The National Study of Internal Medicine Manpower: XX. The Changing Demographics of Internal Medicine Residency Training Programs

Methods

The NaSIMM has been collecting data on residency programs in internal medicine annually since 1976-1977. The project is conducted by the Center for Health Administration Studies at the University of Chicago and is supported by the constituent organizations of the Federated Council for Internal Medicine. In July 1990, July 1991, and July 1992, the NaSIMM mailed questionnaires to the program directors of all internal medicine residency programs accredited by the Accreditation Council for Graduate Medical Education. Programs that did not respond were sent two follow-up mailings in the following fall. Those who still failed to respond were contacted by telephone and urged to complete the questionnaire. These efforts resulted in the following percentages of responses: 92% of 427 programs in 1990-1991, 93% of 421 programs in 1991-1992, and 90% of 419 programs in 1992-1993.

The NaSIMM questionnaires (available on request) cover a broad range of topics including but not limited to the number and characteristics of residents, the relative number of qualified applicants, and the current activities of the previous year's first- and third-year residents. Not all items are updated annually. In 1990-1991 and 1991-1992, we asked the programs to report resident characteristics at the level of the individual resident, whereas in 1992-1993 we asked only for aggregate numbers of residents in various categories by program year as we had done before 1990-1991. This change was made because some program directors could not produce individual listings.

For programs that failed to respond, we estimated the number of residents and their characteristics from previously collected NaSIMM data and data from other sources (for example, the Census of Graduate Medical Trainees conducted by the American Medical Association and the NRMP). No estimates were made for programs that did not respond for more than 2 consecutive years: two programs in 1990-1991, one in 1991-1992, and none in 1992-1993. For partial respondents, we estimated some or all of the residents' characteristics for 62 programs in 1990-1991, 45 programs in 1991-1992, and 49 programs in 1992-1993.

Programs and tracks are defined differently by the NRMP and the NaSIMM, which makes comparisons difficult. The NaSIMM defines a program as all residents training under a single director. In the NRMP, program tracks with the same institutional affiliation may be listed separately. The NaSIMM, for example, identified 84 primary care tracks in 1992-1993, whereas the NRMP listed 75. This happened in part because some programs recruit residents into both their primary care and categorical tracks under one NRMP identifier. To maximize the comparability of the two data sets, we include only categorical tracks (traditional 3-year programs), preliminary tracks (1 year that leads to another specialty), and primary care tracks (3-year programs with a special emphasis on primary care), thereby excluding 198 R1s in medicine-pediatrics (4 years of combined training) and other tracks. We also exclude 251 R1s in 24 programs (most of which are military) that did not participate in the 1992 NRMP.

Results

Demographic Characteristics

Table 1 shows the distribution of R1s and total residents across program tracks and their demographic characteristics for the academic years 1990-1991, 1991-1992, and 1992-1993. The percentage of R1s in the various tracks changed little in this period. In 1992-1993, 68% of R1s were in categorical tracks, 25% were in preliminary tracks, 5% were in primary care tracks, and 1% were in medicine-pediatrics tracks. In 1992-1993, the number of categorical residents increased 6%, the number of primary care residents increased 10%, and the number of preliminary residents increased 10% compared with the respective numbers from 1990-1991. The medicine-pediatrics track contains the fewest residents of any of the tracks and decreased by 37 residents over the 3-year period.

The sex and race characteristics of internal medicine R1s are also changing. For example, the proportion of female internal medicine R1s increased from 30% in 1990-1991 to 32% in 1992-1993. In addition, whereas the percentage of black and Hispanic R1s (5% and 6%, respectively) has not changed, the percentage of Asian R1s increased from 21% to 28%, an increase of 701 persons.

The most pronounced developments reflected in Table 2 are the number and proportion of internal medicine residents who are FNIMGs. The number of R1 FNIMGs increased by 868 over the 3-year period, a percentage increase of 8% (from 24% to 32%). By comparison, in 1976-1977 the NaSIMM recorded 10% FNIMGs, a percentage that remained relatively constant until 1987-1988 (Figure 1) [7]. The number of R1s who were U.S. citizen international medical graduates (USIMGs) decreased from 6% to 4% of all R1s. However, the number of R1s who are IMGs (both FNIMGs and USIMGs) is now 36% of R1s entering internal medicine residency training programs. In contrast, USMGs have decreased from a high of 86% in 1976-1977 to 63% in 1992-1993; most of the decrease has occurred since 1987-1988.

