This Article Abstract Submit a response to this article Services E-mail this article to a friend Alert me to new issues of the journal

P E R S P E C T I V E S
S P E C I A L I S T S  

15 March 2005  
  

The Physician Workforce Crisis: 
Where Is The Evidence?

If improved health is our goal, then it must be part of the evidence.

By David C. Goodman

ABSTRACT:

Barbara Starfield and colleagues present evidence that having more primary care physicians and fewer specialists per capita is associated with lower mortality rates. These analyses stand in contrast to those guiding national workforce policy recommendations, which use current physician-to-population ratios as a normative reference for future physician requirements. Proponents of physician training expansion need to establish a body of evidence that having additional physicians will lead to improved health and well-being of patients and populations. Research to date, in contrast, indicates that physician workforce levels, particularly of specialists, are not a primary factor in determining health outcomes.

The physician workforce crisis is upon us. The Council on Graduate Medical Education (COGME) has strongly advised that we need to train more physicians. COGME offers several reasons for the anticipated shortage, including population growth, particularly of the elderly; economic growth and its historical correlation with physician supply; and the declining work effort of physicians. To maintain the current physician-to-population ratio, adjusted for these and other factors, the council recommends that the number of physicians entering residency programs increase by 3,000 over the next ten years to partially remedy an anticipated shortfall of 85,000 physicians by 2020. The sufficiency of specialist supply is a particular concern of COGME.¹

The policy and political visibility of the physician workforce crisis would need to rise substantially to compete with existing and prominent health care problems—the future solvency of Medicare, the growing uninsured population, and persistent deficits in the quality of medical care.² These health care system ills have been extensively researched in numerous academic papers and national committee reviews. Implementing the COGME recommendations will be a lengthy and expensive process and is unlikely to occur quickly enough to meet COGME’s goal. Before we are carried away by proposals for medical school and graduate medical education (GME) training expansion, it is reasonable to critique the corpus of evidence that underlies the crisis of a physician workforce shortage.

What constitutes evidence in workforce planning? Evidence has always had a minor role in shaping physician workforce policy. The majority of workforce analyses are exercises in projecting supply and requirements adjusted for various factors. First, doctors and populations are enumerated. Then, future physician-to-population supply ratios are projected using different assumptions about rates of training, retirement, and death and changes in clinical effort. Requirement projections generally use current rates of utilization by population characteristics (that is, age, sex, race, locale, and insurance coverage) to forecast demand as populations increase according to these characteristics. That is the field in a nutshell.³ The “evidence” for the workforce crisis is the current national physician-to-population ratios and medical services use rates projected into the future. These methods never examine the relationship between physician supply and the health of patients and populations. Should we simply accept the connection of these two as self-evident?

To leave untested the assumption that today’s supply of physicians is optimally sized for the population’s health is to reveal workforce analysis as the methodologic underachiever of health care research. Why should we settle for less evidence than for a new drug, screening test, or health promotion program? Even the expansion of the State Children’s Health Insurance Program (SCHIP) has been scrutinized for its effectiveness on outcomes.⁴ The research may be late in coming, be poorly done, or ultimately fail to support the new health intervention, but the need for evidence of improved population is a given.

Testing the effect of workforce levels and composition on population health is well within the capabilities of current epidemiological methods. This research has made some progress, despite the considerable technical challenges and scarce funding.

The evidence of Starfield and colleagues. Barbara Starfield and her colleagues have risen to the challenge by studying the relationship between physician supply and mortality rates.⁵ Like several other studies, this analysis relies upon the natural variation in the workforce across areas. This variation, in itself, should raise doubts about the normative value in using a national physician-to-population ratio to judge requirements. How can it be that regions have twofold differences in supply and appear so similar in other characteristics? Furthermore, how can these supply differences indicate differing physician requirements?

If we conceded that physicians do not locate where health needs are necessarily greater, then populations with higher supplies of physicians might still realize some additional health benefit. Starfield and colleagues found that a greater supply of primary care physicians was generally associated with lower county mortality rates, while a greater specialist supply was associated with higher mortality. If improving the health and well-being of population remains our goal, we need more generalists and fewer specialists, today and in the future.

All observational research is subject to the threats of chance, bias, and confounding, and this study is not without these limitations. Counties are used as units for measuring physician supply, which could lead to biased physician-to-population ratios as patients travel into adjacent counties seeking care. Some counties, such as Los Angeles, encompass very large areas and may contain numerous heterogeneous environments of physician supply. In addition, physician classification in this study is not well connected to mortality rates by a explicit causal pathway. How could the supply of pediatricians (included among primary care physicians) influence mortality rates for conditions that affect mostly adults? The study is cross-sectional, leaving an ambiguous temporal relationship between cause and effect. The absence of patient-level data also limits the confidence that the ecological covariates (county measures) adjust adequately for known confounders. Physician supply is conceptually an ecological measure, but the effects of confounding by age, sex, race, and socioeconomic status are individual characteristics that are summarized, in this study, to an area level. If future studies are to improve the standards of evidence in workforce research, they will need to employ designs that rectify as many of these limitations as possible.

