manipulate user-chosen
following sections with a focus on its goals, subgroups of patients; an ability which is re-
research design, uses, and the computer system quired for data analysis. Although direct en-
which makes it operational. hancement of patient care is not a primary ob-
jective of these data banks, some benefit may be
GOALS OF PILOT DATA BANK derived in this area by the introduction of
structured, computer retrievable medical records13
A primary goal of the NINCDS data banks is to for review and use by the attending physician.
demonstrate their utility for clinical research.
Adaptation or development of research designs and DATA BANK DESIGN AND OPERATION
methodological guidelines appropriate to the
data banks are necessary to achieve this goal. The pilot Traumatic Coma Data Bank consists of
Careful study design, which can provide data four federally-funded university hospital centers
that are useful for clinical research, requires (Appendix I), called Clinical Data Bank Centers.
that data are collected in a manner which is (clinical centers) Each center has as Principal
consistent over time among both data collectors Investigator, a neurosurgeon who is primarily
and subjects.10 The NINCDS data bank research responsible for all activities involved in the
design specifies systematic, organized methods data bank. The Principal Investigator leads a
for the implementation and operation of the team composed of other physicians, nurse practi-
clinical data bank; for data identification and tioners, research assistants, and data entry
collection; and for determining the limitations persons. A separate federally funded Data Bank
of the data for clinical studies. Maintenance Center, which is at the Stanford
University Medical Center, provides the computer
Studies arising from the clinical data bank are support to store and retrieve data. The Clinical
observational. An observational study seeks to Data Bank Centers, the Data Bank Maintenance
describe what is happening in an environment, Center, and the Office of Biometry and Field
and is an important method for research on human Studies (OBFS) in NINCDS collaborate on all
problems in situations where one is unable to phases of the projects. OBFS, which sponsers
ethically experiment. Observational studies can this program, also offers technical expertise
incorporate epidemiologic procedures for the se- in systems design, computer science, biosta-
lection of cases and the collection of data. The tistics and epidemiology.
carefully planned observational study, which The design of the Traumatic Coma Data Bank in-
utilizes statistical methods for data analysis
can provide descriptive information concerning cludes specification of research questions; of
disease stages, associations among symptoms, data element definitions; of data collection -
intervention, complication and outcome, and protocol ,s and procedures; and of quality assurance
generate leads for clini al experimentation. methods. Each of these elements is essential
Dambrosia and Ellenbergil have recently described to an effective collaborative observational study
their views on the utility and and limitations useful for clinical research.
arising from medical data bases. The research questions for traumatic coma concern
A second goal of the pilot data banks is to offer, issues of patient management; specifically the
through the computer system, administrative aids relationships among types of emergency care,
for patient management such as structured medical early symptom presentation and outcome; the
records, automated discharge notes, and tickler association of intensive care monitoring, treat-
lists designed to assist in scheduling follow-up ment and outcome; and identification of the
visits. It is hoped that these benefits will pro- determinants of prognosis.
vide continuous motivation to collect consistent
data. Consistent, accurately recorded data are Vocabulary
necessary for good clinical research. It has
been shown that physicians are more likely to ad- The research questions provide a strategy for
here to protocols, to utilize computers, and to the selection of data elements to be included in
collect consistent data if they are provided rap- the data bank, which is termed the vocabulary.
id and useful feedback from these systems.12 The vocabulary includes data elements which
describe the patients' background, accident
The development and use of data collection forms details, transport modes, emergency room care,
which can serve as the medical record and which clinical history, symptoms, laboratory tests,
are also compatible with data entry into the com- interventions, and outcome. In addition, the
puter for research purposes is an objective of
Clinical Data Bank Center staffs have concentrat-
ed on detailed data capture of the complex,
the pilot data bank. In most clinical studies often simultaneous events occurring during inten-
data are abstracted from existing medical records sive care, and on developing a more comprehensive
or are collected by means of a form developed and set of outcome measures than are currently util-
utilized for a single, specific study. In con- ized. The Principal Investigators have mutually
trast, most currently automated patient record defined and agreed to the of definition for each

