US Graduate Medical Education, 2004-2005: Title and subTitle BreakTrends in Primary Care Specialties

The American Medical Association (AMA) and the Association of American Medical Colleges (AAMC) collaboratively maintain a comprehensive database on physicians training in residency and fellowship programs that are accredited by the Accreditation Council for Graduate Medical Education (ACGME). Our purpose in this report is to describe major trends over the past 9 years among residents in family medicine, internal medicine, pediatrics, combined internal medicine/pediatrics programs, and obstetrics/gynecology. Residents in these programs are most likely to become primary care physicians. However, as many residents pursue subspecialty training in fellowships, we also describe the trends in fellows training in ACGME-accredited subspecialties of internal medicine and pediatrics.

Our analysis bridges 1995-1996, a period characterized by a widely held belief that the United States was training more-than-adequate numbers of physicians but needed more primary care specialists,¹ to 2004-2005, when a new consensus began to emerge that we may face shortages of physicians in most specialties and subspecialties.^{2 - 3} Between these periods, there were fundamental shifts in the substrate of medical school graduates entering GME in the United States. The number of graduates from US allopathic medical schools (USMDs) remained stable, but an increasing percentage were women; the number of graduates from osteopathic medical schools (DOs) expanded greatly, exceeding the capacity of osteopathic GME positions; and, although graduates from foreign medical schools (IMGs) entered US GME at a fairly constant rate, they were increasingly likely to be US citizens (USIMGs).^{4 - 6} A detailed analysis of how these medical school graduates have been distributing themselves into primary care residency programs over the past decade seems especially timely in light of current calls for a 15% increase in graduates from US medical schools over the coming decade.^{2 - 3}

The AMA and the AAMC jointly sponsor and administer the National GME Census through GME Track, an Internet-based AAMC product.⁷ The 2 organizations also jointly maintain a database of ACGME-accredited training programs and of the residents and fellows in them. We update this database annually by adding to it the approximately 19 000 new residents who match into programs through the National Resident Matching Program (NRMP) as well as from information gathered through the AAMC’s follow-up report of medical schools. From May to December 2004 we surveyed directors of programs accredited by the ACGME about educational characteristics of their training programs. We included combined programs, which prepare residents for practice and board certification in 2 or more core specialties that are ACGME-accredited. Information about ACGME-accredited programs is maintained on FREIDA Online, an Internet-based public information source on GME programs that is available to medical students and residents.⁸

In July 2004, we surveyed program directors about their active, transferred, and graduated residents for academic year 2004-2005. We provided program directors with lists of residents and fellows from our database and asked the directors to confirm or modify the training status of trainees who were present in their programs the prior year; to add new physicians to their program whom we did not already have in our database; and to confirm, edit, or add demographic information. Our database is cross-sectional and may differ somewhat from the cumulative database maintained by the ACGME, to whom programs must continually provide information as part of the accreditation process. Since the National GME Census tracks all active, transferred, and graduated residents and fellows in all ACGME-accredited programs, the data collected are more comprehensive than data tabulated by the NRMP and other matching programs.

In this article, we use data from the survey of academic year 2004-2005 as well as from surveys of 3 previous years (1995-1996, 1998-1999, and 2001-2002) to examine trends in the 4 major specialties training primary care physicians. Because our GME database is considered an administrative database, data collection and the type of analyses we describe are waived from requirements for review by an institutional review board.

We surveyed 8246 active programs, of which 6979 (84.6%) confirmed the status of all of their active physicians-in-training, accounting for 97.3% of active trainees. An additional 13 programs (0.2%) confirmed some but not all of their trainees (accounting for 17 active residents or fellows), 141 confirmed that they did not have any trainees (1.7%), 334 programs (4.1%) confirmed the status of nonactive trainees (graduates and transfers) but did not have any currently active trainees, and 779 programs (9.4%) did not confirm the status of any physician training in the program (including 572 programs that apparently did not have any active trainees). A total of 95.7% of all physicians in our database had their status confirmed (eg, active, graduated, or withdrawn). Physicians whose status was not confirmed were “advanced” into the next year of training (n = 2665 [2.6% of active residents]) or “graduated” based on expected graduation date (n = 1472 [4.2% of graduated residents or fellows]).

