I. Islands of Rationality

The tradition of decision making based on professional paternalism does not deal well with the complex tradeoffs created by modern technology. Rates of elective surgery and other discretionary interventions, which now are determined in large part by practice style and geographic variations in resources, should be determined by the choices informed patients make. To accomplish this “right rate,” patients must participate in the decision making process; to do so, patients must understand what is known, as well as what is not known, about the outcomes that matter to them. Further, patients must be enabled to choose according to their own preferences, even if they ultimately decide to let their doctors decide for them.

This reform will require a new model of clinical decision making. Fortunately, the time is ripe; the escalation in medical spending over the past three decades has created an environment in which it has become possible for patients to challenge the paternalistic role of physicians as agents and sole decision makers. Employers, as payers, have promoted the growth of managed care, which challenges the autonomy of physicians, imposes rules on clinical medicine, and substitutes the managed care company as the decision maker. This transfer of agency power to third parties—payers, insurance companies, and health maintenance organizations—has opened a national debate about the role of the patient in the choice of medical care.

A new model of the doctor-patient relationship is emerging. Shared decision making recognizes the complex tradeoffs that patients must make in the choice of medical care, and addresses the ethical requirement to fully inform patients about the risks and benefits of treatments as well as the need to insure that patients’ values and preferences play a prominent role in medical decision making.

The shared decision making model holds promise for establishing health care markets in which the right rate of service is determined by the choices made by informed and empowered patients. Shared decision making has been implemented in several clinical studies. The studies provide evidence about both patients’ willingness to participate in decisions about their own care, and the rates at which patients choose certain procedures when they are fully informed about the risks and benefits of their choices. Most patients willingly participate in shared decision making, even when, as in the case of early stage prostate cancer, decisions are complicated and difficult because medical science provides no clear evidence that invasive treatment extends life expectancy. The studies of shared decision making also provide initial benchmarks for addressing the question, Which rate is right? The preliminary evidence indicates that the amount of discretionary invasive care now prescribed in the United States might substantially exceed the amount that informed patients actually want.

II. Setting Limits on Hospital Capacity

While shared decision making and patient-level outcomes research hold promise for creating more rational approaches to making choices among available treatments, those strategies do not effectively address global variations in the supply of resources and medical spending. Much of medicine is not driven by well-articulated medical theories that are (at least conceptually) testable by randomized clinical trials or other forms of outcomes research. Hospitalization is often an effort—sometimes a desperate effort—to hold the tide against the inevitable. The quantity of care provided under these circumstances is often limited only by supply. Judgments about how much care is enough must be grounded in an understanding of the relationship between health care capacity and utilization—on how available resources are used. Decisions about how much is enough must also focus on global outcomes. In the case of the supply of acute care hospital resources, the size of the physician workforce, and the level of health care spending, the primary focus should be on the marginal effects of resources and spending on the health outcomes of populations.

The nation is already moving to reduce hospital capacity, although the reduction is happening at a faster rate in areas where resources are already low (such as California) than in areas where resources are high (such as large urban areas on the East Coast). The nature of the relationship between hospital supply and utilization, and the failure to find evidence that more is better, are indications of the validity of using low-resource, low-utilization areas to define reasonable limits. Using such areas as benchmarks, it is possible to estimate the magnitude of potential savings, which could be realized, if high-resource, high-utilization regions were constrained to the level of low-resource, low-utilization regions.

The Dartmouth Atlas web site will allow analysts to use the existing databases, including Medicare claims and the Virginia claims database, to ask these kinds of “What if?” questions. What if resources and utilization in higher-rate areas were reduced to the level of similar, but lower-rate areas? How many hospital beds would need to be closed? How many doctors would be required in the workforce? How many more (or fewer) procedures would be performed? The Atlas raises these questions. The upcoming internet site will make it possible to query the data in additional ways, allowing the analyst to create scenarios for change and to examine the possible repercussions.

III. Setting Limits on the Physician Workforce

The size of the physician workforce in the United States has been determined by factors that have little to do with patient demand for health care, and much to do with federal policy and the needs of training institutions as they are currently structured. In the late 1970s it was widely assumed that the United States faced a physician shortage, which led to policies which encouraged an increase in the number of medical schools and the enlargement of medical school class sizes.

The federal government, through the Medicare program, is the primary source of funding for the training of physicians in residency programs, providing an estimated $70,000 for every resident in training in 1992. The number of specialty residency positions, however, has been determined by the training institutions themselves, aided by an accreditation process that focuses on academic standards, not the numbers of specialists needed by the populations served by the training institutions.

From 1970 to 1996, the supply of clinically active physicians in the United States grew by about 67%, from 113.1 per 100,000 residents to 188.9. During this period, the number of specialists almost doubled, increasing from 63 specialists per 100,000 residents of the United States to 123 per 100,000. The supply of generalist physicians increased from 49 to 65 per 100,000 residents. By 1996, about 66% of the physician workforce were specialists. In Virginia in 1996-97, the numbers of physicians per 100,000 residents ranged from fewer than 100 per 100,000 to almost 250. The average number of physicians in the total physician workforce in Virginia was 170 per 100,000 residents of the state, about 10% lower than the national average.

But how many physicians are really needed? Traditionally, workforce requirements have been focused on the basis of either needs-based or demand-based planning models, both of which are seriously flawed.

