Secondary Logo

Journal Logo

Research Reports

U.S. Graduate Medical Education and Physician Specialty Choice

Jolly, Paul PhD; Erikson, Clese MPAff; Garrison, Gwen PhD

Author Information
doi: 10.1097/ACM.0b013e318285199d
  • Free
  • AM Rounds Blog Post


Many believe that the United States already faces a critical physician shortage, and the problem will only worsen as more Americans gain access to health insurance under the Affordable Care Act and additional millions enter the Medicare system.1,2 Whereas previous projections showed a baseline shortage of 39,600 doctors in 2015, current estimates by the Center for Workforce Studies of the Association of American Medical Colleges (AAMC) bring that number closer to 63,000, with worsening shortages through 2025.3 The projected shortages are not only in the primary care fields of family medicine, internal medicine, and pediatrics but also in surgery, emergency medicine, cardiology, oncology, and other fields.


Licensure authorities require medical school graduates who wish to become licensed to practice medicine in the United States to participate in a residency, a program of graduate medical education (GME) that takes place in a teaching hospital and is accredited by the Accreditation Council for Graduate Medical Education (ACGME). Newly graduated physicians determine their choice of specialty by selecting a residency program in a core specialty, often followed by advanced training in a subspecialty residency or fellowship. Although both MD-granting and DO-granting medical schools have raised enrollments substantially,4,5 these changes cannot increase the number of practicing physicians without concomitant growth in the number of residency positions available for GME in teaching hospitals.

Medicare partially funds residency positions, allowing teaching hospitals to recover a portion of the costs of GME when residents serve patients who are eligible seniors. Since 1997, however, the Balanced Budget Act has limited this funding.6,7 Without an increase in government support from Medicare, Medicaid, and other government programs, teaching hospitals must find other sources of revenue to pay for the training of new physicians.

Although numbers of residency positions increased to some degree during the last decade and a half,8 analysts raise a serious concern that the pipeline for the production of the physician workforce is inadequate for future needs.2 It is imperative to continue to monitor the structure and size of this pipeline, which was the goal of the research presented here. We report the numbers of residents and trends by specialty and subspecialty, paying special attention to the primary care disciplines of family medicine and internal medicine/pediatrics, and to those internal medicine and pediatrics residents likely to remain in primary care. We also report the numbers of U.S. MD (US MD) graduates, DO graduates, and international medical graduates (IMGs) enrolled in each specialty.


For the purposes of this report, residency programs accredited by the ACGME are grouped into five categories:

  • Core specialties (n = 25), which can be entered directly from medical school or with one preliminary year, for example, anesthesiology;
  • Combined specialty programs (n = 20), which combine a core specialty with one or more other specialties or subspecialties, and which can be entered directly from medical school, for example, internal medicine and pediatrics;
  • Subspecialty residencies and fellowships (n = 116), which require completion of an earlier residency program, for example, cardiology;
  • Sub-subspecialty fellowships (n = 4), which require prior completion of a subspecialty residency or fellowship, for example, interventional cardiology; and
  • Preliminary programs (n = 4), which are one-year training programs in internal medicine, pediatrics, or general surgery, or a transitional year that will ordinarily be followed by entry into a residency in a core specialty.

We include graduates of osteopathic medical schools who participate in ACGME-accredited programs, but not participants in unaccredited programs or programs accredited by other bodies such as the American Osteopathic Association (osteopathic internships and residencies) and the American Board of Obstetrics and Gynecology (obstetrics–gynecology fellowships).

To support the analysis in this report, we created comprehensive tables of all residents and of new residents in program year 1 in all accredited specialties and subspecialties. For this report, we have summarized the data; all the details for the small as well as the large specialties and subspecialties are available in four supplementary tables (

Sources of GME data

In preparing this report, we used data derived from the National GME Census (GMETrack) system of the AAMC and the American Medical Association (AMA). We included data from the censuses of 1996 through 2010, but the analysis was focused on residents on duty on December 31 of 2001, 2004, 2007, and 2010.

