Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 33
4
What Is Outcomes Research?
JOHN E. WENNBERG
In "The Structure of Scientific Revolutions" Thomas Kuhn suggests that sci-
ence is what scientists do. I want to take my cue from him and will discuss
what our research team is doing that we call outcomes research. I then will
return to Kuhn to make some more general observations about the status of out-
comes research and the emergence of evaluative clinical sciences as a response
to medicine's current predicament.
WHICH RATE IS RIGHT?
For some time, our research group has been conducting a series of investiga-
tions of the outcomes of treatment for benign prostatic hyperplasia (BPH).
These investigations are part of an effort to come to terms with the puzzling
geographic variations in rates of use of medical care. By the mid 1970s, I had
become convinced that these variations exist primarily because of differences in
physician practice style, and that a resolution of the puzzle required direct
inquiry into the effect on outcome of the various opinions or clinical theories
underlying these practice styles (1~. By the early 1980s, Dan Hanley, David
Soule, and Alice Russell had organized leading physicians in Maine into study
groups to investigate the variations (2~. These study groups became the focus
of an alliance between practicing physicians and researchers that made many of
our studies possible.
Of these studies, the assessments of alternative treatments for BPH have
become the most widely known, and it is fair to say that it is in the groping for
solutions to the problems associated with BPH that our understanding of out-
comes research has developed the furthest (3-7~. The BPH study group, orga-
33
OCR for page 34
34
JOHN E. WENNBERG
nized in 1982 by Robert Timothy, consists largely of practicing urologists. Our
BPH assessment team consists of research urologists, medical care epidemiolo-
gists, decision analysts, statisticians, psychometricians, computer scientists, and
experts In interactive videodisc technology (see Table 4.11.
At issue is the treatment of prostatism or obstruction of the urinary tract due
to benign hyperplasia of the prostate gland. BPH is a very common condition,
affecting the majority of men by the seventh or eighth decade of life. One com-
TABLE 4.1 Members of the BPH Assessment Team
Discipline
Members
Location
Biostatistics Klim McPherson Oxford, U.K.
Clinical Decision Analysis Michael J. Barry
Albert G. Mulley
Boston, MA
Boston, MA
Computer Science/Interactive Eric Baumgartner Hanover, NH
Videodisc Technology J. Robert Beck Hanover, NH
Joseph V. Henderson Washington, DC
Harold C. Lyon, Jr. Hanover, NH
Barbara Sasso Hanover, NH
Coralea Wennberg Hanover, NH
Medical Care Epidemiology/ Nicolas Black London, U.K.
Claims Data Studies Thomas A. Bubolz Hanover, NH
Marsha M. Cohen Manitoba, Canada
Elliott S. Fisher Hanover, NH
E. Robert Greenberg Hanover, NH
David J. Malenka Hanover, NH
Dale McLerran Hanover, NH
Aviva Ron Tel Aviv, Israel
Leslie L. Roos Manitoba, Canada
Noralou P. Roos Manitoba, Canada
John H. Wasson Hanover, NH
John E. Wennberg Hanover, NH
Medical Ethics Charles Culver Hanover, NH
Psychometrics/ Tavs Folmer Anderson Copenhagen,
Survey Research Denmark
Floyd J. Fowler Boston, MA
Urology Reginald Bruskewitz Madison, WI
Abraham Cockett Rochester, NY
John A. Heaney Hanover, NH
H. Logan Holtgrewe Annapolis, MD
Ernest Ramsey Manitoba, Canada
Stuart M. Selikowitz White River Jet., VT
Robert P. Timothy Portland, ME
OCR for page 35
WHAT IS OUTCOMES RESEARCH?
