вход по аккаунту


Toward an Understanding of Patient Outcome Measurement.

код для вставкиСкачать
Official journal of the American Rheumatism Association Section of the Arthritis Foundation
It is difficult to objectively measure the effects
of arthritis or its treatment, and agreement on satisfactory endpoints has not been reached. Each proposed
measure, from ring size to 50-foot walking time to
articular index, has had its several proponents and its
many critics. Now, new questionnaire-based instruments attempting to measure patient “outcome” have
appeared (1-7), couched in an unfamiliar jargon and
using suspicious terms such as “quality of life.” These
questionnaires, it is implied, measure the truly important things. Yet the supporting literature is often dull
and hard to understand, and it is difficult to tell what is
science and what is self-serving. As new articles
appear in the literature and more investigators include
such measures in their studies, an evaluation of outcome measurement appears timely. This report attempts to describe the present state of the art, the
future directions, and the emerging problems, and to
suggest potential applications.
Five distinctions are important in order to understand the rationale of outcome measurement.
From the Department of Medicine, Stanford University
School of Medicine.
Supported in part by grants to the American Rheumatism
Association Medical Information System (ARAMIS) (AM21393)
and to the Stanford Arthritis Center (AM20610) from the
Address reprint requests to James F. Fries, MD, Department of Medicine, Stanford University School of Medicine, Stanford, CA 94305.
Submitted for publication December 8, 1982; accepted in
revised form January 13, 1983.
Arthritis and Rheumatism, Vol. 26, No. 6 (June 1983)
Process versus outcome
First, there is the matter of “process” versus
“outcome.” “Outcome” is the end result, while
“medical process” is merely what happens along the
way. Death is an outcome; measurement of blood
salicylate level is but part of the medical process.
Disability is an outcome; a latex fixation titer measures part of the disease process. Clearly, it would be
better to measure outcome than process (8-1 I), since
it is a truism that the ultimate goal of health care is to
improve the outcome of patients. But what is meant by
“outcome,” and how may it be measured?
Outcome must reflect the values of patients and
of society, and many of these values are intuitively
obvious. For example, the patient desires to be alive,
to be functioning, to be free of symptoms and discomfort, to have few (if any) side reactions from the
therapeutic endeavors, and to remain financially solvent (including ability to pay insurance premiums)
through the course of a long illness (13).Conversely,
a pure measure of medical “process,” such as erythrocyte sedimentation rate or level of antibody to DNA,
has no such intrinsic importance either to the patient
or to society. The equivalent folk saying is that “the
operation was a success, but the patient died.”
Process measures achieve their value only to
the extent that they serve as accurate proxies for
outcome measures. Thus, if the sedimentation rate
accurately quantitates “fatigue,” and if “fatigue” is
part of patient discomfort, and discomfort is an outcome, then the sedimentation rate may have value as a
surrogate outcome measure, conveniently and accurately ascertained. If antibody to DNA accurately
guides choice of therapy and accurately predicts ulti-
mate renal insufficiency, then it may be useful as an
early and quantitative proxy for the outcome of uremia. But to the extent that these tests (or others such
as fluorescent antinuclear antibody or latex fixation
titers) represent epiphenomena only loosely correlated
with ultimate outcome, they represent false endpoints.
The confusion is compounded by measures in
common use which are not clearly either process or
outcome (12). Is “50-foot walking time” (or “joint
count” or “grip strength”) a “process” or an “outcome” variable? In fact, these variables have some
characteristics of each. It is possible to have many
tenderjoints yet to function well with minimal discomfort; it is possible to have a weak grip strength but to
have good overall functional capacity; it is possible to
have a slowed walking time but to be doing very well.
These variables should be considered to represent
“components” of patient outcome; only to the extent
that they are correlated with more general measures of
outcome do they serve as generally useful endpoints.
The present drive toward more comprehensive endpoints receives much of its impetus from dissatisfaction with traditional endpoints such as grip strength or
walking time. There is emerging recognition that correlations between measures of medical process and of
patient outcome are generally weak, and that spotty
analysis of components of patient outcome yields an
incomplete picture.
Hard versus soft data
Second, there is the matter of “hard” versus
“soft” data. “Hard” data are clearly to be preferred.
