Presidential Guest Speaker

Mr. President, distinguished members of the Society, and guests: your president, Dr. James Weinstein, gave me a great honor when he asked that I deliver the first Harry Farfan Memorial Lecture. Dr. Farfan's distinguished career as a leader of medicine is an inspiration to his students and colleagues at McGill University and to many of you in this room. Today, more than ever, medicine needs leaders, and the Society will miss his wisdom and guidance. In his presidential address, Dr. Weinstein provided us with a broad vision of leadership. Although the health care system may seem out of control-indeed, some in the United States say that it is disintegrating before their eyes-and although the expectations of patients and society, spiraling costs, and the pressures for market reform threaten to alter forever the social contract between providers and society and disrupt the bond between physician and patient, we should not despair, but lead. We must be clear about who we are and what we represent; be provocative and proactive rather than reactive; and seek to provide enlightened leadership. Dr. Weinstein asked me to focus my remarks on the challenge posed to the leadership of medicine by documentation of practice variations and the findings of outcomes research. Epidemiologists traditionally have applied the tools of their trade to the documentation of variations in the incidence of disease among populations and the investigation into the causes of those variations. Medical care epidemiologists, on the other hand, investigate the health care system itself. They measure the quantities of health care available to and used by local and regional populations; they chronicle patterns of variation in costs and use; and they study the relationships between the use and outcomes of care. This research addresses directly the problem of spiraling costs, and it also sheds light on the requirements for enlightened leadership at this time of revolutionary change. I believe that the predicament of medicine has occurred in large part because society has asked physicians to assume a role they cannot reasonably be expected to perform-that of rational agent for patients and society. Physicians have been expected to act simultaneously in the best interests of both, assuring that only needed care is prescribed and setting limits on the size and growth of health care markets. This rational agency has been assumed to be sufficient, on a global level, to keep the supply of resources in equilibrium with patient demand. Appealing as this model may be, it has not kept the demand for health care within reasonable bounds. Indeed, a growing body of evidence, much of which comes from medical care epidemiology and outcomes research, points to the opposite effect-that of supplier-induced demand as a determinant of health care use. The challenge to leadership today is to chart a course of remedy. We must recognize the limits to rationality in medicine while seeking to strengthen medical science; reform the doctor-patient relationship to ensure that patient rather than professional (or payer) values and preferences determine the use of care; and work to develop new ways of limiting capacity that ensure the life-long productivity of professionals and the dynamic reallocation of health care resources to more productive strategies. ▪ Dealing With the Limits of Rationality in Medicine Much of medical practice is not governed by explicit theory, much less theory validated by scientific study, and much of medical care cannot be readily subjected to rationalization through patient-level outcomes research. Indeed, much of what we do is better understood as an effort to solve problems, using available resources, than as theory-driven behavior. What I have called the “threshold effect” on the decision whether to hospitalize patients with specific conditions illustrates the murky nature of much of medical decision-making. Physicians regularly are faced with hospitalization decisions: should I, for example, hospitalize this 80-year-old widower with chronic pulmonary obstructive disease and a touch of pneumonia? Or should I treat his problems in an ambulatory setting? Over the whole population of patients, the aggregate decisions about whether to hospitalize such patients have a huge economic impact. Yet for most conditions, there are no rules that determine when to hospitalize; even medical texts are mute on the subject. Our work in medical care epidemiology has made clear that, for most medical conditions, the decision about whether or not to hospitalize a specific patient is strongly influenced by the supply of hospital resources available for use. Although the rational agent theory would predict that professional decision-making would control capacity, practice variation studies show that just the opposite is true. We have found that capacity affects clinical decision-making, and that the available supply of hospital resources will be used, up to the point of scarcity, regardless of how great the capacity might be. This “threshold effect” is strongest for medical, rather than surgical, causes of admission. The cities of Boston, Massachusetts, and New Haven, Connecticut, have provided us with a remarkable example. For historical reasons that have nothing to do with levels of illness or medical need, Boston has many more beds per capita than does New Haven. We can observe the effect of this greater capacity in the treatment of medical conditions such as chronic obstructive