The percentage of USMGs in categorical tracks is less than that in any other track. In 1991-1992, USMGs composed 64% of residents in categorical tracks, 87% of residents in preliminary tracks, 72% of residents in primary care tracks, and 86% of residents in medicine-pediatrics tracks. Both FNIMGs and USIMGs represented larger percentages of the residents in categorical tracks than in any of the other tracks.

To investigate the relation between types of medical school graduate and program characteristics, we divided the population of residency programs into four groups of equal size based on the percentage of R1s who were USMGs in 1992-1993 (Table 2). The proportion of R1s who were USMGs ranged between 0% and 18% in programs in group A and between 19% and 70%, 71% and 95%, and 96% and 100% in programs in groups B, C, and D, respectively. Approximately 55% of internal medicine international medical school graduate (IMG) residents train in group A programs. This identification of relatively large numbers of programs almost entirely comprising USMGs or IMGs agrees with our earlier analyses of data from the 1988-1989 NaSIMM [9].

Group A programs tend to be located in the Northeast and Midwest. In fact, three states account for more than half of these programs (New York has 29, Illinois has 12, and New Jersey has 10). Pennsylvania, Michigan, Maryland, and Washington, D.C., also contain a disproportionate number of these programs. Group A programs are less likely to be primary affiliates of a medical school (9% for group A compared with 16% for group B, 37% for group C, and 49% for group D) and tend to be in smaller hospitals (33% have 500 beds or more compared with 42% for group B, 51% for group C, and 57% for group D). These programs are also more likely to be located in community teaching hospitals (78% for group A, 62% for group B, 48% for group C, and 37% for group D) and municipal hospitals (12% for group A compared with 9% for group B, 7% for group C, and 3% for group D), and are less likely to be located in university hospitals (7% for group A, 23% for group B, 41% for group C, and 51% for group D). However, group A programs are similar to group D programs in that they are more likely to be located in core counties of large metropolitan areas (63% for group A, 49% for group B, 44% for group C, and 66% for group D). Programs in groups C and D tend to be larger in terms of numbers of residents, and these residents are more likely to be women (Table 2).

Perceived Quality of Applicants

To understand how program directors react to the changing nature of internal medicine residencies, we asked them to subjectively assess the number of qualified applicants they receive. The most striking finding is that applicants from 1992-1993 were evaluated more positively than in earlier years (Table 3). Although it is true that programs that attract more R1s who are USMGs generally rate their applicants more positively Table 2, the shift to a more positive evaluation of the applicant pool was not limited to this group. Whereas in 1990-1991, 12% of programs in group A reported more qualified applicants, in 1992-1993 26% did so. Similarly, the percentage of programs in group D reporting more qualified applicants increased from 24% to 36% between 1990-1991 and 1992-1993, and for group C the percentage increased from 13% to 31%. For group B the increase was smaller, from 16% to 20%. Of equal importance, the percentage of programs reporting fewer or significantly fewer qualified applicants decreased across all groups between 1990-1991 and 1992-1993: group A, from 51% to 29%; group B, from 49% to 42%; group C, from 47% to 36%; and group D, from 33% to 27%. Despite the positive change in the most recent data, it is important to recognize that more programs report fewer or significantly fewer qualified applicants than programs reporting more qualified applicants. Only in group D did this relation reverse in 1992-1993.

Placement of First-Year Residents

Most R1s continue as second-year residents (R2s), and approximately two thirds of R1s continue in the same program (Table 4). In 1992-1993, 7% of R1s transferred to a different internal medicine residency in their second year of training compared with 5% in 1989-1990. About one third of R1s do not enter a second year of training in internal medicine. This includes the approximately 25% of R1s who are in preliminary programs and are not expected to continue. The largest group of persons leaving internal medicine residencies enter anesthesiology residencies, and their numbers increased from 394 in 1989-1990 to 526 in 1992-93, or from 5% to 8% of all R1s. The other specialties that draw from internal medicine R1s do so in considerably smaller and more consistent numbers.