Three other distinct sources of data on numbers of residents participating in GME are

  • Reports from the National Resident Matching Program (NRMP). Each year, the NRMP publishes reports on the numbers of applicants matched in residency programs.9 Although useful for identifying trends in specialty choice, especially by U.S. seniors, data from the NRMP exclude several thousand residents who, each year, find positions outside the main residency match.10
  • Data published by the ACGME. ACGME reports comprehensive statistics on GME annually.11 Their data are collected independently of the National GME Census sponsored by the AAMC and the AMA.
  • Data included in the annual medical education issue of the Journal of the American Medical Association (JAMA), typically published in September of each year. Although the JAMA medical education journal publication also uses National GME Census data,12 the AAMC processes the data differently from the way the AMA processes the data for JAMA. The AMA reports data from a snapshot taken from the survey data in the spring of each year, whereas the AAMC analysis uses data from all reporting years to determine the best indication of which residents are on duty in each year.

We also made an adjustment to recent-year counts to allow for late reporting of data. Because we can count new 2007 residents using 2007, 2008, or 2009 data, it is possible to determine the increase in counts for 2007 residents that are derived from data collected in 2008 and 2009. We adjusted the 2010 rates up by 3.58% for new entrants into residency and by 2.16% for total residents, based on historical trends in late reporting, therefore ensuring greater comparability of results across time. Otherwise, the most recent data would appear lower simply because of late reporting and would not be an accurate reflection of resident counts. Because of this adjustment and the difference in counting methodology, the AAMC is able to report several hundred more residents than does the AMA.

Tracking specialty choice

GMETrack does not directly distinguish residents who enter internal medicine, pediatrics, or general surgery as a required preliminary year before entering a core specialty such as anesthesiology or radiology, but it includes a start date and end date for each annual residency report. Where the difference in dates in an internal medicine, pediatrics, or general surgery residency report is less than or equal to 400 days, we assumed that it was a preliminary internal medicine, pediatrics, or surgery experience (we use 400 days rather than 365 days to allow for those occasions when a resident starts early or finishes late). To determine the number of program year 1 residents in each specialty, we counted the number entering their first residency that was not a preliminary or transitional year residency.

To examine the trends in choice of core specialty, we defined a cohort of all residents beginning a first year in a core specialty, with or without first completing a preliminary or transitional year residency. All residents who pursue certification in a specialty must pass through this logical portal.

Estimating primary care outcomes

In this work, we used a new method for estimating the number of physicians who will practice primary care. From the number entering training programs in family medicine, internal medicine, pediatrics, and combined internal medicine/pediatrics, we subtracted the number entering subspecialty internal medicine and pediatrics fellowships in that same year. We assumed that the number of the current year’s cohort that will enter fellowship training approximates the number from the cohort from three years ago that is entering fellowship training today. If the fraction that chooses to subspecialize is changing only slowly, the approximation will be a good one, but it may under- or overestimate the number subspecializing when the fraction is changing rapidly. For example, if the fraction is increasing, basing the estimate of the number of this year’s new residents who will enter subspecialties three years later on the number entering subspecialties today will underestimate the subspecialty number.

Salsberg et al8 calculated a subspecialization fraction by dividing the number completing the core specialty into the number entering related subspecialties in the following year. This is similar to our approach, but our method allows a determination of the subspecialization fraction for the current year. Both methods include physicians who delay fellowship training by one or more years, because all new fellows are counted whether they go into fellowships immediately after completing the core residency or after a hiatus. Analysis of longitudinal data shows that it is not uncommon for physicians to return for fellowship training one, two, or even more years after completing a core residency. (For example, of 650 residents who completed three or more years of internal medicine in 2002 and subsequently entered a cardiology fellowship, only 441 did so in 2003, whereas 143 began cardiology in 2004, 34 in 2005, and 32 even later.)

An earlier version of this analysis, using somewhat older data, was published by the Macy Foundation in the proceedings of a conference held in October 2010.13 Our work is based on more recent data and uses refined estimation procedures.

The institutional review board of the American Institutes for Research determined that the study protocol was exempt from IRB review.


