Access to primary care is associated with improved health outcomes1 and decreased costs,2 yet 30% of the U.S. population reports difficulty accessing primary care physicians (PCPs).3 Many researchers predict a worsening PCP shortage, especially in rural and underserved areas,4,5 as demand for physicians increases because of population growth, population aging, and increased use among people newly insured through the Affordable Care Act.4 To increase regional production of physicians to help alleviate local shortages, stakeholders have advocated for changes in both undergraduate medical education (UME) and graduate medical education (GME).6,7
In 2006, the Association of American Medical Colleges8 (AAMC) recommended a 30% increase over 2002 enrollment in MD-granting (allopathic) medical schools by 2016 to accommodate predicted physician shortages, particularly in locations that had or were expected to have large population growth. Expansion of MD-granting schools has since focused on increasing matriculation within existing schools and establishing new schools, but this growth is poorly correlated with population expansion and other indicators of regional need.9 At the same time, DO-granting (osteopathic) medical schools doubled their enrollment.10 Overall, the total number of first-year medical students grew by nearly 50% between 2002 and 2014.11 Yet despite these UME enrollment increases, little new federal or state government investment in GME has occurred—particularly in primary care programs—and thus concerns remain about the capacity of the GME system to train graduates to meet population health needs.11,12
GME represents a public monetary investment, with annual federal and state funding exceeding $15 billion as of 2014.12 The Balanced Budget Act of 1997 placed a cap on the number of Medicare-funded residency positions within each existing teaching hospital in an attempt to control spending. However, between 2002 and 2012, GME expanded at a rate comparable to before the cap, with total GME trainee growth of 17.5% over the 10-year interval.11,12 New GME positions, funded largely by state or other nonfederal sources, have been created disproportionately in subspecialty training and non-primary-care specialties,13,14 resulting in 13% growth in trainees entering GME programs that culminate in initial board certification and 40% growth in those entering continuing GME programs (i.e., fellowships).15 These changes have increased opportunities for primary care graduates to subspecialize and have decreased those predicted to solely practice primary care.16
As calls continue for the physician workforce to increase accountability to local needs,7,12,17–19 it is critical to understand mechanisms to align UME and GME training to meet those local workforce needs. The AAMC reports that 47% of GME graduates remain in the state where they completed their last training.20 Additionally, those who complete medical school and residency within the same state have even higher levels of retention in that state.20,21 Thus, states have the opportunity to make significant impacts on the local physician workforce through strategic changes in GME. In this study, we aimed to answer the following questions: Is primary care GME (PCGME) growth within states correlated with workforce needs or with UME expansion? Does PCGME growth correlate with population growth?
We described the total change in PCGME trainees (a net PCP production estimation, described below) by state as a function of changes in trainees in primary care residencies and fellowships for those specialties. We then compared changes in PCGME trainees between 2002 and 2012 with the 2000 PCP-to-population ratio and with the change in UME graduates and the population growth over the same period.
To determine changes in PCGME trainees between 2002 and 2012, we used 2014 American Medical Association (AMA) Physician Masterfile data. The Physician Masterfile contains data on filled GME positions, as reported for each state by the Accreditation Council for Graduate Medical Education (ACGME). (It does not include trainees in osteopathic-only training programs, and thus, they are not included in this analysis.) To estimate this state-level investment in PCP production, we summed the postgraduate year 1 (PGY-1) trainees in core primary care programs—that is, PGY-1 trainees in three-year programs in internal medicine (excluding preliminary/transitional positions), family medicine, and pediatrics (core trainees)—and then subtracted the first-year subspecialty fellowship trainees available to those same specialties for the same year (fellowship trainees).16 This estimates net PCP production, and these net PCPs are referred to as PCGME trainees in this article.
Fellowships are entered after the terminus of a three-year primary care residency program, and there are both ACGME-approved and non-ACGME-approved fellowships. Options for ACGME-approved subspecialization vary depending on the preceding specialty: internal medicine has 22 fellowship options,22 family medicine has 4,23 and pediatrics has 19.24 Board certification in many internal medicine and pediatric subspecialties does not require concurrent internal medicine board certification; however, in family medicine, all subspecialties are considered to be certified areas of concentration and require family medicine board certification to be active.