35
man treatment for BPH is an operation, a prostatectomy. The rate of prostatec-
tomy varies strikingly among neighboring communities. In some places about
10 percent of men undergo this operation by age 85, while in other communities
the proportion can be as high as 50 percent. In 1987, some 340,000 prostatec-
tomies were performed for BPH at a cost of roughly $4 billion. Another com-
mon treatment for BPH is watchful waiting. In communities with low rates of
prostatectomy, proportionately more men with BPH are treated by this alterna-
tive strategy. Clinical practice in these communities emphasizes the viewpoint
that prostatectomy is an elective procedure, reserved for those with truly bur-
densome symptoms.
The assessment of BPH started with a discussion between the researchers
and the members of the study group, who represented many geographic areas in
Maine with considerable variations in rates of surgery. This discussion focused
on establishing the different points of view or clinical theories behind these geo-
graphic variations. Together with a review of the scientific papers published on
BPH, it uncovered an important and unsettled controversy concerning the indi-
cations for the operation. Many physicians hold the theory that prostatectomy
should be performed early in the course of BPH as a preventive measure. They
reason that if the operation is delayed, the patient will be older and at higher
risk when it finally becomes unavoidable; if the operation is delayed, life
expectancy is reduced. According to this theory, watchful waiting is not a rea-
sonable option for most patients. Other physicians argue that the need for the
operation is not inevitable, that it does not improve life expectancy for most
patients and that the primary reason for surgery is relief of symptoms and
improvement in the quality of life. According to this theory, watchful waiting is
a reasonable option for patients who prefer to live with their symptoms to avoid
the risks of surgery (4-6~.
The assessment team tested this conflict in theory and reached several con-
clusions. Evidence from the literature and from claims data demonstrated that
the preventive theory was incorrect: an operation in patients with uncomplicat-
ed BPH which most patients have very likely causes a slight decrease in life
expectancy. The assessment thus confirmed the opinion of those physicians
who felt the operation was justified because it reduces symptoms and improves
the quality of life. Interview studies with patients before surgery and at three,
six, and twelve months after surgery documented changes in symptom and func-
tional status related to the operation. These studies showed that the value of the
operation for most patients is that it is better than watchful waiting for reducing
symptoms and improving quality of life. However, these gains are available
only to patients willing to take the risks of the operation, which include death,
failure to improve symptoms, impotence, and incontinence. The decision to
undergo the operation is thus highly dependent on patients' preferences for out-
comes and their attitudes toward risk.
This substantial clarification of theory occurred without a prospective clini-
OCR for page 36
36
JOHN E. WENNBERG
cat trial; the steps undertaken and the disciplines involved are listed in Table
4.2. Note the emphasis on study groups with physicians to learn about their the-
ories of treatment, and focus groups with patients to learn about the expecta-
tions and outcomes that matter to them. The literature review and the claims
data provided enough information to define the actuarial expectations for the
operated and unoperated patient (without chronic retention) and to test the pre-
ventive theory of early prostatectomy. We had to depend on the placebo arm of
a few drug trials and a scant series of four studies of BPH patients treated with
watchful waiting for estimates of the natural history of symptoms and the fre-
quency of crossover to operation. We found no satisfactory information on
symptom relief after surgery in the literature, so we conducted our own study.
The results provide convincing evidence that the operation is much more effec-
tive than watchful waiting as a treatment for BPH symptoms. Thus, there is no
dispute about the "main effect" superiority of surgery over watchful waiting.
Decision analysis, used to test the preventive theory, also demonstrates that the
decision to undergo the operation depends on patient attitudes toward impo-
tence, incontinence, and the operative mortality rate. The subjective factors of
risk aversion and personal tolerance for disease symptoms thus emerge as
important elements of rational choice.