The total number of residency positions is growing slowly

As shown in Table 1, between the years 2001 and 2010 the number of residents in ACGME-accredited programs that were reported to GMETrack increased by 13,655 (13.6%). In 2010, there were 4,754 residents in preliminary programs and 89,142 residents in core specialty and combined specialty programs, programs available to physicians without prior GME or with only a preliminary year of GME. There were 20,007 in subspecialty training programs, including 439 in sub-subspecialty programs that require prior completion of a subspecialty residency. The total number of residents and fellows in 2010 was 113,903.

Table 1:
U.S. Residents and Fellows in Program Year 1, 2001–2002 to 2010–2011, by Program Type*

Of the 13,655 additional residents reported in 2010 when compared with 2001, core specialties accounted for 8,322 (60.9%) of the additional positions, which was a growth of 10.6% in the total number of residents in core specialties. The total number of residents in subspecialty and sub-subspecialty programs increased by 5,651 (39.4%). The numbers of residents in preliminary programs and combined specialty programs changed little.

Table 2 shows the number of program year 1 residents in each specialty and related subspecialties for the same years. In 2010, there were 1,964 (8.4%) more residents in program year 1 positions in core specialties and combined specialties than in 2001. The increase in the number of program year 1 residents in subspecialty residencies was 2,834, a 44.5% increase. The number of subspecialty participants has increased much more rapidly than the number of new entrants in core specialties. Substantially increased subspecialization is apparent in almost every core specialty.

Table 2:
Residents and Fellows in Program Year 1, 2001–2002 to 2010–2011, by Specialty*

Figure 1 displays the number of program year 1 residents for the most recent year of our data for internal medicine and its subspecialties, graphically demonstrating our approach to estimating primary care outcomes. The full length of the bar represents 6,693 new residents in internal medicine training programs (excluding preliminary programs) in 2010. Shown as offsets are the numbers entering each of the subspecialties of internal medicine in 2010, a total of 3,845. The difference, 2,848, is the number of new residents in 2010 expected to practice general internal medicine. (Some physicians with subspecialty training nevertheless practice primary care.)

Figure 1:
Predicted career paths of 6,695 internal medicine residents who began program year 1 in 2010. The full figure represents 6,695 new categorical internal medicine residents reported to GMETrack in 2010, offset by the number of new fellows in each of the internal medicine fellowship programs in 2010. The difference is an estimate of the number of new residents who will not enter subspecialty training. Source: GMETrack Database, AAMC Data Warehouse, 2012.

The decline in interest in primary care may be ending

Table 3 and Figure 2 show the result of applying our primary care analysis for the four primary care specialties: family medicine, internal medicine, pediatrics, and internal medicine/pediatrics. Between 2001 and 2010, the number expected to enter primary care decreased from 8,624 to 8,084 (6.3%) as the number subspecializing in both internal medicine and pediatrics continued to increase. Of the new entrants in 2010 into categorical internal medicine, 3,845 (57%) are expected to enter subspecialties (compared with 51% in 2001); 1,145 (42%) of new entrants into pediatrics will subspecialize (compared with 28% in 2001). However, when we look at the changes year by year, the data for recent years suggest a possible change in the trend. The number and percentage of new program year 1 residents expected to enter primary care for the most recent year are essentially unchanged from the prior year, offering some hope that the decline in interest in primary care careers may be ending.

Table 3:
Projected Numbers and Percentages of U.S. Residents Who Will Remain in Primary Care*
Figure 2:
Projected primary care outcomes for 2001 to 2010. Primary care includes family medicine, general internal medicine, general pediatrics, and internal medicine/pediatrics. Source: GMETrack Database, AAMC Data Warehouse, 2012.

The number of DO graduates in ACGME-accredited residencies is increasing much more rapidly than the number of US MD graduates or IMGs

Table 4 shows the distribution of US MD, DO, and IMG residents in each specialty and subspecialty by type of undergraduate medical education program. The increase in all residents reported to GMETrack since 2001 is 6,534 (9.7%) for US MDs, but it is 4,127 (15.2%) for IMGs and 3,406 (69.0%) for DOs in ACGME-accredited residencies.