Nearly 87% of program directors (n = 7163) completed the accompanying program survey; 6499 (78.8%) of the programs completed both the program survey and confirmed the status of all active physicians-in-training. Of the remaining 664 program directors who completed the program survey, 97 (1.2%) confirmed the status of some of their active residents or fellows; 129 (1.6%) confirmed that they had no active trainees; 354 (4.3%) had no active trainees to confirm according to our database but did not specifically indicate so on the survey; and 84 (1.0%) did not confirm any active trainees whom we had in our database. A total of 957 program directors (11.6%) confirmed all or some of their active residents or fellows (or confirmed that they had none) but did not complete the program survey. Only 126 programs did neither survey (1.5%).

As of December 31, 2004, the National GME Census counted 101 291 residents and fellows enrolled in ACGME-accredited and combined specialty GME programs for the 2004-2005 academic year. Table 1 presents trend data from the past 6 years on the number of programs, number of total residents, number of residents in graduate year 1 (GY1) positions (entry-level positions that neither require nor preclude prior GME), the number of GY1 residents in GY1 positions who have not had any prior GME, and the total number of residents and fellows by medical school origin. This is the first time we counted more than 100 000 physicians in our census of ACGME-accredited programs. The total number of GY1 residents decreased slightly from last year but continued to hold steady at around 24 000. The number of GY1 residents without prior training (ie, those entering US GME for the first time) was 22 788, the largest that we have recorded. The number of GY1 residents with prior training decreased to 1001, the lowest since the National GME Census began reporting this number for data from 1993-1994.⁹ In 1999-2000, residents who had already received some training made up 12.5% of GY1 residents; currently they make up only 4.2%. Additional data on all physicians-in-training are provided in Article .

For the first time since 1996-1997, the number of DOs in ACGME-accredited programs did not increase. Between academic years 1996-1997 and 2003-2004, the number of DOs in allopathic GME had increased an average of 8.6% per year. The number of DO trainees decreased from 5838 in 2003-2004 to 5675 this year. Conversely, the number of USMDs, graduates of Canadian medical schools, and IMGs all increased slightly.

Table 1 also presents the number of specialty and subspecialty programs and the number of residents and fellows training in them for the past 6 years. We define specialty programs as those leading to initial board certification (plus transitional-year programs), and subspecialty programs are those leading to subspecialty board certification, as defined by the American Board of Medical Specialties. For the third consecutive year, the number of residents in specialty training increased, now at 86 975. The number of specialty training programs continued to decline; the current number of 4151 is 117 fewer than in 1999-2000. The number of subspecialty programs, on the other hand, grew by 72, a 1.8% increase from last year, and the number of fellows in these programs is 709 more than last year, a 5.2% increase. The 1-year increase in the number of fellows in subspecialty programs is more than the gain in the number of residents in specialty programs (n = 618, a 0.7% increase).

We charted the trends in primary care training by examining this year’s data as well as data from 3 previous years, 1995-1996, 1998-1999, and 2001-2002. We included residents from all years of training from the specialties of family medicine, internal medicine, obstetrics/gynecology, pediatrics, and combined internal medicine/pediatrics programs, and fellows from the subspecialties of internal medicine and pediatrics. The 5 specialty areas were classified as primary care in accordance with the health care professional shortage area primary medical care designation criteria of the Bureau of Health Professions of the Health Resources and Services Administration.¹⁰