1334
item in the vocabulary. elevated 30 degrees. Protocols requiring specit-
Data Collection Forms ic tests or other medical interventions that
would change the standard care of patients are
The order in which data are routinely collected not appropriate for the Traumatic Coma Data Bank.
in the patient care setting provide a structure The development of the collection protocols was
for the data collection forms. Some of the forms a difficult task; it was found that if cominon
are in a format similar o the Time-Oriented Re- goals were shared and fostered among investiga-
cord developed by Fries , in which data are col- tors, the development was easier.
lected in columns, by date and time (Figure 1).
The data element codes, definitions, instructions,
TRAUMATIC COMA DATA BANK and data collection protocols comprise the Data
EVALUATION AND TREATMENT Dictionary which is a part of the Administrative
FORM TYPE 3 TOR NO
Manual for each of the centers. Eligibility cri-
DATE teria and informed consent procedures are outlin-
3 TiME ed in Administrative Manuals along with flow
6 i OBSERvER Flosp tai Keeps List of Codes) I = Emergency Room diagrams to decribe case ascertainment, follow-up
3-9 = Nurse
10-20 = Physician
procedures, and personnel rosters. The Admini-
PLACE OF EVALUATION
strative Manuals are reviewed at six month inter-
1 = Pldce of njirv 4 = In Transit to Data Bank Hospital' vals for possible revision.
2 In Tr&isitrv Forst Hospital 5 = At Emergency Room of Data
3 = At Ernergensc Room of Bank Hospital'
|Fist HospItai Data Collection
-%ELACL^;CGCAL EVALUATION
20 , BEST EYE OPEN/NG EChoose One) In each clinical center all patients who are el-
I -"Nre
2= To Pan
6
7
=
=
Injur.ediSwollen
barbiturites, Narcotics. or
igible for the study are followed for two years,
3 To Sound Pharmacologic Paraiyss or until the project ends, patient dies, or is
4 - Spoitaneous B = O-her otherwise lost to follow-up.
5 - Pe:x!ie Tarsorrhaphy
B:ST MrOTOR RESPONSE (Choose One}
1= None
2 - Extensor
7 - Limb iniury;lmmobilization
B - Spinal Cord lbjury
All data are collected by examination of the
Fti.9'sL l
patient or other original source by, or under
Time-oriented records are used for data collected the supervision of the Principal Investigator.
in a repetitive manner during the hospital stay. Data are collected at specific time intervals
Other forms are formatted for a single encounter depending on the patient's status. For example,
or procedure, such as demographic histoW, pa- data are collected every eight hours during
tient history, and surgery (Figure 2). For an- intensive care, and then daily during in-pat'ient
care. In addition, an event triggered data col-
TRAUMATIC COMA DATA BANK lection method is utilized. For example, if in-
PATIENT INFORMATION (Cont'd.) tracranial pressure or blood pressure is elevated
beyond a specific range, additional data are col-
DETAILS OF PATIENT ADMITTANCErrRANSFER lected for a specific time period. Event trigger-
H45 PATIENT ADMITTED/TRANSFERRED TO DATA BANK HOSPITAL ed data collection helps to organize the many
00. = Direct From Scene of Accident patient changes and interventions which occur
002 = From Home (Not Scene of Accident1 in a short time frame in an intensive care unit.
003 = From Non-Hospital Provider
= From Other Huspital

H56 COLLABORATING HOSP!TAL Training


(See Code List irn Ad-ministrative Manual)
H46 TRANSPORT MODE TO FiRST HOSP!TAL Training programs were developed to teach data
- Self
2 = HeletiveiFriend
4 = Helicopter/Plane l entry procedures at each of the clinical centers.
5 = Police
= Ambulance 6 = Other Data collection personnel have attended meetings
H57 TRAINING OF TRANSPORTER TO FIRST HOSPITAL
with the Principal Investigators in order to
0 =None become thoroughly familiar with the data defini-
1 = Police/Firenmen (if less than EMT 1 training) tions, instructions, and data collection pro-
2 = EMT 1 (Bas;c EMS) cedures. Workshops on evaluating patient sta-
3 = EMlT 2 (Advanced EMS)
4 = Nurse tus on follow-up and on data retrieval methods
5- Physician
6 =Other.__ _
occurred.
7 = Unknown
F.i Quality Assurance
alytic purposes, it is important that the forms Quality assurance methods are desi.gned for pa-
contain conventions which deal with categories tient selection, data collection, data entry, and
of missing responses such as "no data", "unavail- data retrieval. Some methods are objective and
able data", and "not applicable". staightforward, such as computer edits and com-
prehensive forms design. For example, a computer
Procedures and Operations Manual edit might prevent an illogical relationship,
e.g., pregnant male. Other methods are more sub-
Where appropriate, specifications for the method jective and approximate, requiring on-site moni-
of data collection are expressed by collaborative- toring and frequent problem-sharing. Specific
ly defined protocols. For example, intracranial methods include the option of central reading of
pressure, if monitored, must be with the head CT scans; compilation of data summaries to check
for anomalous results within and among centers;

