I.   Background

A.   Purpose

The mission of the Health Resources and Services Administration (HRSA) includes facilitating   and supporting national, State, and local workforce planning efforts.  To help meet this mission, HRSA has supported research on physician workforce issues and maintains the Physician Supply Model (PSM) and Physician Requirements Model (PRM). ^{[1], [2]}

This report describes the various components of the PSM and PRM and presents findings from the literature and original research to provide the context for the data, assumptions and methods used to project the future supply of and demand for physicians.  Projections from the PSM and PRM are presented, and the adequacy of future physician supply is discussed.  This information is important to policy makers, health researchers, educators, professional associations, physicians, and the public to better understand current physician workforce issues and trends and their implications for the adequacy of the future supply of physicians.

B.  Review of Physician Workforce Projection Studies and Approaches

Over the past 70 years, numerous commissions and studies on physician workforce issues have contributed to our understanding of the physician workforce and helped shape policies and programs that affect the supply of physician services.  These studies have documented important aspects of physician supply, the determinants of demand for physician services, and concerns about the maldistribution of physicians by geographic location and by specialty. 

A relatively small number of these studies make original projections of the supply of and requirements for physicians to assess the adequacy of future supply.  It has been stated that whereas projections of the supply of physicians often achieve agreement, estimates of the requirements for physician services rarely do (Tarlov, 1995).  The lack of consensus regarding current and projected physician requirements is in part due to conceptual disagreements—such as whether physician requirements should be based on a clinical definition of need, based on what society is willing to pay for, or based on some other benchmark.  Disagreements also exist regarding the size and impact of trends affecting future supply and demand.  It has often been said that projecting future requirements is an art, not a science.   Despite the uncertainties regarding projections of impending physician shortages or surpluses, the policy responses to these projections have real consequences in terms of the size and specialty composition of physicians being trained; the cost, quality, and accessibility of health care services; and the careers and income of individual physicians. 

In the words of Snyderman, Sheldon and Bischoff (2002) regarding projections of the adequacy of physician supply, “because past predictions were so far off the mark, we need to understand why, before we can confidently predict the [adequacy of the] future supply of physicians (p. 167).” In this spirit, we synthesize and critique the extant literature on the physician workforce and compare the findings in the literature with the data and assumptions used in the PSM and PRM.

The following brief history of physician workforce modeling summarizes key studies that make original projections of physician supply, requirements, or both.  For each study, we give a brief summary of the approach, the key assumptions made, and major critiques of the approach used.  The approach and assumptions used to model specific components of physician supply and requirements are described in more detail throughout this report.

• The Committee on the Costs of Medical Care (CCMC, 1933) was perhaps the first major study of physician requirements in the United States.   CCMC defined physician requirements based on a clinical assessment of population needs.  Using information on the prevalence of disease, the committee determined the number of physician encounters that would occur for the U.S. population in 1929 under the premise that all health care needs would be met.  Combined with estimates of minutes of physician time per patient encounter and average time physicians spent in patient care per year, CCMC determined that the Nation  needed 140.5 physicians per 100,000 population.  A major criticism of the approach used by CCMC is that the requirements estimate did not reflect economic realities.  Even if the physician supply were increased to equal CCMC’s estimates of requirements, economic constraints and patterns of health care utilization would result in some people not seeking needed services (based on CCMC’s definition).  Another critique is that the needs-based approach is open to the biases of the expert panel formed to determine parameters for the model.
• The Surgeon General’s Consultant Group on Medical Education (the Bane report, 1959) was the next major study to assess the current and future adequacy of physician supply.  Using a needs-based approach similar to CCMC, this report predicted a growing shortage of physicians, with a projected shortage of 40,000 physicians by 1975.  This report was instrumental in influencing the Federal Government to subsidize and expand medical schools, and to allow more foreign-trained physicians to immigrate to the United States.