The BPH assessment had immediate practical value for clinical decision
making because it clarified controversies, established correct theory, and pro-
viding detailed probability estimates, some of which had not been previously
studied. The effect of practice style on variation in prostatectomy rates was
traced to an incorrect belief in the preventive theory of early prostatectomy and
TABLE 4.2 Synopsis of BPH assessment: Prostatectomy versus watchful waiting
Steps Used in the BPH Assessment
Evaluative Clinical Sciences:
Methods/Disciplines
Identify treatment theories and evaluate
Identify and develop measures of relevant
outcomes from patient's and from
physician's points of view
Undertake non-experimental studies to
estimate (missing) outcomes probabilities
Integrate information from all sources to
test preventive theory and evaluate
importance of patient utilities
Structured review of the literature;
meta-analysis; focus groups with
. . .
physicians
Study focus groups; algorithms
using claims data; patient interview
instruments
Claims data studies; interview studies
of surgery patients
Decision analysis used as "thought
experiment"
OCR for page 37
WHAT IS OUTCOMES RESEARCH?
37
failure to take patient preferences into account when recommending prostatecto-
my. The remedy for unwanted practice-style variations, we concluded, requires
the active engagement of the patient in the decision. It requires informing
physicians and patients of the risks and benefits of prostatectomy and its alter-
native, watchful waiting.
ENGAGING PATIENTS IN THE DECISION PROCESS
We are seeking new ways to engage patients in the decision process and to
make our detailed probability estimates available to patients and physicians in
"real time" for use in clinical decision making. To do this, we have developed a
"BPH Shared Medical Decision Making Procedure" based on interactive
videodisc technology. This interactive program can be used in the physician's
office after standardized information on health status and physical condition is
entered into the data base. The patient is asked to provide information about
symptoms, functional status, and the strength of his feelings about them. This
information allows us to identify the relevant prognostic subgroup to which the
patient belongs so that a patient can be given the best available estimate.
The patient is then shown an audiovisual narration depicting the available
choices, their possible outcomes, and associated probabilities. Interviews with
two physician-patients (one who chose watchful waiting and the other prostatec-
tomy) convey to the patient that there is indeed a choice; if physicians choose
differently, so can their patients. Other interviews are testimonies about the
principal outcomes, including an example of a complication associated with
either choice. The interactive computer feature of the Shared Medical Decision
Making Procedure means that information on the probabilities for outcomes is
specific to the patient's subgroup, according to symptom severity and age. The
patient can also exercise options to learn more about issues of particular con-
cern and can review the presentation. At the end, the patient is given a printed
synopsis to discuss with family and physician. The physician then helps the
patient make a decision.
The BPH assessment conducted so far does clarify controversies in basic
clinical theory and provides better estimates than previously available for symp-
tom status. But important uncertainties remain about some probabilities that are
important for patients with BPH who are trying to evaluate risk. The
attributable postoperative (30-day) risk of death for prostatectomy remains
unclear. For patients who underwent surgery, the sample on which estimates of
symptom improvement and incontinence are based is small (about 400), so the
standard deviations are sometimes large. For untreated patients, the situation is
even less satisfactory. Given a watchful waiting situation, we were unable to
obtain an accurate estimate of the chance of a second episode of acute retention.
This information is important for helping patients who present with acute reten-
tion to decide on treatment. The characteristics of the watchful waiting sub-
OCR for page 38
38
JOHN E. WENNBERG
group at risk for progression to chronic obstruction (with the potential for blad-
der decompensation) are also unclear.
The interactive videodisc program can help fill in these missing probability
estimates because it is also a tool for outcomes research. At the first viewing,
the patient can be enrolled in a prospective study of outcomes based on his
choice of treatment a cohort study that we call a preference trial. For out-
comes research, the patient must return later (e.g., after 3, 6, or 24 months),
regardless of type of treatment, at which point additional information about
health and satisfaction with choice is entered into the data base. These data,
accumulated over time and for many patients, can be used by researchers to
update the information presented to future patients. In this way, the device par-
ticipates in its own update and helps improve the scientific basis of medicine.
We also plan research protocols to improve our understanding of the "framing"
effects of our data displays and to learn more about helping patients make deci-
sions consistent with their preferences. An important objective of these
research protocols is to learn more about, and improve the role of, the physician
as counselor and "cognitive" advisor to patients.