Such data are accurately determined, are numeric, do
not depend upon subjective impressions, are accurately reproducible, and identical values will be obtained
wherever the variable is assessed. Hematocrit (packed
cell volume) measurement gives such “hard” data.
Laboratory errors are few with the hematocrit count,
there is a high degree of standardization, and one can
be quite confident that values measured at one institution would be closely similar to values which would be
measured at another. Quantification of hematocrit
values is continuous on a long scale, and repeat
measurements in the same patient will give values
which are quite close. In contrast, “soft” data evoke
much less confidence. A patient complaint is generally
considered “soft” data. Such information is subjective, of suspect reproducibility, and is expected to
vary greatly from one location to another and from one
observer to another. Traditionally, “hard” data come
through the physical sciences, and “soft” data from
the social sciences or from human interactions such as
the taking of a medical history.
These distinctions, however, become blurred
on close examination. Numbers can be misleadingly
precise. For example, with variables such as antibody
to DNA or serum complement, available evidence
suggests variability on repeat testing of up to 20%, and
variability between laboratories may be much larger
than that (13). When critically analyzed, the creatinine
clearance and the 24-hour protein excretion are found
subject to measurement errors of large degree. Additionally, “hard” measures have fallen into disuse in
arthritis clinical trials because they are not very sensitive at detecting differences between treatments. A
recent survey of experts in clinical trials showed the
five most valued endpoints in rheumatoid arthritis to
be the global assessment by the physician, the joint
count, self-report of pain, self-report of morning stiffness, and grip strength, in that order (14). None of
these are “hard” measures in the traditional sense.
Table 1 presents an illustrative summary of studies
(13-15) and unreported experience with these measures.
On every one of the central characteristics that
define “hard” data (precision, quantitatioh, reproducibility, interobserver variation), most arthritis laboratory variables are inferior to the new questionnaire
instruments developed by social science research
techniques. New measures of disability assessment,
for example, are more reliable as measurements than
are sedimentation rate, antibody level, or grip strength
(4,5,13,15). Thus, it is crucial that the terms “hard”
and “soft” be either deleted from our vocabulary or, if
retained, used to refer to the actual measurement
characteristics of a variable rather than to the origin of
data (the laboratory or from the patient).
Validated versus not validated
Third, there is the new jargon of measurement
methodology. The new terms, drawn from the discipline of survey research (15-19). have an unfamiliar
ring. Outcome assessment research staffs include PhD
investigators with social science backgrounds; these
disciplines have their own vocabularies only slightly
less complex than those of medicine, and remarkably
different. Discussions of topics such as “face validity”
and “content validity” are unfamiliar to many physicians, and even the statistics used to buttress the
methodologic arguments are unfamiliar (3,20).
Table 1. Endooints for arthritis studies
Process measures only
Sedimentation rate
Latex fixation titer
Antibody to DNA
Surrogate measures
Grip strength
Walking time
Joint counts
Patient global assessment
Physican global assessment
Outcome measures
Psychologic discomfort
Physical discomfort
Financial impact
Estimated reproducibility
(same lab or
Estimated reproducibility
(different lab
or observer)
to change
2 tubes’
2 tubes*
1 grade*
1 grade*
I grade*
* Studies in rheumatoid arthritis not reported.
There is an important contrast between the new
questionnaire and the old measures. Traditional measures of arthritis impact were never analyzed for their
measurement characteristics; if they had been, our
disillusion would have come much earlier. It has been
seldom, if ever, asked whether the sedimentation rate
actually measures what we use it to measure, whether
it remains stable with successive measurements on
successive days, or whether different laboratories
would get the same results from the same specimen.
Our traditional measures fail the simplest tests of
measurement characteristics, yet these deficiencies
have received surprisingly little attention.
In contrast, the measurement characteristics of
the new instruments have been carefully studied
(4,5,7,15,2&23). This attention to detail came in part
because it is standard practice in the social sciences,
and in part because of recognition that general acceptance of “soft” measures would depend upon especially solid documentation.