• The Graduate Medical Education National Advisory Committee (GMENAC, 1981) undertook a study of the adequacy of physician supply in response to concerns that efforts to increase the number of trained physicians in the United States  during the 1960s and 1970s would create a national oversupply of physicians. GMENAC used an “adjusted-needs model” that, like the CCMC approach, starts with the premise that physician requirements can be calculated based on the prevalence of disease, an estimate of physician time required per patient to provide needed services, and an estimate of the number of physicians needed to meet the aggregate time requirement for patient care activities.  GMENAC went a step further than CCMC and adjusted downward their needs-based estimates to reflect some economic expectations of the health care system.  The committee concluded that the increased number of graduates from U.S. medical schools (USMGs), combined with the increased number of international medical graduates (IMGs) practicing in the United States, would eliminate the shortage of physicians in most specialties and would greatly alleviate (although not eliminate) the maldistribution of physicians by geographic area.  GMENAC predicted a surplus of 70,000 physicians (13 percent) by 1990, and a surplus of 145,000 (22 percent) by 2000.  In response to this report, the Federal Government reduced general funding for medical schools, but inadvertently increased the financial incentives for teaching hospitals to admit more international medical school graduates to train in the United States (Blumenthal, 2004). The medical need model used by GMENAC was updated in 1990 by interdisciplinary panels of health professionals (Abt Associates, 1991).  Critiques of needs-based modeling also apply to the adjusted needs-based approach.
• The Council on Graduate Medical Education (COGME), formed in the mid-1980s, was created with the mandate to provide advice and recommendations to Congress on the supply and distribution of physicians in the United States.   COGME has issued a series of reports looking at different aspects of the physician workforce.   COGME (2003) used preliminary projections from the PSM and PRM, adjusted for trends in economic growth and other assumptions, to project physician supply and requirements under alternative scenarios.  This study concludes that without a modest increase in U.S. medical school capacity there will be a shortfall of approximately 85,000 physicians, mostly specialists, by 2020.  This projected shortfall contrasts with  COGME’s 1994 report which projected an overall surplus of 80,000 specialists by 2000, and a surplus of 120,000 specialists by 2020.  The dramatic change in findings is attributed to three factors:

1. The 1994 requirements projections are highly sensitive to trends in managed care, following the work by Weiner (1994), but projected trends in the continued growth of the strictest forms of health maintenance organizations (HMOs) failed to appear.
2. In response to reports that there was a growing over supply of specialists, during the mid-to-late 1990s a growing proportion of new medical graduates chose primary care specialties over non-primary care specialties.  Changes in Federal graduate medical education (GME) financing also favored training generalists over specialists.
3. The 2004 projections are sensitive to COGME’s assumptions of the impact of economic growth on future physician requirements, following the work of Cooper et al. (2002) that suggests continued economic growth will increase demand for all physician services, particularly those of specialists.

• Workforce studies responding to the growth of managed care (e.g., Weiner, 1994; Hart et al., 1997;  Gamliel et al., 1995; Wennberg et al., 1993;  and others) found smaller physician-to-population ratios for populations enrolled in HMOs.  Using physician-to-population estimates from a small number of closed-panel HMOS as indicators of an efficient number and mix of physicians, Weiner (1994) projected a surplus of 163,000 physicians in 2000, with the entire surplus consisting of specialists.   Hart et al. criticized Weiner’s study for not including specialist care provided by out-of-network physicians and, after adjusting for out-of-network care, found physician-to-enrollee rates much closer to national physician-to-population rates than those found by Weiner.  One major criticism of using closed-panel HMO staffing patterns to estimate future physician requirements is that the projections are quite sensitive to assumptions regarding the proportion of the future population enrolled in closed-panel HMOs.  Enrollment in closed-panel HMOs grew slower than predicted by many analysts during the early and mid 1990s, partly because of a backlash from consumers desiring greater access to care than provided by closed-panel HMOs.  Another critique of using closed-panel HMOs as a benchmark is that selection bias in who enrolls in HMO plans complicates extrapolating HMO physician staffing patterns to the rest of the U.S.  population.   Weiner (2004) is an update of the 1994 study and finds that physician-to-population ratios at the three pre-paid group practices studied are about 25 percent lower for primary care physicians and 32 percent lower for specialists compared to national ratios. Weiner concludes that the United States could continue to provide adequate care with substantially fewer physicians if the entire system were to function more like HMOs.
• Workforce studies sponsored by individual physician specialties were motivated in part by the controversies surrounding the predictions of Weiner and others in the 1990s of impending surpluses as well as the desire to better inform educators and individual specialties about the likelihood of future demand.  A sample of the physician specialties studied to assess the adequacy of the current and future supply include: emergency medicine (Holliman et al., 1997; Moorhead et al., 1998), endocrinology (Hogan et al., 2001), gastroenterology (Meyer et al., 1996), general surgery (Jonasson, Kwakwa, and Sheldon, 1995), nephrology (Neilson et al., 1997), optometry and ophthalmology (Lee, Jackson and Relles, 1995), orthopedic surgery (Lee, Jackson and Relles, 1998),  physiatry (Hogan et al., 1996), pulmonary and critical care medicine (Schmitz, Lantin, and White, 1998), and urology (Weiner, McDaniel and Lowe, 1997).  These studies investigated a range of workforce issues, including issues related to scope of practice and training requirements.   These studies used different approaches and assumptions, but generally relied on either a utilization-based approach or an HMO benchmarking approach.