Depending on the type of fellowship completed, individuals may transition to practicing clinical medicine related to only the subspecialty or to combining specialty care with primary care services. The distinction between providing primary care services versus primary care is important. While many subspecialists provide primary care services, they do not provide the first-contact, continuous, comprehensive care that PCPs do.1,2 It is important to note that we were unable to account for, or exclude, PCPs working as hospitalists; it has been reported that some GME institutions retain less than 10% of their internal medicine residents in primary care as a result of losses to fellowships and hospital medicine.19
For each state, we recorded the percent change in PCGME trainees between 2002 and 2012 and compared it with the 2000 PCP-to-population ratio. (For the purposes of this analysis, we considered the District of Columbia to be a state.) Additionally, as population changes adjust the impact of each physician in a given state, we examined how changes in population growth influenced PCGME growth. We determined the number of active PCPs in direct patient care in each state using 2000 and 2010 AMA Physician Masterfile data. We excluded residents, fellows, and physicians who were classified as retired, semiretired, temporarily not in practice, or inactive for other reasons. We used the number of active PCPs and state population data from the 2000 and 2010 U.S. Census to calculate the PCP-to-population ratio for the respective years. We used the 2000 and 2010 state population data to weight correlations and rates of change per population, as well as to determine population growth.
For each state, we also compared the percent change in PCGME trainees with the percent change in UME graduates between 2002 and 2012. We used publicly available data from the AAMC25 and the American Association of Colleges of Osteopathic Medicine26 (AACOM) to determine the number of medical school graduates in 2002 and 2012 by state. Four states do not have medical schools, but they partner with the University of Washington School of Medicine (UWSOM) to provide state-subsidized positions within UWSOM classes for their residents through the WWAMI (Washington, Wyoming, Alaska, Montana, and Idaho) Regional Medical Education Program. We obtained medical school graduate data for these states directly from the UWSOM.
All analyses were performed using Stata 13 (StataCorp LP, College Station, Texas). This study was exempted by the American Academy of Family Physicians institutional review board.
From 2002 to 2012, a total of 54 core trainees were lost and 625 fellowship trainees were added, with the summa tive effect of a net loss of 679 PCGME trainees. This represents a PCGME trainee national decline of 7.1% and median state decline of 2.7%. Thirty-one states had positive core trainee growth, and 40 had positive fellowship trainee growth (see Supplemental Digital Appendix 1 at http://links.lww.com/ACADMED/A416). Fellowship trainee expansion was greater than core trainee expansion in 29 states. In Washington, DC, and Maine, both core and fellowship trainees decreased.
Figure 1 illustrates the growth of core trainees compared with the growth of fellowship trainees among the 25 states that produced 85% of PCGME trainees. Of these 25 states, 7 (28%) had net growth of PCGME. For most of these 7 states, this resulted from greater growth of core trainees compared with fellowship trainees. The other 18 states (72%) had net loss of PCGME, the reasons for which varied. Many states decreased core trainees while adding fellowship trainees, while other states had no or minimal core trainee growth that did not maintain pace with large fellowship trainee growth. In a national context, in 2002, 43.5 PCPs per 1 million people began residency training, of whom 9.4 went on to fellowships—and thus represent a loss from primary care (Table 1). In 2012, there were fewer core trainees and more fellowship trainees, creating a net loss of 5.2 PCPs for each 1 million persons from 2002 to 2012.
With regard to population need, states varied in the robustness of their primary care workforce. The median number of PCPs per 100,000 people in 2000 among all states was 63. PCGME trainee growth over the study decade in relation to the 2000 PCP-to-population ratio varied widely across states without an apparent pattern or trend (r = −0.06; Figure 2, panel A). Of the 15 states with a PCP-to-population ratio below the median, 7 had PCGME decreases > 10%, while 8 had PCGME growth > 10%.
The total number of annual medical school graduates increased nationally by 19.0% from 2002 to 2012, with a median state increase of 12.5%. In many states, large changes (increases or decreases) in UME graduates did not correlate with corresponding growth in PCGME (r = 0.17; Figure 2, panel B). More than half of states had ≥ 10% growth of UME graduates, yet the majority of these states had declines in PCGME trainees.