There are at least six frequently reported types
of validity which indicate qualities that evidence
should ideally possess: 1) Face validity asks if the
instrument, on its face, looks as though it might
measure a reasonable endpoint. 2) Content validity
asks if the instrument actually measures the content
implied by the name. 3) Criterion validity asks about
the accuracy of the measurement. 4) Discriminate
(concurrent) validity asks if the measure is sufficiently
sensitive to different states to be able to distinguish
between two groups of patients who are known to
differ. 5 ) Construct validity asks if the variable can be
correlated with a “gold standard.” There are no
perfect “gold standards” in arthritis against which to
compare a proposed variable; thus, comparisons are
generally made with another imperfect measure, a
form of construct validity sometimes called “convergent” validity. 6) Predictive validity, sometimes
termed “empirical” or “criterion-related’’ validity, is
a practical test which asks if the measure can predict
future differences.
Short term versus long term
Fourth, there is the distinction between short
and long term results (24). Rheumatoid arthritis extends over an average course of 25 years. Health
status at any one time is a combination of present
events and accumulated past damage. Obviously, an
optimum therapy must be one which minimizes the
overall impact of the disease over time, neither disregarding the present status nor neglecting long term
difficulties. Results at some arbitrary single point in
time are certainly incomplete and potentially misleading (Table 2).
The new measures take a much longer view of
the impact of arthritis, and are designed for the assessment of slowly moving changes over long time periods. Particularly in the area of disability, these measures are thus relatively insensitive to short term
change, but effectively measure long term trends. The
necessity for long term measures is obvious, since
there are many clinical examples of short term tactics
that are deleterious over the long run. Indiscriminate
use of corticosteroids (Table 2) serves as a prime
example. Other examples include reliance on strong
narcotic agents, apparently successful with some short
term measures but almost certainly detrimental over
the long term, and use of cytotoxic agents with documented medium term benefit but with long term risks
of neoplasia. Implicit in the new outcome assessment
is an approach toward serial monitoring of health
status emphasizing the trends and tempo of the illness.
Univariate versus multivariate
Finally, there is a problem of the multiple
influences upon patient outcome (1,5,25). Nonbiologic
inputs clearly can affect outcome and cannot be ignored, yet our education and training systems are
designed to identify and modify biologic mechanisms.
It is clear that patient motivation, patient compliance,
education level, socioeconomic status, arrangement of
health care resources in the community, payment
mechanisms, public and patient education programs,
and family support, among other factors, may be as
important determinants of disability as are the biologic
activity of the disease or the extent to which that
biologic activity may be controlled by medicinal
agents. There are multiple influences, and some of
them require a broader model of health.
This broader model of disease leads to a certain
amount of ill-defined unhappiness with the new measures. The patient needs what we were not trained to
give and what, in many instances, we have not the
power to effectuate. We even hear physicians state
that patient outcome is, of course, very important, but
that affecting patient outcome is beyond the scope of
Table 2. Two probably true but contradictory statements
In rheumatoid arthritis, 20 ml of prednisone daily is clinically and
statistically superior to placebo in reducing synovitis, decreasing
morning stiffness, increasing walking speed, decreasing the
number of tender joints, improving grip strength, and decreasing
the sedimentation rate.
In rheumatoid arthritis, 20 ml of prednisone daily, as compared
with non-steroid treatment, leads to increased mortality,
increased rate of development of disability, increased symptoms
due to long term side effects, more hospitalization, and increased
direct and indirect costs of disease.
the medical care system. The medical care system, in
this view, cannot accept responsibility for the general
well-being of the patient; the medical task is seen as
that of bringing scientific biologic knowledge to the
reversal of disease-induced biologiC abnormalities.
Such unfortunate statements contradict the Hippocratic oath and the medical tradition, and seem certain to
accelerate erosion of the public trust.
Moreover, the concept of long term outcome
resulting from multiple health influences takes one
away from the traditional experimental setting (26).
The univariate experiment has formed the basis of
biologic science, and the randomized controlled clinical trial represents the clinical analog of the univariate
laboratory experiment. In the experiment, the experimenter alters a single variable, and then observes the
effects. Multivariate problems involving multiple “risk
factors” are not susceptible to such reductionist analysis, since the various factors interact. Truly long term
analyses using contemporary human experimentation
protocols are not practical; they are neither ethical nor
affordable, and accumulating “dropouts” prevent definitive analysis. Accepting a broader concept of disease unfortunately mandates developing clinical
knowledge within a more complicated framework.