1335
use of common diagnostic criteria; standardized Each Data Bank Center has a Datapoint 1800 intel-
protocols for patient evaluation; agreement on ligent CRT processor with dual floppy disk drives
specific times for patient observation and col- and printer, for communication with the Stanford
lection of data; determination of compliance computer. The Datapoint 1800's are microproces-
to th data definitions; and patient management sors. To take full advantage of their potential,
aids. 4During the data collection test phase, each center is encouraged to use them for admin-
cases were reviewed and discussed among all of istrative assistance to their patient care process,
the Principal Investigators and their staff. as well as for the data entry and retrieval
This collaboration served to identify areas needs of the data bank.
of ambiguity in the instructions and definitions
and many items were revised or clarified. The TOD System
During the operational phase of the pilot, sam-
ples of the completed forms are periodically TOD is programmed for data entry and editing ei-
compared with items stored in the computer. An ther online or through batch transmission via
external observer will visit each of the Data commercial communications lines such as Telenet
Bank Centers to independently examine and dia- and Tymshare. Central to the TOD system is a
gnose patients and record information. schema which is composed of the attributes of
each data element. The TOD system was designed
COMPUTER TECHNOLOGY for an out-patient setting and is based upon a
patient-clinician encounter which is called a
The computer system utilized in the pilot data VISIT. Data collected in the first and subse-
bank integrates the storage and retrieval cap- quent encounters, as well as laboratory tests
abilities of a Data Base Management System (DBMS) and therapies, are stored by VISIT number and
which are useful for research with the benefits date. Background and demographic data which do
of a local computer which can serve the patient not change over time are stored in a separate
management needs of each clinical Data Bank file. All patient medical history data are
Center (Figure 3). The specific attributes re- collapsed into a single visit, called VISIT 0.
The Traumatic Coma Data Banks, however, must con-
sider the handling of patients in an acute care
Data Bank hospital setting. In the hospital, patients are
Center Daow Bank examined, treated, and evaluated several times a
'N
CGntam day. A patient's status, as well as therapy, may
change dramatically within a few hours, especial-
ly in an intensive care unit. In order to accu-
rately describe a patient's course, these changes
must be collected and stored. In addition, a
specific event type, such as a surgical procedure,
or arteriogram, is more relevant to hospital data
collection than the VISIT concept. For these
reasons, the TOD system was modified to record
data elements by time, date, and event type as
well as VISIT.
DATAPOINT INTERFACE
While the TOD system manages data in the central
Dapdnt 180 data bank in a timely and flexible manner, it
Mloproc_o was not designed to meet the needs of acute care
environments in geographically dispersed centers.
ViuRee 3 However, a microprocessor or intelligent terminal,
which itself is a small computer, is an appropri-
quired of a DBMS for the clinical data bank ate tool for the patient care setting. Rather
include: ease of use by nontechnical personnel; than choosing "dumb" terminals, which can only
adaptability to change; flexibility in access transmit and receive data, the Datapoint intelli-
and retrieval of changing subsets of data; modes gent terminals were selected for the Data Bank
of linkage to statistical packages and higher Centers. A "front-end" general purpose software
order programming languages; and security with package was developed for the Datapoint, which
respect to system intergrity and confidentiality. allows data to be entered, edited, and stored
locally by time and date. This software package
The clinical data base management system utilized also sorts the collected data and prepares it for
is a modified version of the Time Oriented Data subsequent transmission in a batch mode over a
(TOD) System and is maintained by a subgroup conunercial line to the TOD system which accom-
of the American Rheumatism Association Medical plishes more complex editing.
Information System (ARAMIS). TOD-ARAMIS, which
was developed specifically for the clinical The microprocessor software operates with menu
environment, is housed on an IBM 3033 at Stanford processing, in which a non-programmer such as
University Medical Center, and is managed by the data entry person can choose the relevant
ORVYL, a powerful operating system. form(s) for data entry from a list of options.
It produces screen images which replicate the

