In 1933, the Committee on the Costs of Medical Care (CCMC) published its historic treatise, "The Fundamentals of Good Medical Care." It was a bold effort to describe the dimensions of the physician workforce in precise, quantitative terms. However, for reasons that are inexplicable, this model, a creature of the 1920s, came to dominate physician workforce planning for the rest of the 20th century.

The CCMC was chaired by Henry Moore, Secretary of the Interior, and it included a distinguished interdisciplinary panel. Over the preceding five years, it had issued 27 reports covering various aspects of health care services. However, "Fundamentals" was its major report, and (at $2.50) its most expensive.

The thesis underlying the CCMC's approach was drawn from Olin West, MD, Secretary of the American Medical Association and member of the CCMC. Dr. West identified the outstanding problem confronting medicine as "the delivery of adequate, scientific medical service to all the people," a statement that bears relevance today. Seizing upon this statement, the CCMC undertook to define "adequate" by applying the principles of science.

Basic Tools for Workforce Analysis
The CCMC began its study by systematically cataloguing the host of conditions and disorders that physicians must be concerned with. Focusing on "adequate," it limited its scope to "the essential services," since "medical care is a medical and not an economic concept" (a point that grossly underestimated the impact that the economy would have in the future). It then applied the principles of science to this process by quantitating the prevalence of disease, determining the exact number of physician encounters required for each and designating the time (in minutes) for each encounter. The CCMC's unique and enduring contribution was to establish two basic tools for workforce analysis: reconstructing the system from its component parts and measuring the parts using the "metric of time."

Applying these tools, the CCMC concluded that, in the aggregate, good medical care in 1929 required exactly 283,131 hours of physician time. Assuming that each physician devoted 40 hours per week, 50 weeks per year to these tasks, "less than the present heroic working schedule," the system would need 140.5 physicians per 100,000 of population, a figure that was 10 percent greater than the existing supply. Moreover, it concluded that 18 percent of these physicians should be specialists in one of the 10 specialties then recognized. This exhaustive exercise was presented in 302 pages of text and tables, but it included a warning that, if the reader "expects to find here the finality of judgment and precision of detail, he is doomed to disappointment." And doomed we have been.

GME Model of Workforce
Almost 50 years later, another creature of government, the Graduate Medical Education National Advisory Committee (GMENAC), reached into the past for a model that it could use to determine the number of physicians that were required in each of the specialties. While retaining the CCMC's core methodologic tools, it modified the approach of the CCMC to create its "Adjusted Needs Model."

Like the CCMC's earlier model, GMENAC's was based on "dissecting the intricacies of the pluralistic health care system" from an epidemiologic perspective. It studied the prevalence of disease and used expert panels to build a consensus regarding the proportion of individuals with each disease who need treatment, the time required for that treatment and the number of physicians necessary to provide that time.

As was evident in the CCMC's model, GMENAC's dependence on disaggregating and reconstituting the universe of care, coupled with its need to assign the metric of time to both the elements of care and the effort of physicians in providing it, seriously handicapped its ability to determine what actually was occurring. However, GMENAC went one step furrther. Failing to heed the advice of the CCMC that "it is impossible to determine, once and for all time, the services that will represent an adequate application of medical knowledge and skills to the needs of the people," GMENAC proceeded to extrapolate its calculations 20 years into the future, predicting that there would be a 30 percent surplus of physicians in the year 2000. Although this prediction proved to be excessive, it has had a strong influence on health policy discussions.

COGME's Approach to Workforce
With the increasing availability of data on clinical practice in the early 1990s, GMENAC's successor, the Council on Graduate Medical Education (COGME), adopted the Demand-Utilization Model for workforce planning. Rather than relying on epidemiologic data, this model assessed the requirements for physicians based on actual measurements of services provided. For this, it drew upon the resources of national databases, such as the National Ambulatory Medical Care Survey, the National Hospital Discharge Survey and Medicare claims data. However, like its predecessors, the Demand-Utilization Model attempted to recreate physicians from their component tasks and to standardize them by applying the metric of time. Not surprisingly, it, too, failed. For example, only four years ago, COGME projected that there would be a 47 percent surplus of specialists in the year 2000.

As managed care emerged, a new avenue of analysis, the requirements model, appeared. It was based on physician utilization in staff/group model HMOs. These seemingly "closed systems" should, it was reasoned, be able to account for all of the care provided and all of the time necessary for physicians to provide it. Moreover, by applying the metric of time, all of this could be expressed as full-time equivalent (FTE) physicians.

However, the HMOs from which this model was built represent a small and shrinking segment of clinical practice, and the assumptions and extrapolations required to describe the entire system from this narrow pedestal are complicated and tenuous. As a result, the conclusions have been far from the mark. In what was characterized as "the most complete forecast to date," carried out on behalf of COGME in 1994, Weiner predicted that 65 percent of all specialists (165,000 physicians) would be in excess supply by the year 2000, a prediction that led to a call for the closure of 20 U.S. medical schools, a sharp decrease in specialty training and the curtailment of funding for international medical graduates.

Thus, beginning with the CCMC's report in 1933 and continuing through GMENAC's report in 1980 to COGME's various reports in the 1990s, physician workforce studies have been dominated by a linear, mathematical mode of thinking based on dissecting and reconstituting the health care system and standardizing its components according to the metric of time. The imprecision in this process is legion, and the errors associated with applying it to a multiplicity of diseases, an array of services and a diversity of both patients and physicians are enormous. Using it to project future needs further compounds the error. Moreover, it does so in ways that are not always apparent in the final product.

This has been the American way for seven decades. Is there an alternative?

New and Improved: The Trend Model
Over the past year, I have had the great pleasure of working with an expert panel comprised of representatives from the Council of Medical Specialty Societies (CMSS) to consider how to best conduct studies of the specialty workforce. The deliberations of this panel recently concluded by endorsing an alternative model that I call the "Trend Model." This model does not dissect and reconstitute the current system but, rather, accepts its complexity and diversity. It projects future demand by using a process of trend analysis. However, rather than using the metric of time, it employs a statisticcal approach, assigning a vector, magnitude and probability to each of the trends considered.

The dominant trend is the economy. Even in 1933, the CCMC recognized that "compelling economic forces" influence the distribution of physicians and that "the practice of medicine depends upon the consumers of medical services as much as on the practitioners of medicine." These economic forces are even stronger today. But other trends also are important, such as technology, demographics, physician productivity and the changing roles of nonphysician clinicians. Moreover, these trends are complex and interdependent, and most are influenced by the underlying economic dynamics.

Building from current realities, the Trend Model is constructed in a manner that is deterministic and objective. It also allows the introduction of value judgments concerning issues such as costs, access and training. However, by compartmentalizing such judgments, it frees the basic analysis from bias and permits the juxtaposition of alternative formulas for future workforce needs.

Looking to the Future
Attempts to analyze the physician workforce reveal how imperfect the science is. It must accommodate to inconsistencies and ambiguities in existing data and uncertainties about the future. Approximations of that future, however, are needed to guide the important training decisions of today. I believe that it's time to move beyond the workforce models of the 1920s and to adopt contemporary approaches that reflect the complexities of modern health care. The Trend Model could fill that need. But, as a wise sage observed, "prediction is very difficult, especially when it involves the future."

Richard A. Cooper, MD, is Professor of Medicine and Health Policy, Director of the Health Policy Institute at the Medical College of Wisconsin, and principal investigator for the CMSS Specialty Workforce project.