Five areas of recent progress in outcome assessment have combined to provide instruments of
practical utility.
Identification of dimensions
“Outcome” is a general term beyond intuitive
grasp. The thoughtful critic, arguing that outcome is
not measurable as such, will note that combination of
probability of death, disability, and financial impact
into a single number implies that one knows the
relative importance of each. Since such relative importance is not established and since it appears to vary
between individuals and over time, it follows that
outcome cannot be quantitated as a single number.
The concept of outcome as consisting of several
separate dimensions provides a partial escape from
this dilemma.
The question is whether variables which describe the impact of arthritis operate independently, or
whether there are natural aggregations (clusters) of
items which measure aspects of the same conceptual
entity. In a list of potential outcome variables (such as
walking time, cost of doctor visits, the ability to button
a shirt, duration of morning stiffness), are there natural
associations which might represent “dimensions”? Is
there a classification by which we can begin to define
what we mean by “outcome”? Understanding of a
hierarchic structure (Table 3) has begun.
The importance of such hierarchic structures is
not in the detail but in the concept. Outcome as
dimensions which collectively include all impact of
arthritis upon the patient is defined conceptually. In
turn, dimensions may be broken down to subdimensions, subdimensions to components, and components
to individual items. Items are discrete variables which
may be measured with precision. With a hierarchic
model, one may stop at any desired level of generality.
Individual measurement instruments need not measure everything; they need only specify what they are
measuring. For particular purposes a particular dimension, subdimension, or component may well require
elaboration in greater detail.
Until recently, such theoretic structures were
without direct verification. Now, however, comparison of instruments has shown natural clustering of
items around dimensions such as disability. A study
comparing the Arthritis Impact Measurement Scales
(AIMS, Boston University Arthritis Center) and
Health Assessment Questionnaire (HAQ, Stanford
Arthritis Center) instruments has shown that these
separate instruments measure the same dimensions
(27). For example, physical disability scores on HAQ
and AIMS instruments have high inter-instrument
correlation coefficients of 0.91 and inter-instrument
pain correlations of 0.70. Correlations between components of the same dimension are similarly high,
usually ranging from 0.6 to 0.8.
In contrast, correlations between dimensions,
even in the same instrument, are much weaker, usually 0 to 0.3. In other words, discomfort and disability
are only weakly correlated, as are disability and cost,
discomfort and approaching death, and other correlaTable 3.
tions between dimensions. These results confirm clinical intuitions and emphasize the difficulty of speaking
of outcome as a single entity. But they indicate the
possibility of describing outcome by a small number of
discrete dimensions. Disability, discomfort, psychologic outcomes, cost, and death have been identified
as separable dimensions; the full number of dimensions seems likely to be between 5 and 8.
Validation of multiple well constructed
The term “instrument” usually refers to a questionnaire, and the questionnaire is generally patientadministered. Questions also may be asked in a faceto-face interview by a physician, a paramedical
professional, or an outcome assessor, or be asked over
the phone. The well constructed self-administered
questionnaire, however, offers significant advantages.
It is far less expensive, and hence more practical, to
gather information without major expenditure of professional time. Moreover, standardization between
institutions is easier when one does not have to worry
about the different ways in which different observers
might ask the questions. Thus, problems of observer
bias are reduced by the self-administered questionnaire.
The field of survey research is rich with studies
indicating design, format, and wording features that
allow instruments to be, for the most part, culturally
independent and easily understood. The dozens of
steps in questionnaire development, testing, and refinement are subtle but exhausting. Statistical techniques are used to eliminate redundant variables and
to increase the reproducibility of the responses. Openended feedback from subjects helps to clarify areas of
ambiguity. An iterative process continues, usually for
a period of at least 18 months to 2 years, before a final
instrument is ready.