pulmonary disease, pneumonia, and diabetes. Hospitalization rates (and per capita hospital costs) for residents of Boston are about 60% higher than for residents of New Haven. In both communities, medical care is delivered by physicians who are faculty members of some of the most prestigious medical schools in the world. Yet we have observed that the supply of hospital resources, rather than science-based knowledge, is the determining factor for use and costs. In the Dartmouth Atlas of Health Care,1 we have been able to show that the same threshold effect of hospital bed supply determines the rates of use of hospitals for medical admissions across the entire United States. The phenomenon is universal, with the exception of the decision about whether to hospitalize patients with a few specific conditions, such as hip fracture. Determining how long the interval should be between visits to the physician's office provides another example of problem-solving behavior that has little theoretical grounding in patient-level outcomes science. When should I schedule a follow-up visit for this 75-year-old retired school teacher who has congestive heart failure and diabetes that is under control? In 6 months? Three months? Again, aggregated over the entire population, the seemingly unimportant distinction has a huge economic impact. Cutting the interval between revisits in half effectively doubles the costs of caring for patients with these conditions, although we have no evidence that it correspondingly improves the patients' health, well-being, or longevity. Again, although the choice of practice pattern has enormous effects on costs, medical texts are unable to supply us with science-based evidence about what should underlie the rules of thumb we use to determine the interval between visits. It is critical for the leadership in medicine to understand the importance of capacity in setting the thresholds for clinical intervention. Whether one is aware of it or not, capacity sets the standard for how medicine is practiced in the community and at what cost. We must address the social policy issues raised by the threshold effect of supply. We must ask, and find answers to, the questions of “Which rate is right?” and “How do we determine the appropriate level of resources and their use?” Answers to these questions depend on choosing the proper point of intervention. This, in turn, depends on coming to terms with the limits of rationality. In medicine, we are inclined to convert uncertainty about the outcomes of care into a clinical research agenda in which specific hypotheses are subjected to experimentation-to randomized clinical trials. For many examples of scientific uncertainty, this is the only way out. But to conduct a formal experiment, the theory to be tested must be reasonably well specified and the criteria for defining who is eligible for inclusion in the experiment tightly defined. Because there is a threshold effect of supply on use across a vast array of clinical conditions and problems, many of our management theories are too implicity associated with supply even to be recognized by clinicians as testable hypotheses. The issue of the appropriate level of supply of hospital resources or physicians is too complex to be approached as a series of patient-level clinical trials; there are too many studies, too many outcomes, too many subgroups, and too many new and changing technologies and theories. In the end, we cannot hope to establish a market that rationalizes supply on the basis of the decision rules and practice guidelines that emerge from outcomes research and needs-based planning. When medical theory is weak and decisions are driven by supply, strategies aimed at making physicians more perfect agents will not succeed in containing capacity. Although the resources might be mustered to establish the specific outcome probabilities for hospitalizing my hypothetical 80-year-old widower (compared with treating him on an outpatient basis) or for scheduling my 75-year-old retired schoolteacher for a 3-month, rather than a 6-month, revisit interval, the net contribution to the resolution of supplier-induced demand would be zero. Our professional bias is to use available resources up to the point of their exhaustion. The nature of the threshold effect is such that any resources that are made available by rationalizing the outcomes of one medical theory will be expended on the huge number of problems that remain unrationalized. The leadership of medicine must come to terms with the likelihood that clinical science will not provide the “condition-outcomes-value for patient” evidence required to rationalize the per capita supply of hospital beds or the numbers of clinically active doctors. Supplier-induced demand cannot be rationalized by this paradigm. The leadership must help society come to terms with uncontrolled costs by helping society set limits; to do that, we need a test of reasonableness. The Dartmouth Atlas of Health Care provides many examples of communities in the United States with lower than average resource allocations where the global (population-based) indicators of health needs and outcomes appear to be as good or better than regions with higher than average resources. We need to study these benchmarks and learn from them; and when more is not better, we need to learn to live with less. ▪ Building Islands of Rationality While much of what is done in medicine may never be rationalized through formal hypothesis testing at the patient