Although each data segment is separated by 3 years, there is some overlap because residents may appear in consecutive segments, especially in obstetrics/gynecology, where the training program is 4 years long, and in other specialties where residents may have extended their training (eg, as chief residents). We did not include the family medicine subspecialties of geriatric medicine or sports medicine, as the number of fellows and programs is consistently small; in 1995-1996 there were 14 programs and 22 fellows¹¹ and in 2004-2005 there were 99 programs and 118 fellows (see Article ). Subspecialty programs in obstetrics/gynecology are not accredited by the ACGME and are therefore not part of the National GME Census. We categorized residents by the type of medical school (US or Canadian allopathic, osteopathic, and foreign) and, in the case of IMGs, whether they were native or naturalized citizens of the United States at the time of GME.

The total number of programs in family medicine and internal medicine/pediatrics initially expanded and then contracted during this time frame, while the number of programs in internal medicine, obstetrics/gynecology, and pediatrics all contracted (Table 2). Subspecialty programs in internal medicine contracted and then expanded (Table 3). The number of pediatrics subspecialty programs continuously grew from 472 to 649 during this period.

Examining the total numbers of residents training in these areas shows the rise and fall in primary care interest since the mid 1990s, with a decrease followed by an increase in the number of fellows training in the subspecialties. Despite these trends, there are still more residents training in pediatrics, internal medicine, their combined programs, and in family medicine at the end of this time span compared with the beginning, a trend not evident for obstetrics/gynecology. After a drop in the number of fellows training in internal medicine and pediatric subspecialties in 1998-1999, there has been a steady rise.

We gauged the popularity of primary care specialties by noting the number of residents in these areas who are USMDs. More competitive specialties and subspecialties (as evidenced by their overall match rates and rankings in the NRMP) tend to attract more USMDs.^{12 - 13}

The composition of residents pursuing family medicine training has dramatically changed in recent years (Figure). Riding a crest in primary care popularity through the 1990s, 77.6% of family medicine residents were USMDs in 1998-1999. That percentage has fallen to 51.7%, a relative drop of one third and an absolute drop of 41.1% (from 8232 to 4848). At the same time, the number and percentage of family medicine residents who are IMGs (both US citizens and noncitizens) more than doubled.

The trends for the percentage of USMDs in pediatrics, internal medicine, and combined internal medicine/pediatrics programs followed parallel paths, although the percentage of USMDs was much higher for the combined programs, followed by pediatrics, then internal medicine. In 1995-1996 and 1998-1999, obstetrics/gynecology attracted proportionately more USMDs than other primary care areas; however, there are now 1039 fewer USMDs training in obstetrics/gynecology than there were 9 years ago. This loss allowed gains for the other types of trainees, most notably in the number of USIMGs, which grew from 72 residents in 1995-1996 to 256 in 2004-2005 (a 255% increase). The percentage of residents who are USMDs is currently greatest for internal medicine/pediatrics programs, numerically growing from 853 to 1180, a 38.3% increase.

The growth in the total number of fellows in the pediatric subspecialties over this period, an increase of 55.7%, was almost entirely due to the near doubling of USMDs in 9 years, from 813 to 1617. The percentage of USMDs increased from a slight majority in 1995-1996 (52.9%) to more than two thirds in 2004-2005 (67.5%). The proportion of USMDs in internal medicine subspecialties has followed a course similar to that of the pediatric subspecialties but with a lesser recent increase.

The total number of non-US IMGs training in the primary care specialties and subspecialties dipped during the middle of this period but is now the same as in 1995-1996 (14 723 in 1995-1996 and 14 724 in 2004-2005). However, the number of USIMGs who are residents in primary care specialties has almost doubled (from 1799 in 1995-1996 to 3358 in 2004-2005). The number of USIMGs who are fellows in subspecialties of internal medicine and pediatrics has also increased (from 622 in 1995-1996 to 906 in 2004-2005, a 45.7% increase). In 2004-2005, USIMGs represented 1 in 13 residents in primary care specialties and 1 in 12 fellows in the subspecialties of internal medicine and pediatrics.