The larger body of evidence. No single study is perfect in its methods or definitive in its findings. Starfield and her colleagues discuss their work in the larger context of other studies that show that more specialists are not always better, and may at times be worse, for patients. These studies employ a variety of study designs, some restricted to narrow populations (for example, newborn or Medicare) to employ multilevel cohort studies that avoid the limitations inherent when studying all physicians in relation to total mortality.⁶ Taken together, the evidence is reasonably consistent—above certain workforce levels, there is little marginal benefit in higher specialist supply. What is affected? Usage rates are higher, as are expenses and the intensity of care. Greater provision, in contrast, of primary care does appear to offer health benefits to patients.

The important advantage of this type of evidence is that it is grounded in the fundamental goals of health care that we hold in common: improvement of the health and well-being of patients. Designing workforce policy to perpetuate current physician-to-population ratios is an arithmetic exercise dressed up as research. If improved health is our goal, it must be part of the evidence. The paper by Starfield and colleagues is a reminder that these investigative efforts continue, even as COGME proposes costly medical education expansion using labor models that are unvalidated by patient outcomes.

The author thanks George Little and Scott Shipman for their suggestions in the preparation of this paper.

NOTES

1. The Council on Graduate Medical Education (COGME) advises the U.S. Department of Health and Human Services (HHS) and Congress on the physician workforce and its training. See Council on Graduate Medical Education, “Minutes of Meeting, July 24–28, 2004,” www.cogme.gov/minutes07_04.htm (3 February 2005). As of 3 February 2005 the latest COGME report had not been officially published and released by HHS. Despite the report’s uncertain future, the findings and recommendations have been cited by Association of American Medical Colleges, AAMC Reporter, November 2004, www.aamc.org/newsroom/reporter/nov04/cogme.htm (3 February 2005); American Academy of Family Physicians, AAFP Direct, 13 August 2004, www.aafp.org/x28808.xml (3 February 2005); and American Medical Association, American Medical News, 13 September 2004,
www.ama-assn.org/amednews/2004/09/13/prsc0913.htm (3 February 2005).
2. S. Heffler et al., “Health Spending Projections through 2013,” Health Affairs, 11 February 2004, content.healthaffairs.org/cgi/content/abstract/hlthaff.w4.79 (4 February 2005); Institute of Medicine, Care without Coverage: Too Little, Too Late (Washington: National Academies Press, 2002); and IOM, Crossing the Quality Chasm: A New Health Care System for the Twenty-first Century (Washington: National Academies Press, 2001).
3. Vector Research Inc., Development of Integrated Requirements for PAs, NPs, CNMs, and Physicians (MDs and DOs): Final Report, Pub. no. VRI-HRSA-15 FR95-01(r) (Ann Arbor, Mich.: Vector Research Inc., 1995); and COGME, Eighth Report: Patient Care Physician Supply and Requirements: Testing COGME Recommendations (Rockville, Md.: Health Resources and Services Administration, 1996).
4. P.C. Damiano et al., “The Impact of the Iowa S-SCHIP Program on Access, Health Status, and the Family Environment,” Ambulatory Pediatrics 3, no. 5 (2003): 263–269.
5. B. Starfield et al., “The Effects of Specialist Supply on Populations’ Health: Assessing the Evidence,” Health Affairs, 15 March 2005, content.healthaffairs.org/cgi/content/abstract/hlthaff.w5.97.
6. D.C. Goodman et al., “The Relation between the Availability of Neonatal Intensive Care and Neonatal Mortality,” New England Journal of Medicine 346, no. 20 (2002): 1538–1544; E.S. Fisher et al., “The Implications of Regional Variations in Medicare Spending, Part 1: The Content, Quality, and Accessibility of Care,” Annals of Internal Medicine 138, no. 4 (2003): 273–287; and E.S. Fisher et al., “The Implications of Regional Variations in Medicare Spending, Part 2: Health Outcomes and Satisfaction with Care,” Annals of Internal Medicine 138, no. 4 (2003): 288–298.

David Goodman (david.goodman{at}dartmouth.edu) is a professor of pediatrics and of community and family medicine at the Center for the Evaluative Clinical Sciences, Dartmouth Medical School, in Hanover, New Hampshire.

DOI: 10.1377/hlthaff.w5.108
©2005 Project HOPE–The People-to-People Health Foundation, Inc.