level, when medical care theory and outcomes are well specified, outcomes research can be decisive in establishing a scientific basis for the practice of medicine. For a number of common conditions, however, the critical appraisal of the quality of medical evidence reveals many shortfalls. There are serious gaps in the scientific basis for medical decision-making, even though many untested medical theories are strongly-even passionately-held. Where these gaps exist, physicians cannot know the outcomes of care simply because the necessary studies have not been done or cannot be interpreted because of such things as inadequate sample size, poor methodology (e.g., failure to randomize), or because the studies do not measure the outcomes that matter to patients. Building islands of rationality depends on the ability of outcomes research to clarify the theoretical basis for choice and to provide the tools essential to measure the outcomes and subjective states that are crucial to clinical decision-making, particularly when rational choice depends on tradeoffs that belong to the patient. Leadership in medicine must, for this to happen, recognize the importance of undertaking systematic programs of outcomes research to establish the scientific basis for clinical decision-making. My colleagues and I encountered an example of just such leadership. In the 1980s, we began meeting with leading urologists in Maine who wanted to address the reasons for practice variations in their own communities-to discover why the rates of prostate surgery for benign prostatic hyperplasia differed so much from one community to another. It became apparent that, although the decision rules the physicians used for making decisions to operate were similar, their theoretical reasons for undertaking surgery differed. Some of the Maine urologists undertook surgery as a preventive measure early in course of benign prostate disease because they believed that earlier intervention, by preventing complications to the kidney and bladder, would lead to an overall increase in life expectancy. Other urologists were less pessimistic about the overall course of untreated benign prostatic hyperplasia; they believed the major role of surgery was the reduction of symptoms and improvement in the quality of life.2 We searched the surgical literature to look for a critical test of the preventive versus the quality of life theory for prostate surgery for benign disease. No such test was found. There had been no randomized clinical trials. The measurement tools for the “main effect” under the quality of life theory-reduction in symptoms and improvement in quality of life-had not been standardized, and the literature consisted largely of anecdote. Although the competing theories were based on radically different assumptions, the measurement of the need for surgery used by both the proponents' theories was the same: urine flow. Our research group, working with the leadership of the urology community in Maine, undertook a series of outcomes research studies to clarify the theoretical basis for undertaking surgery and to develop measurement instruments that reflected the outcomes that mattered to patients. As it turned out, the preventive theory did not withstand a well-constructed test. Dr. Michael Barry and Dr. Al Mulley combined data from mortality tables, large claims databases, and the literature to conduct an actuarial analysis of the impact on life expectancy of early surgery. They found a slight increase in mortality among cohorts of early stage benign prostatic hyperplasia patients who had been treated with surgery. Their conclusion was that the incidence of life-threatening complications from delayed surgery was too small to the complications of early or surgery. Dr. focus with patients to learn about the outcomes that mattered to the basis of this Dr. Barry and Dr. measurement tools that made it to the symptoms that were of to patients. This us to discover that patients were about the of particularly those patients with levels of symptoms had different subjective to their Some who had symptoms were by them; symptoms were were not We used this to develop that patients' about and their of about The of the tools of outcomes research to some in the theoretical and the basis for and benign prostatic was found to in a reduction in symptoms for most patients, but not for and differed in how they the and For with a level of preferences for surgery or the on the tradeoffs between the for the to which their symptoms and the patients' own about the impact of surgery on although the preventive and the quality of life school of had to urine as the for on the need for urine turned out to have little with The importance of the measurement of symptoms to the of a the which is in clinical use. The importance of patient preferences as the basis for rational choice of treatment is recognized and has been into the guidelines by the for Health Care and Rationality in medicine also depends on into In where the of patients and providers are well the challenge is to the of care to that wanted outcomes are In where rational decision-making the active of the patient, the for leadership is the reform of the doctor-patient me provide examples of practice variation from the Dartmouth Atlas of Health Care that the need for that professional leadership. of clinical have provided evidence that population for the early of is in from among more than or for in this is a clinical by patients, and policy however, is by the that is not