The problem with self-administered question-
A partial classification of patient outcome
Component examples
Item examples
Level ground, stairs, using a
fork, buttoning clothes
Pain Analog Scale, sleep disturbance
Surgical visits, rheumatologist visits, days absent,
lost jobs
Physical function
Walking, eating, dressing
Physical, psychologic
Pain, anxiety, depression
Direct cost, indirect
Doctor visits, lost work
naires is that physicians are skeptical about the accuracy of patients’ responses, and are concerned that
data gathered in this manner may be unacceptably
soft. However, available data point strongly in the
opposite direction-toward the relative inconsistency
the best of
of physician history-taking (18-20 )-and
the new instruments are fully “validated.” Patients
are timed while completing the questionnaire. They
are given the same questionnaire 2 weeks later to
determine retest effects. Results are compared in the
same patients when the same questions are asked by
questionnaire, interview, or over the telephone. Visits
are made to the patient’s home to determine if actual
performance matches responses to questions about
function. Serial administrations of the questionnaire to
the same population over time quantitate sensitivity.
Use of the questionnaires as endpoints in clinical trials
indicates their relative sensitivity as compared with
the traditional endpoints. Patient costs are abstracted
from hospital and doctor records and compared with
services self-reported by patients. Third party carriers’ computer records can be matched with costs
reported by patients.
These same validation studies are repeated in
multiple locations to insure generalizability . A standard set of validation tests is repeated in every study
to insure that results can be validated with a random
subsample of patients. These many steps involve
physician investigators, biostatisticians, epidemiologists, and other staff. If fully accounted, the cost of
developing a single well-validated questionnaire may
be a quarter of a million dollars or more. But the
resulting instrument gives “hard” data; its measurement characteristics are fully validated, and its accuracy is established.
Several such instruments are currently available (4,5,7). In widest use are the HAQ and AIMS
questionnaires, which have been used in thousands of
patients and in multiple settings. Other new instruments share many of the same characteristics to a
lesser degree. In general, validation results affirm the
honesty and intelligence of patients. Well-constructed
instruments minimize confusion, and outright dishonesty in patient response is almost nonexistent.
Development of efficient instruments
An information science aphorism states that the
quality of data is inversely related to the quantity of
data requested. Procedures to reduce redundancy
have allowed construction of much shorter instruments than previously used, thus improving data quali-
ty. For example, if two items on a proposed questionnaire are extremely highly correlated, e.g., above 0.9,
one may be eliminated while keeping essentially all the
information represented by that response. In a series
of steps, the shortest questionnaire sufficient to reliably describe a dimension can be developed.
Our HAQ questionnaire, for example, was reduced from 67 original items to 20 final items grouped
into 8 components. Aggregation beyond this point
resulted in information loss, but performance actually
improved as we pruned questions toward the present
number. The resulting HAQ questionnaire measuring
disability and pain dimensions is completed by the
patient in a period of 5 minutes or less, and may be
scored and entered onto the chart by an office assistant
in less than 1 minute. Hence, outcome assessment
now may be performed in a practical manner, at a cost
considerably less than that of a single laboratory test.
The HAQ instrument has now been administered over
21,000 times to over 7,000 patients (5).
Unfortunately, not all dimensions can presently
be assessed so efficiently. Present cost questionnaires
still require extensive verification time and take much
longer, perhaps 30 minutes, for the patient to complete
(6). The question of the minimal yet sufficient set of
questions to assess psychologic discomfort is, likewise, still an area of controversy (27).
Demonstration of relevance
Outcome measurement is designed to represent
the values of the patient. That is, it is supposed to
measure the truly relevant outputs of the health care
system. Yet until recently, evidence that patient Values actually were being captured by such approaches
was scant. The cooperative study between the Boston
University and Stanford University Multipurpose Arthritis Centers documented relevance to patient Values. A “gold standard” was established by asking
rheumatoid arthritis patients to indicate the total overall impact of their arthritis, on a scale from 0 to 100.
The disability and discomfort dimensions explained
about two-thirds of the total variance represented by
patient self-reported global values. This is an extraordinarily strong result, and indicates a high concordance between what is measured by these instruments
and the values of the patient (27).
Establishment of normal values
Normal, healthy individuals consistently have
scores of 0 on HAQ disability and HAQ discomfort
scales. Patients with rheumatoid arthritis average
about 1.1 on a scale of 0 to 3 in most large rheumatoid
populations, the disability index rising approximately
0.08 per year of disease. Scleroderma patients average
about 0.7. Osteoarthritis patients report lesser degrees
of disability, with the mean in most populations averaging approximately 0.4,and with a stronger relationship to chronologic age of the patient than to the
duration of symptoms. In contrast, discomfort scores
are much more constant between diseases, with osteoarthritis and rheumatoid arthritis patients scoring essentially the same at 0.8 on a 0 to 3.0 scale (5). Thus,
there are now techniques to compare populations and
pretreatment status of groups, against established multicenter figures (43).
refined cost instruments, better validated approaches
to measurement of psychologic health, and instruments which monitor and detect long term therapeutic
toxicities. Specific component areas require elaboration by validated instruments; “patient satisfaction”
and “social networking” represent two such areas.