1336
order of the data on the the data collection and
protocols;
record (Figure 4). During data entry, data are
2. Design new patient record forms to aid the
PATIENT OFFLINE DATA ENTRY AND UPDATE clinicians and to facilitate the process
of data entry into the computer;
PAGE Past History Fl = JUMP F3= ABORT F5 = WRITE 3. Require follow-up data such as long-range
outcome on patients to permit estimates of
M.R.N. 00001 PAGE NO. 2A ENTRY DATE 01 Jan 1980 TIME 1200 prognosis; and
4. Develo% quality control procedures for data
collection for clinical research.
OLD CHANGE
P11 Location in Community ........................ 1
In the early stages of the data bank planning and
development, a communication barrier existed
P435 Mental Status Problem ......................... 2 among the various disciplines, partly because of
the subject-specific jargon that tends to pervade
P18 Employment Status .......................... 2 one's professional speech patterns. Numerous
P19 Unemployed/Retired for Medical Reason ......... 0 - meetings, perserverance, and increased confidence
in the data banks dissipated these barriers and
P425 Performance Class ........................... 1 2 common goals emerged which replaced stereotyped
P426 Placement Clss ............................ 1 -
discipline viewpoints.
F 94i.e LI The data bank projects have taken advantage of
edited for valid numeric ranges, alpha/numeric the specific definitions required by DBMS to
field consistency, code lists, and special for- integrate the computer technology with research
mats such as dates. Data are protected by a methodology. As with many medical applications
password system and confidential items such as of DBMS, this planned integration might have been
name and address are never transmitted to the overlooked, if the project did not involve a mul-
central data bank. Changes to the TOD schema tidisciplinary team.
update a local Datapoint program which automat-
ically generates changes to the local entry ASSESSMENT AND FUTURE DIRECTIONS
program. The data collected locally are stored
on a diskette by patient, data element, date, The Traumatic Coma Data Bank is a pilot effort.
and time. The data housed on the diskette are During the three-year pilot, each Data Bank
useful for generating information on an individual Center will collect and enter data on a minimum
patient such as course during hospital stay, of fifty patients annually. At the end of the
discharge and referral notes. These data will pilot, NINCDS will carry out a heuristic assess-
also be used to produce lists of patients such ment of the data bank according to the following
as those needing follow-up visits, and simple goal s:
patient counts.
a. Ability of diverse institutions to partici-
When data are received by TOD from the local pate, collaborate, and benefit from a
system, they are stored in an ENTRY file. Data computer data bank network.
are then transposed into the central data bank b. Ability to collect data of high quality in
and maintained separately for each Data Bank the data bank.
Center; that is, arranged in a matrix ordered by c. Preliminary utility of data for (1) deter-
data item, patient, date, and time within center. mining new research leads and (2) patient
Transposed data are accessed to generate subsets management.
of patients for research purposes. Data analyses d. Ability to provide guidelines and protocols
can be run on the Stanford system or data can be for expansion to additional centers and
sent back to the local centers for use with sta- other neurologic disorders, if appropriate.
tistical packages and higher order programming
languages. In addition, the biostatisticians involved in the
pilot program will examine the consistency of
An aspect of the computer system which may be the relationships among variables across the clin-
addressed during the latter stage of the pilot, is ical centers in order to determine the appropri-
the feasibility of establishing an automated link ateness of aggregating the data for research
between the microprocessor and monitoring equip- studies.
ment used during intensive care. The computer system aspects of the data bank are
MULTIDISCIPLINARY PROJECT TEAM also being assessed under an independent contract
which will eventually outline specifications for
The implementation and operation of the Traumatic an optimal clinical data bank system.
Coma Data Bank has involved a multidisciplinary
team of neurosurgeons, biostatistians, computer Long Range Goals
scientists, and epidemiologists. The team has If the pilot data bank is successful, a full-phase
had to:
clinical data bank will be supported so that the
1. Develop standard data element defi-nitions long range goals of the data bank may be realized.