Overall, the number of DOs in these areas increased 66.6%, from 2119 to 3530. This growth was not uniform, however; the number of DOs increased by 2 or more times in obstetrics/gynecology, pediatrics, and pediatric subspecialties but by less than 50% in both family medicine and internal medicine/pediatrics.

Comparing the ratio of subspecialty fellows to residents from each medical school group provides an indication of the relative likelihood of graduates from any group pursuing subspecialty training. In pediatrics, the current ratio of fellows to residents is highest for non-US IMGs (0.35), followed by USIMGs (0.34), USMDs (0.30), and DOs (0.17). In internal medicine, the current ratios are 0.52 for USIMGs, 0.46 for USMDs, 0.34 for DOs, and 0.31 for non-US IMGs.

In 1995-1996, 34% of all residents (including non–primary care specialties) were women, increasing to 41.8% in 2004-2005 (Table 4). The proportion of primary care residents who were women grew from 42.9% to 52.5%. While for each period the percentage of women in internal medicine programs mirrored the percentage for all GME specialties combined, other primary care specialties attracted higher complements of women. In family medicine, women increased from 42.3% to 51.9% of the total; in pediatrics, from 61.4% to 69.1%, in internal medicine/pediatrics, from 43.8% to 52.8%; and in obstetrics/gynecology, from 57.9% to 74.5%. The growth rate in the percentage of women training in obstetrics/gynecology was 28.7%. There are 600 more women training in obstetrics/gynecology this year compared with 1995-1996, even though the specialty overall has 300 fewer residents. During this same period, the percentage of women in subspecialty training increased from 23% to 31.9% for internal medicine and from 42.3% to 53.7% for pediatrics.

The number of residents and fellows in ACGME-accredited training programs is at a historic high. Most of the overall growth in trainees in the last 5 years has been fueled by an increase in fellows training in ACGME-accredited subspecialties. The rise in the number of fellows in ACGME-accredited subspecialty programs in internal medicine and pediatrics must be considered within the context of the growth in the number of accredited subspecialties. During the time span of our study, 2 new internal medicine subspecialties were added to the ACGME accreditation process: interventional cardiology (121 programs, with 206 fellows currently) and sports medicine (2 programs, with 1 fellow). Although physicians would have been training in these areas prior to formal ACGME accreditation of their programs, they would not have been counted by the National GME Census. Nonetheless, compared with 1995-1996, there are 720 more fellows who are training in internal medicine subspecialties that are not new.¹¹

Pediatrics saw a continuous expansion in subspecialty training; the number of accredited programs grew from 472 to 649 during this time frame. The expansion in the number of programs is explained by the addition of 6 new ACGME-accredited subspecialties: adolescent medicine (currently with 23 programs and 71 fellows), developmental-behavioral pediatrics (22 programs and 36 fellows), emergency medicine (44 programs and 202 fellows), infectious diseases (62 programs and 147 fellows), rheumatology (25 programs and 130 fellows), and sports medicine (8 programs and 10 fellows).

The number of residents in specialty programs has increased over the past 3 years, despite a reduction in the number of programs in which to train.The recent changeable hospital environment, with mergers and consolidations occurring among institutions, has likely led to program mergers (and closures) as well, thus decreasing the number of specialty programs.

The number of residents in GY1 positions who have had prior GME continues to drop. Several factors may contribute to more residents entering GME without prior training. Current residents may be more certain of their initial specialty choices. Medicare GME financing is higher for residents training for their first board certification than for those undertaking additional training; therefore, institutions sponsoring GME may be reluctant to hire residents seeking second residencies. Programs may be accepting IMGs without US GME experience with greater confidence than in the past, because since 1998 IMGs have had to pass a clinical skills examination that also assesses English-language proficiency and interpersonal skills.