Moreover, for specific purposes, elaboration of a
particular area of the outcome model shown in Table 3
will be necessary to detect changes restricted to a
particular area. Since good instruments are difficult
and expensive to develop, the most productive approaches will build upon what is available and will
confine new efforts to areas in which existing instruments remain inadequate. Many such elaborations will
be required, and each may serve as a separate contribution toward a growing and coherent ability to measure long term results.
Further steps toward aggregating and combining dimensional results depend upon knowledge of
patient preferences, which is presently lacking. For
example, are the trade-offs between disability and pain
reasonably constant among individuals, or is there
wide individual variation? If there is wide individual
variation, can the preferences of a particular individual
be inferred by knowledge of clinical, socioeconomic,
or other status? Work in these areas to date has been
exploratory, and no hard answers are available (2830).
The sensitivity of these new instruments in
detecting change requires additional evaluation. How
rapidly do measurements change? What is the sensitivity compared with traditional assessment measures?
What techniques are possible to increase sensitivity?
Is sensitivity greater in some areas, such as discomfort, than in others, such as disability?
Finally, how generalizable are the results? Do
they hold across different socioeconomic strata? Are
they culturally independent? Do they hold across
education levels? Ethnic origins? Are there regional
dzerences? Preliminary answers to these questions
are encouraging, but final answers require further
The development of outcome assessment instruments is at an exciting stage. The underlying
concepts challenge traditional views about the role of
medical care, about the magnitude of strictly biologic
inputs as compared with the social and psychologic,
and about the role of the patient in deciding what
should or should not be done. The developing methodologies broaden the model of health to involve many
factors. The issues raised are those with which we
must all become more familiar.
1. Long term studies of disease should utilize
measurement of patient outcome. In a long term study,
there are no advantages, and many disadvantages, to
using surrogate process measures.
2. In any long term study that attempts to
compare the results of two therapies, all the different
outcome dimensions must be considered, since otherwise it is possible to come to an erroneous conclusion
regarding the most efficaceous drug. For example, a
drug might be favored on the basis of its ability to
prevent disability, even though it accelerates death
and results in much time spent sick or hospitalized.
3. In short term studies, outcome should be
measured whenever possible. However, it should not
yet be depended upon, since these instruments’ sensitivity to short term change has not been established.
4. In controlled, prospective clinical trials, outcome measurement should be added to traditional
protocols, in order to anchor the results obtained in
the patient value system and to gain additional experience with the sensitivity of these instruments.
5. In clinical management of chronic illness, the
patient record should develop as a flow sheet, delineating the trends and tempo of the underlying disease.
Health outcome measurement may well prove to be
the most important part of such serial observations.
Clinical chart formats, such as those used by the
American Rheumatism Association Medical Information System, are now available for such purposes.
Outcome instruments need to provide more
comprehensive coverage. There is need for more
The author is grateful to Byron Brown, Sarah Fries,
Deborah Lubeck, Mitchel Seleznick, Yvonne Sherrer, Patricia Spitz, and Donald Young for critical review, and to Ana
Maria de la Cerda for secretarial assistance.
1. Fries JF, Spitz P, Kraines RG, Holman HR: Measurement of patient outcome in arthritis. Arthritis Rheum
23~137-145, 1980
2. Meenan RF, Gertman PM, Mason JH: Measuring health
status in arthritis: the Arthritis Impact Measurement
Scales. Arthritis Rheum 23:146-152, 1980
3. Liang MH, Jette AM: Measuring functional ability in
chronic arthritis: a critical review. Arthritis Rheum
24:80-86, 1981
4. Meenan RF, Gertman PM, Mason JH, Dunaif R: The
Arthritis Impact Measurement Scales: further investigations of a health status measure. Arthritis Rheum