EXTENDING ASSESSMENT TO ALL RELEVANT
TREATMENT THEORIES
In the course of our work we have encountered a number of other theories or
controversies concerning the treatment of BPH. One is an "old" theory that the
transurethral approach to prostatectomy has better outcomes than the open
prostatectomy. Another is that prazosin and other alpha-blockage drugs used to
treat hypertension are useful treatments for symptoms of BPH. We also found
new techniques for treating BPH based on balloon dilation and identified two
promising drugs undergoing FDA evaluation, a less invasive operative tech-
nique (prostatotomy), and a microwave diathermy treatment which is thought to
relieve symptoms by reducing prostate size through the scarring of tissue. As
our thinking about our role has evolved, we have come to see the advantages of
having the assessment team include all alternative treatments, the old as well as
the emerging. We therefore want to address the evaluative challenges each of
these theories presents.
With the exception of the relative effectiveness of transurethral prostatecto-
my (TURP) versus open prostatectomy (an issue I will address below), these
theories are quite new and few data exist on which to base an evaluation. The
need for prospective evaluation thus arises. There is a debate within our group
concerning approaches to prospective evaluation. We began with the orthodox
view that a double-blind, placebo-controlled randomized clinical trial (RCT) is
the optimal strategy for establishing the probability for outcomes of alternative
treatments, but some of us are no longer sure that this is always the case. Our
debate is not about the recognized difficulties of RCTs, such as the issue of
OCR for page 39
WHAT IS OUTCOMES RESEARCH?
39
placebo controls and patient blinding when surgery is involved, the problems of
costs, logistics, and changing technology that can befuddle studies where end-
points require many years of follow-up, or the problem posed by rare diseases
where cases are few and far between. Our conclusions about the importance of
patient preferences for rational decision making, at least in the treatment of
BPH, raise questions about the role of randomization itself.
The first question concerns the ethical requirements for shared decision mak-
ing and the legal requirements for informed consent. When the outcomes of
alternative treatments for a given condition are asymmetric in some significant
dimension say, known differences in the risk of death patient preferences
with regard to what is known about these risks should influence treatment
assignments. This is a strong challenge for the planning of further studies of
watchful waiting versus TURP, where the uncertainty is not even about the
"main treatment effect" probabilities, but about prognosis in certain subgroups
on the watchful waiting arm. When informed about this uncertainty, few
patients who want to share responsibility for decision making are indifferent to
randomization. In this case, the most efficient and ethical way to obtain the
missing probabilities and characterize patient subgroups may well be a prefer-
ence trial the systematic follow-up of patient cohorts where treatment assign-
ments are made according to informed patient choice rather than by randomiza-
tion.
A similar problem may exist in the evaluation of drugs such as prazosin. In
this situation, the "main effect" probabilities concerning symptom relief are
unknown and the rationale for randomization seems more compelling.
However, Phase I safety studies (and the vast experience gained with prazosin
in its use as an antihypertensive drug) show that mortality, over the short term at
least, is lower than that following surgery, and that the drug does not cause
incontinence. Again, when informed, few patients who want to share responsi-
bility for decision making accept randomization.
The unwillingness of informed patients to accept randomization may not be
the main point, however. The more important issue may be which approach
provides the more useful information. To understand this, we need to know a
good deal more about the confounding effects of study design on probability
estimates, particularly for soft outcomes. A pragmatic view is that the informa-
tion generated by clinical trials is useful only to the extent that it helps decide
choice of treatment. When choice is delegated to the physician, the physician
alone makes the connection between the evidence from clinical trials and the
patient. In this case, the probabilities obtained under randomized designs that
ignored or minimized the importance of patient preferences may be the best
estimates for decisions. Under the shared decision making paradigm, however,
it is not self-evident that classic RCTs provide the most accurate information.