Representation of USMDs in primary care specialty training programs appears to have peaked. Positions that in the past were filled by USMDs are now being filled by DOs and IMGs, particularly in family medicine programs. At the same time, subspecialty programs in internal medicine and pediatrics are increasingly competitive and attracting more USMDs. While USIMGs represent the most rapidly growing cohort of trainees in the primary care specialties, they also appear even more likely than USMDs to pursue subspecialty training in internal medicine and pediatrics. The reasons that US citizens attend foreign medical schools are varied; however, once core training in pediatrics or internal medicine is attained, USIMGs appear to be as interested in fellowship positions and successful in finding such positions as are USMDs.

This year, the number of DOs in allopathic training programs declined. There were 32 fewer DO GY1s without prior training this year compared with last year, the first time in 5 years that the number of DOs has not increased. This decline is not due to a decrease in DO graduates. There were 2713 osteopathic medical school graduates in 2004, a 4.1% increase from the 2607 graduates in 2003.¹⁴ To accommodate increases in DO graduates (projected to be 3342 in 2009¹⁴ ) there has been a steady increase in the number of osteopathic internship positions available, accompanied by greater geographic dispersion. Furthermore, the number of osteopathic residency programs offering a specialty track continues to grow: 549 in 2003-2004 vs 486 in 2000-2001.¹⁵ These programs offer specialty training similar to ACGME-accredited non–primary care specialty and subspecialty programs. Therefore, DOs seeking such specialty practices may increasingly obtain their training in osteopathic programs.

Women constitute a majority of trainees in the 4 primary care specialties (52.2%) as well as in pediatric subspecialty programs (53.7%). Only recently did women become the majority of family medicine residents (51.9%). Although both obstetrics/gynecology and family medicine have seen a decline in the total number of residents, the recent decline in family medicine residents is steeper and due entirely to fewer men. There were 426 fewer residents in family medicine in 2004-2005 than in 2001-2002, but there were 43 more women, representing a loss of 469 men. The most striking growth in women residents has been in obstetrics/gynecology. If this trend persists, most practicing obstetricians/gynecologists in the United States will be women by 2020.¹⁶

As it is likely that women now entering primary care specialties will have different personal and professional expectations than the retiring physicians whom they will replace, practice characteristics within these specialties may change over time. For example, in obstetrics/gynecology, more women than men may seek practice arrangements that allow shorter work hours and periodic leaves of absence to provide time for child rearing and family responsibilities.¹⁷ Female obstetricians/gynecologists are less likely than men to practice obstetrics.¹⁸ Such changes in work effort have been documented in pediatrics, which also has a large percentage of women.¹⁹ It is not unreasonable to expect that other specialties in which the number of women is growing will experience similar changes in practice characteristics.

With some experts calling for a 15% increase in graduates from US medical schools over the coming decade,^{2 - 3} it is essential to understand the current sources of physicians in allopathic residency training. It is not clear whether expanding enrollment in US allopathic medical schools over the next decade would supplement or supplant current sources, since these schools might compete for the same pool of students who currently enter osteopathic medical schools or who attend medical school abroad and return to the United States for residency training.

As with all survey studies, limitations need to be considered in interpreting the results. This year, program directors and coordinators of more than 8200 programs were contacted about their programs and residents. While our surveys are online, and all attempts are made to make them as easy to complete as possible, the surveys are voluntary. Not all programs responded, and of those who did respond, not all answered every question. Nonetheless, our overall response rate of 84.6% is very high, and we believe that our survey data, which we supplemented with data from the NRMP, AAMC, and ACGME, present an accurate picture of US allopathic GME.

There are now more than 100 000 physicians training in ACGME-accredited programs. An increasing proportion of these physicians are pursuing subspecialty training, while the number in primary care specialties has leveled off after a period of popularity in the mid 1990s. The trends we describe suggest that the primary care medical workforce of the future will include more women, more IMGs, and more DOs, information which may inform the current discussions about physician workforce needs.