Needs-based planning relies on experts to estimate the correct number of physicians to meet need and produce optimal outcomes. Unfortunately, the uncertainties inherent in clinical medicine, rapid changes in technology, and the inevitable failure of outcomes research to keep up with innovation mean that even “experts” are unable to accurately predict the need for physicians.

Demand-based planning assumes that the utilization of care is driven by patient demand; the trends in prevailing rates of service are therefore assumed to be the right rates and are used to project the need for physicians. The evidence that the supply of resources and provider preferences influence the rates of use of care for discretionary services is evidence of the futility of using utilization as a measure of patient demand, and consequently its failure as a method by which to project workforce requirements.

Benchmarking—comparing the workforces in different markets to each other and calculating the excess or deficit in the numbers using a selected area as standard—provides a pragmatic alternative for estimating the requirements for a reasonably-sized workforce. The Atlas series has argued that the hiring practices of large, stable, staff-model health maintenance organizations or the population-based physician supply in regions with efficient delivery systems (such as Minneapolis) should be used as benchmarks for estimating the needed and rational physician workforce for a given area. Benchmarks provide a useful measure of the level of need for several reasons:

• Benchmarks provide working examples of the actual deployment of the workforce, realistic guidelines drawn from successful health care plans or regions. In the case of staff-model health maintenance organizations, workforce configurations have succeeded in competition with fee-for-service markets, often in places such as San Francisco where the numbers of physicians per 100,000 residents serving the fee-for-service market is among the highest in the nation. Regions with efficient health care markets are also useful as benchmarks because their workforce configurations serve entire populations, not just the part of the population enrolled in health maintenance organizations.
• There is little or no evidence that patients are harmed because they are served by health plans or systems with constrained workforces, or live in regions with fewer physicians per capita. Indeed, there is some evidence that the current surgical workforce is more than sufficient to meet patient demand for discretionary surgery.
• Finally, while studies of the global impact of marginal increases in the physician supply on population mortality have not been done and should be encouraged, when it is unclear that spending more is beneficial, common sense argues against the status quo (continuing to produce physicians at a rate which increases the nation’s per capita supply), particularly when the trend in the market is toward managed care.

The Dartmouth Atlas web site will provide analysts with the capacity to compare the physician workforce in Virginia’s hospital service areas to workforces in other areas of the United States, such as Minneapolis, and to make internal comparisons between hospital service areas in Virginia. Such benchmarking can inform the debate over the current supply of physicians—specialists, by specialty, generalists, and the total workforce—and help policy makers and providers assess the current deployment of physicians in Virginia and plan for future workforce requirements.

IV. Focusing the Debate: A Summary Statement

Health care markets in Virginia are characterized by wide variations in the supply of hospital beds and physicians, in price adjusted spending, in rates of hospitalization and surgery, and in the intensity of care during the last six months of life. Practice variations challenge basic assumptions about the nature of the health care economy and theories about how it should be reformed. For decades, the health care debate has taken place against the background assumption that more is better, and that constraint leads inevitably to the rationing of efficacious health care. It is time to re-frame the debate over health care reform to address the fundamental issue of value itself: Which rate is right? How much is enough? and What is fair?

The Dartmouth Atlas series of publications, including the Dartmouth Atlas of Health Care in Virginia, suggests certain conclusions and important hypotheses that bear on the debate:

• Patients should be fully informed about what is known and what is not known about the outcomes of available treatment options, and should be encouraged to choose among those options according to their own preferences.
• Outcomes research should become part of the everyday practice of medicine, and routine follow up of patients according to treatment choice should be incorporated into strategies to improve the scientific basis for clinical decision making.
• It is both safe for patients and in the public interest to adopt the level of acute hospital capacity, physician supply, and Medicare spending of efficient benchmarks such as New Haven and Minneapolis.
• In order to achieve fairness, spending among regions should be equalized on a price adjusted basis.

The impact on the health care economy of reform along these lines would be considerable. When informed patients actively participate in the choice of treatment, there is evidence that patients express less demand for invasive treatments than the amount now being provided.

Extrapolations into the future show that if Medicare spending in regions with higher rates than Minneapolis were brought down to that benchmark, the depletion of Medicare trust funds would be avoided or substantially delayed. Indeed, the Minneapolis configuration of resources suggests a level of health care spending for populations of all ages that is far less than the current average for the United States. Within the savings generated by the judicious reduction of resources and spending to the level of such benchmarks, the nation can find the resources to provide access to health care for all Americans.

Virginia, as a cluster of health care markets, has very much the same patterns of variation observed nationally in other editions of the Dartmouth Atlas series. The use of data prepared for the Virginia Hospital Research and Education Foundation by HCIA to construct this edition of the Atlas demonstrates that variations in Virginia reflect a nationwide phenomenon, and that many of the same remedies could be applied: a “right rate” established for the deployment of resources and workforces; the implementation of shared decision making among Virginia citizens faced with choices between alternative medical interventions, most likely resulting in reduced demand for such services as open heart and prostate surgery, and a more fair and equal distribution of spending for basic health care needs among all residents of the state. This Atlas provides a platform on which analysts, providers, policy makers, employers, payers, and patients can begin a discussion of how to address these important questions.