Between 2002 and 2012, the U.S. population grew by 9.7% with a median state population growth of 7.8%. All but one state (Michigan) had positive population growth, with 35 states having > 5% growth and 20 states having > 10% growth. However, among the 25 states accounting for 85% of PCGME production, no state had net growth of PCGME trainees once corrected for population growth (Figure 3). Once adjusted for population growth, the national PCGME decline was 15.3% (unadjusted decline = 7.1%), and the median state decline among all states was 9.7% (unadjusted decline = 2.7%).
Despite states’ critical role in addressing local physician workforce needs, our findings show little correlation at the state level between PCGME growth and ostensible indicators of population need, such as PCP-to-population ratio or UME expansion. Further, the rate of state-level PCGME expansion has been largely unable to keep pace with population growth. Despite a federal funding cap on new GME positions in existing teaching hospitals, these positions have continued to grow, principally among subspecialty training programs (i.e., fellowships). Toward the end of our analysis period, the Health Resources and Services Administration Primary Care Residency Expansion and Teaching Health Center (THC) programs established numerous PCGME positions, although not enough to gain net core primary care trainees. Since 2012, there has been a considerable increase in the number of THC trainees27; however, if subspecialty (i.e., fellowship) positions continue to grow and be filled at a rate similar to the rate of growth in this analysis, it is unlikely that the growth in THC trainees will be large enough to offset the overall declines in PCGME trainees. Furthermore, ongoing funding for these programs is tenuous,28,29 and this modest federal contribution has not yet shifted the GME system away from specialization to address local workforce needs or the current and predicted national deficit of PCPs.
That UME growth, funded in many states under the banner of primary care expansion,30 showed little correlation with growth or even stability of PCGME is concerning. UME total enrollment grew 49% between 2002 and 2014,11 which is much larger than the graduate growth of 19% between 2002 and 2012 seen in our study. The consistency of the GME trends over the decade studied suggests even more growth of subspecialty training positions and continued reduction in primary care output after our study period.15 It has already been shown that UME expansion has little correlation with population growth or other indicators of population need9 and that the state of GME completion is a better predictor of state retention than is state of UME completion.20 Thus, both UME and GME growth have been uncoordinated with each other and with population need. The goal of public funding for UME and GME is to produce physicians to meet local and national health care needs, and yet state efforts have not generally coordinated physician pipeline development with local need.
As proxies for population need, we found that PCGME growth was not correlated with PCP-to-population ratio in 2000 or population growth from 2000 to 2010. However, in considering a publicly funded system that is responsive to local need and how this relates to PCGME, the question of “what specialty distribution would meet population needs?” arises. This question is not easily answered. Populations do need access to a wide array of specialty services, but it is unclear what distribution of specialists leads to improved outcomes. Conversely, a strong, integrated primary care foundation has been shown to produce better overall population health and health outcomes.1,2 Our inclusion of fellowship-trained family physicians likely underestimates their primary care contributions based on limitations discussed in the Method section; however, our inclusion of hospitalists as PCGME trainees far overestimates the net primary care outcomes.
Within the constraints of the current primary care system, there have been various models of delivering primary care that affect population access. For example, some patient-centered medical home models have reduced panel sizes,31 and many clinicians have left traditional practice for direct primary care, which even further reduces panels.32 These panel contractions emphasize primary care need, a need that will only grow from the current GME system. Some have argued that shifting work to nonphysicians, such as nurse practitioners (NPs) and physician assistants (PAs), may improve this burden.33 The optimal use of these clinicians may not be as physician substitutes, although this is frequently their role. Further, although the NP and PA workforce in primary care continues to grow, the majority of both NPs and PAs are now going into subspecialty services.34
This analysis did not include two special subsets: trainees in osteopathic-only residency programs and graduates of international medical schools (IMGs). Nationally, there are approximately 675 PGY-1 primary care trainees in osteopathic-only programs per year.35 Regarding state investment, as a number of osteopathic schools employ an interstate educational model where preclinical education is held at centralized locations, followed by clinical training in different states, it is unclear how graduates of these schools should be counted. Further, in 2014, the American Osteopathic Association, AACOM, and ACGME agreed to a single accreditation system for GME training as of 2020.36 This venture will give osteopathic graduates greater access to subspecialty training, potentially further eroding PCGME trainees. With regard to IMGs, each year thousands are accepted into GME programs, often in primary care. These GME positions were included in our analysis, but represent an influx of medical school graduates not accounted for in the state-level analysis, which further highlights the lack of coordination between the UME and GME pipelines.