Imagine a clinical trial of prazosin in which patients with BPH are randomly
assigned to a preference trial or an RCT. Patient characteristics (co-morbidities,
OCR for page 40
40
JOHN E. WENNBERG
symptom status, demographics, etc.) are obtained before patients are given
information about their choice. A modified version of the Shared Medical
Decision Making Procedure (described above) is needed to present information
uniformly to patients in all cooperating centers. On one arm of the trial, patients
are offered surgery, watchful waiting, or the opportunity to participate in a dou-
ble-blind, placebo-controlled randomized clinical trial of prazosin, based on the
full disclosure of information which the ethics of shared patient decision mak-
ing requires. On the other arm, they are offered a preference trial the opportu-
nity to elect surgery, watchful waiting, or prazosin, based on the same disclo-
sure about what is known (and not known) about the effect of the drug. Patients
who want additional information or advice before deciding on treatment are
counseled by physicians operating under standard protocols who do not admin-
ister treatments. Follow-up to determine outcomes is by observers who are
blinded as to treatment received. I have diagrammed such a trial in Figure 4.1.
The trial would explore a number of interesting problems. Would more
patients elect the drug if they knew they would get it than would elect the dou-
ble-blind trial? Would the subgroup of patients electing the double-blind trial
fairly represent those who elect prazosin with the knowledge that they will get
their choice of treatment? That is, would the probabilities for symptom reduc-
tion and other outcomes estimated in the double-blind trial be the same as those
estimated in the preference trial? It seems to me they would not. Patients who
agree to randomization after a fair presentation of the treatment dilemmas are
very likely different from those who choose prazosin. Moreover, I suspect that
the outcomes associated with freely chosen treatments will be more positive.
SMDP-RCT
Surgery
N__/
C~ Placebo
\
Watchful Waiting
Surgery
Prazosin
\
Watchful Waiting
FIGURE 4.1 Proposed trial of a trial: Preference design versus double-blind randomized design.
OCR for page 41
WHAT IS OUTCOMES RESEARCH?
41
The preference effect I am postulating is the common-sense notion that a treat-
ment chosen by patients who believe it will help them is more likely to have a
positive effect than randomly assigned treatments.
Studies of the effect of patient compliance in double-blinded RCTs provide
some strong hints about the importance of the preference effect. David Sackett
has found five documented instances where patients who comply with treatment
regimens have substantially better outcomes, including survival, than noncom-
pliant patients, even on the placebo arm of the trial. Which set of probabilities
is most relevant to the everyday clinical situation? Under the shared decision
making paradigm, the answer seems intuitively clear. The correct probabilities
are those based on preference trials, where patients knowingly choose treat-
ments on the basis of information about probabilities for the outcomes relevant
to them. However, reliance on preference trials makes sense only if we can dis-
tinguish therapeutic effects from the effects of preference, placebo, and compli-
ance. The trial I have proposed to you, because it also studies patients who
choose watchful waiting and provides a complete accounting of choice and out-
comes on both major arms of the study, may allow us to disentangle these
effects.
Under the paradigm of shared decision making, what has been viewed as a
distinct boundary between clinical research and everyday clinical practice
promises to become blurred. The tools that help patients make informed choice
also are useful for organizing clinical studies. The preference trial approach
may prove more acceptable to practicing physicians and increase the possibility
for large-scale, cooperative studies. It is somewhat of a scandal that, although
prostatectomies have been performed since the beginning of this census and
more than 350,000 are now performed annually in the United States, fewer than
400 patients—our series in Maine- have been followed systematically to study
the "main effect" outcome, the effect of the operation on symptom status. The
information on symptom status and crossover to operation for patients choosing
watchful waiting is based on only four studies of a small series of non-operated
patients.
I believe that many more patients will elect experimental treatments under
preference designs than under randomized designs. If this is so, and if we
become confident about the limitations and advantages of preference trials, this
change will greatly speed the process and reduce the costs of technology assess-
ment as well as provide new methods for Phase IV studies. For example, pra-
zosin is now used extensively in everyday clinical practice in this country, but
in order for its manufacturer to advocate its use for the treatment of BPH it must
undergo an extensive series of RCTs. For all this trouble we would know little
more than its relative value compared to a placebo; we would know nothing
about the probabilities that count in clinical practice, such as its relative effec-
tiveness compared to watchful waiting or surgery, and its value when patients
freely choose their treatments. My hope is that our assessment team can broad-
en the investigation of the relative advantages of preference versus randomized
OCR for page 42
42
JOHN E. WENNBERG
clinical trials across the full spectrum of assessment issues we have encoun-
tered, including the merits of balloon dilation, microwave diathermy, and sever-
al new drugs.