Table 4:
Distribution of U.S. Medical Graduate (US MD), Osteopathic (DO), and International Medical Graduate (IMG) Residents and Fellows by Program Type and Type of Undergraduate Medical Education Program, 2001–2002 to 2011–2012*

Figure 3 shows that DOs and IMGs are more likely to become primary care physicians, with nearly half expected to enter primary care compared with one-quarter of US MDs. The percentages of new US MD and DO program year 1 residents expected to enter primary care have increased in recent years, whereas for IMGs the percentage has continued a long decline.

Figure 3:
Estimated percentages of program year 1 residents (U.S. medical graduates, osteopathic graduates, and international medical graduates) who will enter primary care. Source: GMETrack Database, AAMC Data Warehouse, 2012.


The number of residency positions is increasing in spite of Medicare limitations

The Balanced Budget Act of 1997 established per-hospital caps on the number of residency positions eligible for Medicare GME support. Nevertheless, the number of training positions has increased slowly since 1997, in part because of additional sources of funding. The nation’s teaching hospitals self-fund, in aggregate, nearly 10,000 positions above the Medicare cap,14 and some states partially support GME positions under their Medicaid programs. However, the number of states making Medicaid GME payments has declined significantly because of budget constraints. There is no mandate that private payers provide any level of GME support.

New teaching hospitals have provided a second source of additional positions, although the number and size of these programs are typically small. Additionally, new teaching hospitals are not exempt from Medicare’s GME cap and are not able to increase their level of Medicare GME support over time.

The decline in interest in primary care continues to be a concern

The steady decline in interest in primary care careers over the past several years has been worrisome in view of the anticipated increase in demand for primary care services. Although the latest figures give some hope that the annual declines may have ended, we will have to see whether this change in trend is confirmed in future analyses. Over the past decades, interest in individual specialties and subspecialties has, at times, both increased and decreased, and this behavior will likely continue in the future. A trend toward increasing interest in primary care careers could be reinforced if the job market, reimbursement policies for physician services, or other environmental conditions should change.

Although primary care salaries are lower than for other specialties, there are signs that primary care salaries are on the rise, and this could make the field more attractive. At least 27 states are already paying premiums for patient-centered medical homes for Medicaid patients, and an additional 14 states are developing plans to do so.15 Private insurers such as Blue Cross Blue Shield16 and Wellpoint17 have also launched pilots or incentive programs in some markets. Furthermore, the Affordable Care Act is placing a strong spotlight on the need for more primary care physicians and the role they can play in meeting the triple aim of decreasing cost of care and improving quality and outcomes.18

There is increasing specialization in surgery and other specialties

Many new subspecialties have been created in recent years. In 1980 and 1981, certificates were awarded in only 51 specialties and subspecialties.19 Today, the American Board of Medical Specialties recognizes three times that many,20 and new certificates are approved almost every year.21 Whereas more than three-fourths of residency positions remain in the core specialties and combined specialties, the creation of these new subspecialties has contributed to the growth in GME, as physicians return for additional training to qualify them for practice in newly developing areas.

Increased participation in subspecialty training is not limited to internal medicine and pediatrics but also is occurring in family medicine and in numerous other specialty areas, including dermatology, neurology, orthopedic surgery, pathology, psychiatry, diagnostic radiology, and surgery. As Table 2 clearly shows, the trend toward continuing increases in subspecialization is quite general.

An area of special concern is general surgery. Since 2001, the number of new residents in general surgery, excluding those who only do a preliminary year, has declined by 16.3%, and the number expected to enter practice as general surgeons without further GME has declined by 33.3%.

Growth in GME is not keeping pace with growth in undergraduate medical education

In 2010, there were 16,838 MD graduates of U.S. medical schools.4 With 25,345 entry-level positions in core specialties and combined specialties, there were 1.5 positions available for each MD graduate. Of course, osteopathic graduates and IMGs are competing for the same positions.

Over the period 2001 to 2010, new residents in core specialties and combined specialties increased at a compound annual rate of 0.90%. Between 2001 and 2010, first-year US MD enrollment increased at 1.38% annually,4 a rate substantially more rapid. First-year DO students increased very much faster, at 6.64% annually.5

If GME training positions continue to grow only slowly while the number of U.S-trained MD and DO graduates continues rapidly increasing, the number of U.S-trained MD and DO applicants for accredited residency programs will eventually be equal to or will exceed the number of available entry-level positions. Should these trends continue, the approximately 7,000 IMGs who currently enter GME each year will find it increasingly difficult or impossible to find residency positions. If nothing is done, eventually there will even be too few entry-level residency positions to accommodate U.S. graduates.