Our proxies for population need did not account for all possible indicators of need as determined by states; it may be that PCGME or specialty growth relates to an unmeasured but defined need. For example, we did not examine whether state retention was correlated with need, and states with high retention rates may have lower PCGME growth despite large medical school enrollment increases. Other states with large influxes of IMGs or trainees in osteopathic-only programs may also have reduced PCGME growth. Finally, our analysis examined only actual trainees. It may be that states have PGY-1 primary care positions that remain unfilled; however, this would also be true for specialty and subspecialty positions.
GME position growth is affected by caps on Medicare funding, causing reliance on funding from Medicaid, the U.S. Department of Veterans Affairs (VA), or private sources for new growth. Georgia, which experienced one of the largest declines in PCGME during our study period, recently began to incentivize GME development in needed specialties and locations using workforce data and short-term start-up funds.30 Conversely, Florida and Texas, which have added numerous medical schools yet experienced substantial declines in PCGME, are among the states citing the challenges of additional administrative expenses and recruiting teaching faculty support.37 If the GME system were able to create financial partnerships outside Medicare and their federal caps, it is possible that residency programs could sustainably establish the positions required to meet workforce needs and to absorb new medical school enrollees.
Creating state-level financial partnerships may help address local workforce challenges and support pipeline creation. More than $4 billion of Medicaid funds are currently invested in GME.12 States have flexibility in Medicaid GME payments as well as the ability to assess local workforce challenges and allocate funds accordingly. Thus, these state-level partnerships (e.g., with Medicaid) could ensure that changes in GME structure—with regard to access, distribution, and specialty needs—are better aligned. Although many states have invested Medicaid dollars into GME, states often follow Medicare GME funding procedures, missing the opportunity for strategic allocation and strengthened local accountability.38 Although Medicaid provides less GME funding than does Medicare, these funds have the potential to make meaningful differences in local GME reform by helping shape the physician workforce to meet state needs.
The VA also funds many GME positions.39 In 2014, the VA announced funding of 1,500 new residency positions, emphasizing primary care and mental health and targeting facilities with a shortage of physicians, in rural locations, or in Health Professional Shortage Areas. Physicians who train in rural and underserved areas are more likely to remain in similar settings for practice,17,40 and a majority of family medicine graduates remain close to their GME site after graduation.41 Thus, establishing PCGME in primary care shortage areas may help ameliorate physician maldistribution. As of the time of writing, these VA positions were in the process of being distributed. We believe they have the potential to have a large impact on local populations.
The creation of pipeline programming and use of public funds requires accountability to ensure that workforce targets are met and physician distribution is improved. Few states systematically collect data regarding workforce assessment, and those that do have no requirement to use these data to support GME expansion.30 We suggest that GME policies support state workforce needs and require continuous evaluation and feedback to inform decision making and ensure the effectiveness of these publicly funded investments.
Our measures of need were not intended to be comprehensive so much as exemplary and based on previous state-level assessment.9 Because of the known lag time of the AMA Physician Masterfile, we may be overcounting physicians who are no longer in practice or undercounting the total number of PCGME trainees. Because the Physician Masterfile relies on input from the ACGME for the number of filled residency positions, trainees in programs that did not submit data to the ACGME would be uncounted. Finally, as stated in the Method section, we did not account for those entering hospital medicine, a specialty that had large growth during our analysis period42 and thus exacerbates the decline in PCGME trainees.
Recent growth in PCGME trainees is not correlated with state need. It is not correlated with the number of PCPs per population, with increases in UME graduates, or with state population growth. Nationally, total PCGME trainees declined 7.1% (unadjusted for population growth) or 15.3% (adjusted for population growth) from 2002 to 2012. Most PCGME trainee growth occurred in fellowship trainees and had the net effect of eroding primary care output. Existing data are helpful in tracking population need, the impact of workforce policies, and institutional accountability. States should capitalize on opportunities to create explicit linkages between medical education, training, and population need, to strategically allocate Medicaid GME funds, and to monitor the impact of workforce policies and training institution outputs.