.. . ~ . .. . . . .
· · . .
we also need to consider the value or ootn trial Designs in resolving another
treatment controversy: the relative effectiveness of two surgical approaches to
the prostate the transurethral or the open prostatectomy. Over the past three
decades, BURP has virtually replaced the older open operative techniques. This
replacement, which occurred without prospective clinical trials, reflects the
belief that TURP is a safer, less invasive operation that is effective in the long
term. Since the administrative data bases available to our team extend back to
the 1960s, we have compared the outcomes of these two operations over a con-
siderable span of time. The results indicate that the open operation may be
more effective. Patients with an open operation have significantly lower rates
for stricture and reoperation, suggesting that the more complete removal of the
prostate following the open operation, which is less traumatizing to the urethra,
results in better long-term reduction of urinary symptoms. More puzzling and
disturbing is the unexplained elevation in risk of death in the years following
AMP. In separate studies undertaken in Maine (U.S.A.), Manitoba (Canada),
Oxford (U.K.), and Denmark, we find the risk of death is significantly higher in
the five years following operation.
In view of our concern that the higher mortality following BURP may occur
because those patients are sicker than those undergoing open prostatectomy, we
abstracted data on severity of illness from the medical charts at a large universi-
ty hospital. Even after adjustment, however, the 45 percent higher mortality
rate following TURP was unchanged. The consistency of the findings, and our
failure to explain them on the basis of data available in medical charts, led us to
conclude that differences in co-morbidity do not explain the effect and that a
prospective study is needed to evaluate this problem.
Over the past several months, our assessment team has been in contact with
urologists in various countries, including the leadership of the American
Urological Association (AUA). We are extremely pleased that AUA has urged
its members to participate in a large-scale clinical trial. The clinical trial being
planned will not concentrate solely on the open versus TURP controversy.
Since the available data suggest that BURP does harm, we have reasoned that
this harm, if a true effect, will reveal itself no matter what the comparison
group. The clinical trial thus provides the opportunity for direct comparison of
new technologies balloon dilation, microwave diathermy, and the new
drugs with surgery and watchful waiting. We hope that the clinical trial will
provide the essential information on the outcome of all new treatment theories
relevant to treatment of BPH.
We emphasize randomization in this situation for several reasons. It would
be unwise to base the primary test of open prostatectomies versus BURP on a
preference trial, because the necessary methodologic work to understand the
OCR for page 43
WHAT IS OUTCOMES RESEARCH?
limits and advantages of preference trials has not been done. The controversy
here is about the role of patient risk factors as the confounding variables respon-
sible for higher mortality rates following TURP. Well-conducted and large-
scale randomized clinical trials give assurance that patient differences do not
confound treatment effects. Moreover, it is ethical to randomize if the patient
will accept randomization: our assessment team is uncertain whether the effect
is caused by BURP or by differences in patient characteristics that we could not
detect using the retrospective data available to us. The great majority of physi-
cians do not believe the effect is due to the operation.
As a secondary strategy, however, there may be an advantage if practices not
able to participate in randomized clinical trials (because open operations are not
available) participate in preference trials. Under the hypothesis that patient dif-
ferences do not account for the observed differences in mortality following
TURP versus open prostatectomy, TURP appears to reduce life expectancy.
The effect should therefore be observable in a preference trial where BURP is
compared to a control population undergoing watchful waiting or some other
alternative. An advantage of such preference trials would be the speed with
which additional information on this important problem could be expected to
emerge.