into In the Dartmouth Atlas of Health we have been able to show that, the on need and the use of in the United States of the in this example, depends on improvement in the of care. the of the of wanted we must quality that will us the reasons for failure and help us to In and other where the of practice are well understood and the challenge to leadership is to to the of quality improvement and to professional to the in an of The science for clinical decision-making for management is randomized clinical have provided clinical evidence that surgery by is about as as in life expectancy. choice between treatment depends on how about their how they about the of should they of have and the need for or a they The remarkable variations in the use of surgery among different regions in the United which we in the Dartmouth Atlas of Health are with a model of decision-making in which patient preferences determine the use of care. The to rationality in this is rational agency model of the relationship in which patient decisions to their physicians in the that physicians are able to than are patients. The variations in treatment because physicians (and care and do not know the and values of patients, can they know In where patient preferences an rationality the active of the patient in the decision because nothing in the clinical provides the physician with the he or needs to the with can have clinical and and findings but different treatment because they on the subjective they and the rationality in the decision-making particularly for those clinical conditions where provides with different and and where the tradeoffs belong to patients, depends on outcomes research and the of the doctor-patient leadership must understand the and scientific basis for a from the rational agency to the decision-making and medical leadership must a number of conditions in which choice of treatment should depend on patient research and decision-making must establish a new between patient and which can as the basis for into the doctor-patient relationship by or to for Dr. for Leadership The need to set limits on capacity and the to outcomes research and decision-making an professional The members of this society are particularly well to provide a for the leadership. of the major is the of decision-making for from disease. There are strategies for with and problems by to the The and and surgery. research has that surgery many patients a more but with it a small but not that the effort to things better might with of disease, the of to the through their patients with some are a are less Some patients are more about the of treatment than they are about the of their conditions in the opposite The need for leadership in this issue by patients a role in decision-making for disease is in the pattern of variations in the rates of surgical treatment in the United States The numbers of per than between and rate the regions with the rates of surgery are per and the regions with rates per are and are not they can be observed within in the same as is the for and The rates of surgery were much higher in the States and the than in the or in The number of surgery per for in and of the local population, from to point one of the hospital regions in the United and patterns of variation for the surgical management of are of the management of other address the causes of and to these Dr. Weinstein, with Dr. and other members of the patient outcomes research are a for decision-making in research, by of medical theory, the of medical and the patient to his or own preferences in the choice of decision-making the between supply and use the is which rate is by the of the Hospitalization rates for hip are to the local bed supply. Hospitalization rates for variation medical conditions (e.g., are strongly to the local bed supply. point one of the hospital regions in the United the of benign prostatic the of outcomes research and the of decision-making in an market demand well the rates of use of care. The for the professional is decision-making the of available supply, no matter what the among or the of work by the rationality can or decision-making, medical were but that can be directly by the of rationality into the The per capita size and of professional from in another The about need for physicians and associated outcomes of the care they have been and to be based on the that the of physicians are and that work will be of changing and or patient The leadership of medicine needs to recognize the of the model and seek to it with a more appropriate a policy of and the reallocation of effort over the of the be for the The good is that the of medical care under demand a There are many for productive work that is not directly in patient outcomes research, the quality of or the problems of the needs of The is that the to the quality and outcomes of of which are not directly in patient not or even by the many of care. The good is that we are a based on which the to demand at the point of and have the to provide the to the to across the many and complex the health care me you with one more challenge from the Dartmouth Atlas of Health Care The local of from than per residents of to more than in and must be part of which rate is and what is the of between the complex of health care In the numbers of per residents to be higher in the in and in of the the numbers in the and to be The number of in active practice per for in and of the local population, from to more than point one of the hospital regions in the United New