Acknowledgments: The authors wish to thank Susan Skillman for her help in obtaining data pertaining to state-level enrollment at the University of Washington School of Medicine.
1. Starfield B, Shi L, Macinko J. Contribution of primary care to health systems and health. Milbank Q. 2005;83:457502.
2. Macinko J, Starfield B, Shi L. The contribution of primary care systems to health outcomes within Organization for Economic Cooperation and Development (OECD) countries, 1970–1998. Health Serv Res. 2003;38:831865.
3. Hackbarth GM. Pursuit of an expanded physician supply. N Engl J Med. 2008;359:764.
4. Petterson SM, Liaw WR, Phillips RL Jr, Rabin DL, Meyers DS, Bazemore AW. Projecting US primary care physician workforce needs: 2010–2025. Ann Fam Med. 2012;10:503509.
5. IHS Inc. The Complexities of Physician Supply and Demand: Projections From 2013 to 2025. Prepared for the Association of American Medical Colleges. 2015.Washington, DC: Association of American Medical Colleges.
6. Rockey PH, Rieselbach RE, Neuhausen K, et al. States can transform their health care workforce. J Grad Med Educ. 2014;6:805808.
7. Mullan F, Chen C, Petterson S, Kolsky G, Spagnola M. The social mission of medical education: Ranking the schools. Ann Intern Med. 2010;152:804811.
8. Association of American Medical Colleges. AAMC statement on the physician workforce. https://www.aamc.org/download/55458/data/workforceposition.pdf
. Published June 2006. Accessed November 8, 2016.
9. Adler B, Biggs WS, Bazemore AW. State patterns in medical school expansion, 2000–2010: Variation, discord, and policy priorities. Acad Med. 2013;88:18491854.
10. American Association of Colleges of Osteopathic Medicine. Trends in osteopathic medical school applicants, enrollment and graduates. http://www.aacom.org/docs/default-source/data-and-trends/2014-trends-COM-AEG-PDF.pdf?sfvrsn=26
. Published 2014. Accessed March 18, 2015. [No longer available.]
11. Mullan F, Salsberg E, Weider K. Why a GME squeeze is unlikely. N Engl J Med. 2015;373:23972399.
12. Institute of Medicine. Graduate Medical Education That Meets the Nation’s Health Needs. 2014.Washington, DC: National Academies Press.
13. Salsberg E, Rockey PH, Rivers KL, Brotherton SE, Jackson GR. US residency training before and after the 1997 Balanced Budget Act. JAMA. 2008;300:11741180.
14. Weida NA, Phillips RL Jr, Bazemore AW. Does graduate medical education also follow green? Arch Intern Med. 2010;170:389390.
15. Holt KD, Miller RS, Philibert I, Nasca TJ. Patterns of change in ACGME-accredited residency programs and positions: Implication for the adequacy of GME positions and supply of physicians in the United States. J Grad Med Educ. 2014;6:399403.
16. Jolly P, Erikson C, Garrison G. U.S. graduate medical education and physician specialty choice. Acad Med. 2013;88:468474.
17. Council on Graduate Medical Education. Advancing Primary Care. 2010. Washington, DC: Health Resources and Services Administration; http://www.hrsa.gov/advisorycommittees/bhpradvisory/cogme/Reports/twentiethreport.pdf
. Accessed November 8, 2016.
18. Medicare Payment Advisory Commission. Report to the Congress: Aligning Incentives in Medicare. 2010.Washington, DC: Medicare Payment Advisory Commission.
19. Chen C, Petterson S, Phillips RL, Mullan F, Bazemore A, O’Donnell SD. Toward graduate medical education (GME) accountability: Measuring the outcomes of GME institutions. Acad Med. 2013;88:12671280.
20. Center for Workforce Studies. 2013 State Physician Workforce Data Book. 2013. Washington, DC: Association of American Medical Colleges; https://members.aamc.org/eweb/upload/State%20Physician%20Workforce%20Data%20Book%202013%20(PDF).pdf
. Accessed November 8, 2016.
21. Seifer SD, Vranizan K, Grumbach K. Graduate medical education and physician practice location. Implications for physician workforce policy. JAMA. 1995;274:685691.