25: 1048-1053, 1982
5 . Fries JF, Spitz PW, Young DY: The dimensions of
health outcomes: the Health Assessment Questionnaire,
disability and pain scales. J Rheumatol9:789-793, 1982
6. Lubeck DP, Spitz PW, Fries JF: The Health Assessment
Questionnaire (HAQ): assessment of personal economic
costs (abstract). Arthritis Rheum (supp1)25:S149, 1982
7. Helewa A, Goldsmith CH, Smythe HA: Development
and properties of a functional capacity instrument in
rheumatoid arthritis (abstract). Arthritis Rheum
(suppl)25: S24, 1982
8. Donabedian A: Evaluating the quality of medical care.
Milbank Mem Fund Q 44: 166-206, 1966
9. Williamson JW: Evaluating quality of patient care: a
strategy relating outcome and process assessment.
JAMA 218566569, 1971
10. Patrick DL, Bush JW, Chen M: Toward an operational
definition of health. J Health SOCBehav 14:7-13, 1973
11. World Health Organization: The First Ten Years of the
World Health Organization. Geneva, WHO, 1958
12. Tugwell P, Bombardier C, Dok C, Sinclair S, Buchanan
W: Quality of life and symptom-relief end-points clinical
trials (abstract). Arthritis Rheum (supp1)25:S24, 1982
13. Feigenbaum PA, Medsger TA, Kraines RG, Fries JF:
The variability of immunologic laboratory tests. J Rheumatol 9:408-414, 1982
14. Bombardier C, Tugwell P, Sinclair A, Dok C, Anderson
G, Buchanan W: Preference for endpoint measures in
clinical trials: results of structural workshops. J Rheumatol 9:798-801, 1982
15. Bombardier C, Tugwell P: A methodological framework
to develop and select indices for clinical trials. J Rheumatol 9:753-757, 1982
16. Bergner M, Bobbitt RA, Carter WB, Gibson BS: The
Sickness Impact Profile: development and final revision
of a health status measure. Med Care 19:787-805, 1981
17. Patrick DL, Bush JW, Chen MM: Methods for measuring levels of well-being for a health status index. Health
Serv Res 8:228-245, 1973
18. Stewart AL, Ware JE, Brook RH: Advances in the
measurement of functional status: construction of aggregate indexes. Med Care 19:473-478, 1981
19. Babbie ER: The Practice of Social Research. Belmont,
California, Wadsworth Publishing Company, 1979
20. Sackett DL, Chambers LW, MacPherson AS, Goldsmith CH, McAuley RG: The development and application of indices of health: general methods and a summary of results. Am J Public Health 67:423-428, 1977
21. Jette AM: Functional capacity evaluation: an empirical
approach. Arch Phys Med Rehabil 61:85-89, 1980
22. Convery FR, Minteer MA, Amiel D, Connett KL:
Polyarticular disability: a functional assessment. Arch
Phys Med Rehabil58:494-499, 1977
23. Lee P, Jasani MK, Dick WC, Buchanan WW: Evaluation of a functional index in rheumatoid arthritis. Scand
J Rheumatol 2:71-77, 1973
24. Fries JF: Education for outcome. J Rheumatol 5:l-2,
25. Fries JF: Aging, natural death, and the compression of
morbidity. N Engl J Med 303:130-135, 1980
26. Coles LS, Brown BW, Engelhard C, Halpern J, Fries
JF: Determining the most valuable clinical variables.
Methods Inf Med 19:42-49, 1980
27. Brown JH, Kazis LE, Spitz PW, Gertman P, Fries JF,
Meenan RF: The dimensions of health outcomes: a
cross-validated examination of health status measurement (abstract). Arthritis Rheum (supp1)25:S149, 1982
28. Landefeld JS, Seskin EP: The economic value of life:
linking theory to practice. Am J Public Health 72555565, 1982
29. Eraker SA, Politser P: How decisions are reached:
physician and patient. Ann Intern Med 97:262-268, 1982
30. McNeil BJ, Pauker SG, Sox HC, Tversky A: On the
elicitation of preferences for alternative therapies. N
Engl J Med 306:1259-1262, 1982
Без категории
Размер файла
753 Кб
measurements, outcomes, understanding, patients, towards
Пожаловаться на содержимое документа