THE EVALUATIVE CLINICAL SCIENCES AND
THE FUTURE OF MEDICINE
I now turn to the more speculative task of attempting to relate outcomes
research to current struggles about the future of medicine. Again, Thomas
Kuhn provides an appropriate context. He teaches that when the rules and pro-
cedures of a science no longer deal effectively with anomalies of theory or
experimental evidence, a new science emerges. Problems are stated in new
ways; new disciplines establish themselves; and new techniques provide solu-
tions not previously possible. I believe that the emergence of the evaluative
clinical sciences, as a set of disciplines distinct from biomedical science, is a
response to the increasingly obvious anomalies of the model on which clinical
and health care policy decision making has been based. That model is the more
or less passive delegation of decision making to physicians who, to use Kenneth
Arrow's term, act as rational agents for patients and society to assure that health
resources are distributed optimally. The rational agency theory depends on two
key assumptions about physicians. First, that by virtue of the strength of
biomedical science and their own clinical experience, physicians know the
prognosis for the various treatments available to a patient with a given condi-
tion. Second, that physicians choose treatments that maximize patient utility,
and that they can wisely choose the right treatment for the individual patient,
disentangling their own preferences to deal with the problem of interpersonal
differences in utility. This requires successful negotiation of E. B. White's
43
OCR for page 44
44
JOHN E. WENNBERG
observation that one man's meat is another man's poison, in order to make
vicarious judgments about which treatment has the highest expected value for
any patient.
The anomalies of rational agency theory are now evident to a widening audi-
ence. It is increasingly apparent that the large investment in biomedical science
has not resulted in a consensus among physicians on the correct way to practice
medicine, nor has it resulted in orderly assessment of treatment theories. Quite
the contrary, biomedical science, by virtue of its prodigious success, appears to
have increased uncertainty and practice-style variations by offering an ever
increasing supply of new technology and new biomedical ideas. This is behind
the shocking variations in costs and utilization between Boston and New Haven,
where the scientific credentials of local medicine are state of the art. It is also
increasingly apparent that patients want to participate in determining their own
medical fates and that physicians have neither the information nor the decision mak-
ing skills to always choose the treatment that will maximize patient utility.
These challenges, emerging at a time of runaway costs, growing consumerism,
and concern about patients' rights, have effectively dissolved long held assump-
tions about the efficacy, the ethical sufficiency, and the legal basis of the dele-
gated decision making model.
The breakdown is apparent in two ways. On the microscale, the impact is
felt in the changing doctor-patient relationship, where physicians face an
increasingly assertive and sometimes litigious clientele. The passive trust that
manifested itself in the patient's willingness to delegate decision making to the
physician and accept what happens as fate has receded. More and more,
patients demand active involvement in the decisions that determine their medi-
cal futures—even when this participation forces exposure to the reality that
physicians do not know all the facts.
This is not to say that patients must distrust their physicians or that a signifi-
cant minority now do. But it does identify a new relationship based on an
active patient role where decision making is shared, not delegated. In this new
role, the challenge to the physician is to learn how to function in what some
have called the cognitive role as an advisor or counselor who understands the
need to act on imperfect information and who helps individual patients under-
stand their own preferences, given the dilemmas the physician and the patient
face together.
It is perhaps on the macroscale of socioeconomics, of reaction to runaway
costs, that the breakdown is most acutely felt by physicians. Understanding the
flaws in the rational agency theory helps explain runaway costs. Demand is not
controlled by professional consensus on "correct practice"; the supply of man-
power and technologies, medical theory and practice style, are in dynamic equi-
librium. Technologic possibilities, the numbers of physicians practicing
medicine, and the costs of care are now such that many politicians, government
officials, corporate officers, and labor leaders perceive further growth in the
OCR for page 45
WHATIS OUTCOMES RESEARCH?
45
health care sector as a threat to other societal priorities. More and more, gov-
ernment and private corporations are intruding into the doctor-patient relation-
ship in their attempts to control demand.