• Cooper (2000) and colleagues (2002) use an approach that is a radical departure from the major modeling approaches used in the past.  Cooper’s “Trend Model” approach starts with the premise that the health care system is too complex and diverse to build an estimate of physician requirements “from their component tasks and to standardize them by applying the metric of time (2002, p. 2).” Cooper identifies changes in the level of economic output as the dominant force affecting per capita demand for physician services based on an analysis of national trends from the period 1929 to 2000.  The premise that underlies the economic growth argument is that technological advances provide an almost unlimited demand for physician services—specialist services in particular—that are capped only by our ability and willingness to pay.  Changes in technology, physician productivity, demographics, and the changing role of non-physician clinicians also are important determinants of demand.   The overall conclusion of Cooper’s work is that the United States has a looming shortage of specialists, possibly as many as 200,000 physicians by 2020.  Cooper’s approach has been criticized as placing too much emphasis on observed correlations without a clear causal relationship and for the assumption that historical patterns of physician supply are good proxies of physician demand (e.g., see critiques by Barer, 2002; Grumbach, 2002; Reinhardt, 2002; and Weiner, 2002).  Cooper’s requirements are sensitive to the estimated relationships between economic growth and demand for physician services, and there is no consensus on the size of this relationship; they also are predicated on the idea that there is a relatively high ceiling on what patients are willing to pay for medical care.  Still, Cooper’s work illustrates the possible importance of including economic growth as a determinant of physician requirements.
• The Health Resources and Services Administration maintains a physician supply model and two physician requirements models:  The Physician Requirements Model (PRM) and the Physician Aggregate Requirements Model (PARM).  The assumptions, methods, data and findings from the PSM and PRM are described in more detail in Chapters II and III, respectively.  Both the PRM and the PARM are utilization-based models, but use different approaches and rely on different sets of assumptions.  The critiques that apply to demand-based requirements models in general also apply to these two models.  Namely, that they assume the market is roughly in equilibrium in the base year such that inefficiencies in current utilization and delivery patterns are extrapolated into the future.  The PARM produces national estimates of physician requirements for 19 medical specialties.  The PARM divides the population into 108 categories based on age, gender, race/ethnicity, and insurance status.  Using estimates of per capita utilization of physician services by delivery setting (i.e., inpatient care, outpatient care, doctor’s offices, other) and Census Bureau population projections, the PARM projects the number of physician-patient encounters by setting and medical specialty.  BHPr (2003) presents physician requirements projections under alternative scenarios.  Under the baseline scenario that assumes no changes in health care utilization and physician staffing patterns over time, the PARM projects that between 2000 and 2020 physician requirements will increase by 33 percent.  Under an alternative scenario that assumes modest growth in managed care and modest shifts in patient care from inpatient to outpatient settings, projected requirements increase by 26 percent between 2000 and 2020.

Although HRSA’s models can be used to analyze the consequences of specific health care programs and policies, the main purpose of this report is to present long-term trends and assess the workforce implications of these trends.  The next two chapters of this report discuss how current and projected trends could affect physician supply and requirements.  Chapter II uses the components of the PSM as a framework for discussing trends in important determinants of physician supply.  The discussion also provides a context for each component of the PSM and a synthesis of the relevant literature.  Chapter III uses the PRM to help frame the discussion of physician requirements.  A synthesis of the literature and a summary of original research provide context to understand and critique inputs to the PRM.   Chapter IV compares the supply and demand projections and discusses the adequacy of physician supply.  Chapter V discusses trends in physician compensation and the implications of the PSM and PRM projections on physician compensation.  Chapters VI and VII , respectively, discuss issues of physician gender and race/ethnicity as they apply to projecting the adequacy of physician supply.  Chapter VIII contains a summary of the key findings and concludes with a discussion of areas for future research.

This publication was prepared for the Health Resources and Services Administration by the Lewin Group under Contract  Number  HRSA-230-BHPr-27(2).