1337
Long range goals include evaluation of the con- bank developments will add a new and much needed
tribution and effectiveness of the clinical data dimension to the clinical research of not only
banks with respect to improvements in patient traumatic coma but to other neurological
care, research, and teaching; a determination of disorders as well.
whether the clinical data banks can provide ex- REFERENCES
perienced-based computer consultation; evaluation
of the overall costeffectiveness of the network; 1. NIH Pub. No. 79-1910 (June, 1979). National
determination of whether a significant part of Research Strategy for Neurological and Communica-
the data bank system's cost can be defrayed by tive Disorders. U.S. Dept. HEW, PHS, NIH, NINCDS.
providing reports required by various components
of the patient care system. 2. Jennett, B., Teasdale, G., Galbraith, S.,
Pickard, J., Grant, H., Braakman, R., Avezaat,
The computer system used for the Traumatic Coma C., Maas, A., Minderhoud, J., Vecht, C.J., Heiden
Data Bank can, and most likely will, be utilized J., Small, R., Caton, W., and Kurze, T. (1977)
to collect and retrieve the data for clinical ex- Severe head injury in three countries. J. of
periments such as clinical trials. The clinical Neurology, Neurosurgery and Psychiatry,
trial involves the standardization and randomiza- 40:291-298.
tion of the treatment regimens among the study
patients. A long term goal of the data bank 3. Miller, J.D. Sweet, R.C., Narayan, R., and
program is to incorporate clinical trials when Becker, D.P. (1978). Early insults to the injured
the efficacy of a specific treatment or therapy brain. J.A.M.A., 240:5, 439-442.
is under question and the requirements for ran-
domization and protocol can be met. 4. Feinstein, A.R., and Koss, N. (1971) Computer-
aided prognosis. Arch. Intern. Med., 127:438-447.
Another long term goal is to determine the use-
fulness of the clinical data bank in the develop- 5. Weed, L.L. (1968) Medical records that guide
ment and application of prognostic information. and teach, N.E.J.M., 278(11): 593-600.
The usefulness of data bank prognostic information
will depend upon the completeness and validity 6. Ibid, Feinstein
of the data as well as on the ability to develop
rules in using the data. 7. Rosati, R.A., Wallace, A.G., and Stead, E.A.
CONCLUSION (1973). The way of the future. Arch Intern.
Med., 131:285-287.
Carefully collected data on the clinical history
of head trauma and other neurologic disorders, 8. Wiederhold, G., Fries, J.F., and Weyl, S.
in general, is a scarce commodity. The clinical (1975) Structured organization of clinical data
data bank can provide descriptive information bases, AAFIPS - Conference Proceedings 44:479-485.
concerning disease stages, and associations
between symptoms, intervention, and outcome. 9. Rosati, R.A., McNeer, J.F., Starmer, C.F.,
Through observational studies, the Traumatic Coma Mittler, B.S., Morris, J.J., and Wallace, A.G.,
Data Bank may provide leads for clinical experi- (1975) A new information system for medical prac-
mentation and has the potential for supporting tice. Arch. of Intern. Med., 135:1017-1024.
clinical trials.
10. Tukey, J.W. (1977) Some thoughts on clinical
The computer aspect of the Traumatic Coma Data trial especially problems of multiplicity.
Bank, while based upon existing technology, is Science 19, 679-684.
innovative. The system integrates a modified
version of a clinical data base management system 11. Dambrosia, J.M. and Ellenberg, J.H. (1980)
with a software package for the Datapoint 1800 Statistical considerations for a medical
in order to edit, store, and access data in data base, Biometrics. 36:323-332
dispersed geographic areas and aid in providing
patient records, discharge notes, and other 12. deDombal, F.T. Leaper, D.J. Horocks, J.C.
administrative aids to patient care. Staniland, J.R., and McCann, A.P. (1974) Human
and computer aided diagnosis of abdominal pain;
The development of a uniform vocabulary used by further report with emphasis on performance of
several centers is a beneficial product of the clinicians. Brit. Med. Journal. 1,376-380.
data bank program. Several other neurosurgical 13. Weed, L.L. (1971) Quality control and the
centers have requested the data collection forms medical record, Arch. Intern. Med. 127.
and operations manuals for use in their centers.
14. Kunitz, S.C., Havekost, C.L., Gross, C.R.
Recent articles17'18 have stated that research (1979); Pilot data bank networks for neurological
in chronic disease can benefit greatly from the disorders. Third Annual Proc. Symposium of
technology of data base systems. In order to be Medical Care:793-796.
useful for clinical research, however, a clinical
data bank requires an appropriate data base 15. Fries, J.F. (1972) Time-oriented patient
technology integrated with the requirements of records and a computer data bank. J.A.M.A., 222,
research methodology and the needs of the clinical 1536-1542. Fries, J.F. (1974) Alternatives in
environment. It is hoped that the current data medical record formats, Medical Care XII:871-881.

1338
16. Gross, C. and Dambrosia, J. (1980) Quality
assurance for clinical data banks; (Unpublished
manuscript)
17. Abbrecht, P.H., Cox, J.R., Ferguston, F.P.
(1978) Future directions for biomedical engineer-
ing research: recommendations of an evaluation
workshop for the NIGMS Physiology and Biomedical
Engineering Program; IEEE Transactions on Bio-
medical Engineerinng: BME-25:490-493.
18. Laird, M., Weinstein, M.C., and Stason, W.B.
(1979) Sample-size estimation: A sensitity
analysis in the context of a clinical trial for
treatment of hypertension. Am.J. Epid.,
109:4, 408-419.

1339

S-ar putea să vă placă și