22. Accreditation Council for Graduate Medicine Education. Internal medicine. http://www.acgme.org/Specialties/Overview/pfcatid/2/Internal%20Medicine
. Published 2016. Accessed June 3, 2016.
23. Accreditation Council for Graduate Medicine Education. Family medicine. http://www.acgme.org/Specialties/Overview/pfcatid/8
. Published 2016. Accessed June 3, 2016.
24. Accreditation Council for Graduate Medicine Education. Pediatrics. http://www.acgme.org/Specialties/Overview/pfcatid/16/Pediatrics
. Published 2016. Accessed June 3, 2016.
25. Association of American Medical Colleges. Table 27: Total graduates by U.S. medical school, sex, and year. 2009–2010. https://www.aamc.org/data/facts/enrollmentgraduate/148670/total-grads-by-school-gender.html
. Accessed March 18, 2014. [No longer available.]
26. American Association of Colleges of Osteopathic Medicine. Graduates by osteopathic medical college and race and ethnicity 2000–2014. http://www.aacom.org/reports-programs-initiatives/aacom-reports/graduates
. Accessed March 18, 2015. [No longer available.]
27. Health Resources and Services Administration. Affordable Care Act—Teaching Health Center Graduate Medical Education (THCGME) Program. http://bhw.hrsa.gov/fundingopportunities/default.aspx?id=d7c71e90-f4bd-4e3e-8108-b11fa774e38a
. Accessed November 9, 2016.
28. Chen RM, Petterson SM, Bazemore AW. Projected impact of the Primary Care Residency Expansion program using historical trends in graduate placement. Am Fam Physician. 2014;89:518.
29. Chen C, Chen F, Mullan F. Teaching health centers: A new paradigm in graduate medical education. Acad Med. 2012;87:17521756.
30. Spero JC, Fraher EP, Ricketts TC, Rockey PH. GME in the United States: A Review of State Initiatives. 2013.Chapel Hill, NC: Cecil G. Sheps Center for Health Services Research, The University of North Carolina at Chapel Hill.
31. Reid RJ, Fishman PA, Yu O, et al. Patient-centered medical home demonstration: A prospective, quasi-experimental, before and after evaluation. Am J Manag Care. 2009;15:e71e87.
32. Wu WN, Bliss G, Bliss EB, Green LA. Practice profile. A direct primary care medical home: The Qliance experience. Health Aff (Millwood). 2010;29:959962.
33. Iglehart JK. Expanding the role of advanced nurse practitioners—Risks and rewards. N Engl J Med. 2013;368:19351941.
34. National Academies of Sciences, Engineering, and Medicine. Assessing Progress on the Institute of Medicine Report The Future of Nursing. 2016.Washington, DC: National Academies Press.
35. American Osteopathic Association. 2013 AOA Match results. http://www.osteopathic.org/inside-aoa/Education/students/match-program/Pages/match-results.aspx
. Accessed March 20, 2014. [No longer available.]
36. Accreditation Council for Graduate Medical Education. Single GME accreditation system. http://www.acgme.org/What-We-Do/Accreditation/Single-GME-Accreditation-System
. Accessed November 2, 2016.
37. Whitcomb ME. Commentary: Meeting future medical care needs: A perfect storm on the horizon. Acad Med. 2011;86:14901491.
38. Henderson TM. Medicaid Graduate Medical Education Payments: A 50-State Survey. 2013. Washington, DC: Association of American Medical Colleges; https://members.aamc.org/eweb/upload/Medicaid%20Graduate%20Medical%20Education%20Payments%20A%2050-State%20Survey.pdf
. Accessed November 8, 2016.
39. Chang BK, Brannen JL. The Veterans Access, Choice, and Accountability Act of 2014: Examining graduate medical education enhancement in the Department of Veterans Affairs. Acad Med. 2015;90:11961198.
40. Phillips RL, Petterson S, Bazemore A. Do residents who train in safety net settings return for practice? Acad Med. 2013;88:19341940.
41. Fagan EB, Gibbons C, Finnegan SC, et al. Family medicine graduate proximity to their site of training: Policy options for improving the distribution of primary care access. Fam Med. 2015;47:124130.
42. Kuo YF, Sharma G, Freeman JL, Goodwin JS. Growth in the care of older patients by hospitalists in the United States. N Engl J Med. 2009;360:11021112.