The principal limitation of this work is the estimated 5% of residency experiences not reported to the GMETrack system. Because of this limitation, the aggregate numbers of residents reported may slightly understate the actual numbers of physicians in residency training. There is no reason to expect that there are systematic biases, however, and we are confident that the trends identified here are valid.

Funding/Support: The Josiah Macy Jr. Foundation supported an earlier version of this work.

Other disclosures: None.

Ethical approval: The institutional review board of the American Institutes for Research determined that the study protocol was exempt from IRB review.

Previous presentations: An earlier version of this work was reported in the proceedings of a conference entitled “Ensuring an Effective Physician Workforce for America,” sponsored by the Josiah Macy Jr. Foundation and held in Atlanta, Georgia, October 24 and 25, 2010.12


1. Association of American Medical Colleges.. The Complexities of Physician Supply and Demand: Projections Through 2025. Accessed December 11, 2012.
2. Association of American Medical Colleges.. The Impact of Health Care Reform on the Future Supply and Demand for Physicians Updated Projections Through 2025. Accessed December 11, 2012.
3. Association of American Medical Colleges.. Physician shortages to worsen without increases in residency training. Accessed December 11, 2012.
4. Association of American Medical Colleges.AAMC Data Book. 2011 Washington, DC Association of American Medical Colleges
5. American Association of Colleges of Osteopathic Medicine. Student enrollment. Accessed December 11, 2012.
6. Association of American Medical Colleges.. Medicare Direct Graduate Medical Education (DGME) Payments. Accessed December 11, 2012.
7. Association of American Medical Colleges.. Medicare Indirect Graduate Medical Education (IME) Payments. Accessed December 11, 2012.
8. Salsberg E, Rockey PH, Rivers KL, Brotherton SE, Jackson GR. US residency training before and after the 1997 Balanced Budget Act. JAMA. 2008;300:1174–1180
9. National Resident Matching Program, Results and Data: 2011 Main Residency Match. 2011 Washington, DC: National Resident Matching Program
10. Jolly P. First-year residents who began their graduate medical education in 2009–2010 and found their positions within and outside the NRMP match. Acad Med. 2012;87:586–591
11. Accreditation Council for Graduate Medical Education.ACGME Data Resource Book Academic Year 2010–2011. Accessed December 11, 2012.
12. Brotherton SE, Etzel SI. Graduate medical education, 2010–2011. JAMA. 2011;306:1015–1030
13. Josiah Macy Jr. Foundation.Ensuring an Effective Physician Workforce for America. 2010 New York, NY: Josiah Macy Jr. Foundation
14. . U.S. Department of Health and Human Services. Hospital Medicare Cost Report Data—FY2009. Accessed December 18, 2012.
15. National Academy for State Health Policy. . Medical home and patient centered care map. Accessed December 18, 2012.
16. Share DA, Mason MH. Michigan’s Physician Group Incentive Program offers a regional model for incremental ‘fee for value’ payment reform. Health Aff (Millwood). 2012;31:1993–2001
17. Raskas RS, Latts LM, Hummel JR, Wenners D, Levine H, Nussbaum SR. Early results show WellPoint’s patient-centered medical home pilots have met some goals for costs, utilization, and quality. Health Aff (Millwood). 2012;31:2002–2009
18. Iglehart JK. Primary care update—Light at the end of the tunnel? N Engl J Med. 2012;366:2144–2146
19. American Board of Medical Specialties.Annual Report and Reference Handbook—1987. 1987 Evanston, Ill: American Board of Medical Specialties
20. American Board of Medical Specialties.2011 Certificate Statistics. 2012 Evanston, Ill: American Board of Medical Specialties
21. American Board of Medical Specialties.. ABMS Announces Certification in Three New Physician Subspecialties: Pediatric Anesthesiology, Female Pelvic Medicine and Reconstructive Surgery and Complex General Surgical Oncology. Accessed February 24, 2013.
© 2013 Association of American Medical Colleges