The emergence of the evaluative clinical sciences and outcomes research is a
response to the intellectual crisis created by the demise of the rational agency
theory and the loss of faith that biomedical science, through its own internal
logic, assures effective medical practice. It is an effort to provide the method-
ologies, strategies, and knowledge base needed to support a new model for clin-
ical decision making, one that Albert Mulley has suggested be called the shared
clinical decision making model. The model preserves the notion that consumer
preferences and knowledge about outcomes should serve as the regulator of
patient behavior and of aggregate demand. But it also recognizes the inherent
limits in the ability of physicians to determine vicariously what treatments
patients value most. The evaluative clinical sciences seek to make possible bet-
ter clinical decisions based on a fuller understanding of outcome probabilities,
and active participation of patients in selecting treatments.
The outcomes research agenda thus seeks to do something quite new. It
focuses on the systematic evaluation of all of the outcomes that are relevant to
patients mortality, morbidity, complications, symptom reduction, and func-
tional status improvement as well as the physiologic or biochemical indicators
which have perhaps too often been assumed to be valid surrogates for patient
well-being. It focuses on the systematic evaluation of all (reasonably held) the-
ories and alternative practice styles that are relevant to a particular condition or
illness. It emphasizes assessment teams, organized around specific conditions
with ongoing responsibility for evaluating all treatment options, old and new.
The outcomes research agenda thus removes the "double standard of truth" that
has characterized previous evaluation approaches. These approaches have main-
ly concentrated on new drug evaluations, while ignoring evaluations of the use
of approved drugs in novel ways, of surgical operations and other procedures
such as balloon angioplasty, or of the use of hospitals compared to ambulatory
care settings for chronic and acute conditions. Outcomes research focuses on
the development of methods and strategies for conveying information in ways
that activate the patient as a partner in decision making. The research agenda
includes a broad emphasis on learning to help patients make decisions consis-
tent with their own preferences.
The outcomes research agenda focuses on new strategies and methods for
making inferences to improve the validity and efficiency of evaluative research.
Emphasis is placed on developing a proper balance between non-experimental
and experimental methods for making inferences, and exploring the available
options. Decision analysis is emphasized as a tool for organizing thinking, for
conducting "thought experiments" to evaluate whether a particular treatment
controversy is about the probabilities for outcomes, or whether it is really about
the value of outcomes for patients. And, because the Shared Medical Decision
OCR for page 46
46
JOHN E. WENNBERG
Making model emphasizes the importance of patient preferences, He agenda
accentuates Be need to learn He advantages and limitations of basing treatment
assignments in prospective clinical research protocols on patient preference
rather than on randomization.
REFERENCES
1. Wennberg JE, Bunker JP, Barnes B. The need for assessing the outcome of com-
mon medical practices. Annual Review of Public Health 1980; 1 :277-295.
2. American Medical Association. Confronting Regional Variations: The Maine
Approach. Chicago: American Medical Association, 1986.
3. Wennberg JE, Freeman JL, Shelton RM, Bubolz TA. Hospital use and mortality
among Medicare beneficiaries in Boston and New Haven. New England Journal of
Medicine 1989;321:1168-1173.
4. Wennberg JE, Mulley AG, Hanley D et al. An assessment of prostatectomy for
benign urinary tract obstruction. Journal of the American Medical Association
1988;259:3027-3030.
5. Fowler FJ, Wennberg JE, Timothy RP et al. Symptom status and quality of life fol-
lowing prostatectomy. Journal of the American Medical Association
1988;259:3018-3022.
6. Barry MJ, Mulley AG, Fowler FJ, Wennberg JE. Watchful waiting versus immedi-
ate transurethral resection for symptomatic prostatism. Journal of the American
Medical Association 1988;259:301~3017.
7. Wennberg J. Roos N. Sola L, Schori A, Jaffe R. Use of claims data systems to eval-
uate health care outcomes: Mortality and re-operation following prostatectomy.
Journal of the American Medical Association 1987;257:933-936.
Representative terms from entire chapter:
outcomes research