The programs with the fewest USMG R1s (group A) are the least likely to train residents for other specialties (Table 2). In these programs, only 14% of R1s were in preliminary tracks, whereas between 27% and 30% of R1s in the other three groups are in preliminary tracks. Residents in the group A programs are, however, more likely to leave for training in another internal medicine program (group A, 16%; group B, 8%; group C, 5%; group D, 3%). About two thirds of R1s who switch programs for their second year of residency originate from programs in groups A and B.

Placement of Third-Year Residents

Most R3s go directly into subspecialty training (40%) (Table 5). An earlier study by the NaSIMM showed the ultimate subspecialization rate of internal medicine residents to be about 60% [5]; therefore, approximately one third of R3s who enter subspecialty training pursue some other activity between their third year of residency and their first year of fellowship. The other major activities attracting R3s are solo, group, or health maintenance practice (23%); staff positions in hospitals, ambulatory care centers, and emergency rooms (12%); and chief residencies (10%).

The rates of direct entry into subspecialty training vary from 35% for residents in group B programs to 43% for residents in group D programs (Table 2). Earlier studies have shown that FNIMGs subspecialize at a higher rate than do either USIMGs or USMGs [5]. We would thus expect to observe the highest rates of entry into subspecialty training among group A programs and the lowest rates among group D programs. However, reported rates of direct entry into subspecialty training do not follow this pattern. Consequently, we must conclude either that residents in group A and B programs are more likely to pursue some temporary activity before entering a fellowship or that FNIMGs in group C and D programs subspecialize at an exceptionally high rate. We found some support for the first possibility in the higher rates of entry into hospital, ambulatory center, and emergency room staff positions for residents in group A programs (14% for group A compared with 9% for groups B, C, and D). However, this finding is offset by a reverse relation for rates of entry into chief residencies, which is the other most likely temporary activity (7% for group A, 10% for group B, 11% for group C, and 12% for group D).

The National Study of Internal Medicine Manpower and the National Resident Matching Program

In 1991-1992, most program directors reported increased efforts in recruitment (77%) (Table 2). The programs with the largest and smallest proportions of R1s who were USMGs had the smallest increases in recruitment efforts. To understand this finding better, we examined data from the NRMP, which is the major recruitment channel for residents.

The NaSIMM and the NRMP measure closely related but distinct populations. The NaSIMM records the number of residents actually on duty in the various programs, whereas the NRMP records only those positions offered and filled through the NRMP. We found that the NaSIMM consistently measures more R1s on duty than positions offered through the NRMP [10] (Figure 2). The number of R1s has increased steadily from about 6800 in 1978-1979 to 8927 in 1992-1993, an increase of about 30%. The number of R1 positions offered through the NRMP increased from about 5800 in 1978-1979 to 7467 in 1989-1990 and has subsequently decreased to 7403 in 1992-1993. The gap between positions offered through the NRMP and the actual number of R1s varied between 483 and 885 from 1978-1979 to 1990-1991. In 1991-1992 this figure increased to 1243 and in 1992-1993 increased to 1533. The percentage of R1 positions offered through the NRMP has decreased steadily from a high of 94% in 1987-1988 to 83% in 1992-1993, and the fill rate has decreased from a high of 90% in 1985-1986 to a low of 78% in 1991-1992. Consequently, the percentage of R1s obtaining their positions through the NRMP decreased from 82% in 1985-1986 to 68% in 1992-1993; the number of R1s entering internal medicine residency training outside the NRMP increased from 1337 in 1985-1986 to 2865 in 1992-1993.

In 1992-1993, programs with the most R1s who were USMGs ultimately filled only 77 more R1 positions than were offered in the NRMP, whereas programs with the fewest R1s who were USMGs filled 731 more R1 positions than were offered in the NRMP (Table 2). This represents 3% of R1 positions in group D programs and 40% of positions in group A programs. These rates are inversely proportional to the NRMP fill rates. Programs in group A filled 62% of positions offered through the NRMP, whereas programs in groups B, C, and D filled 74%, 88%, and 93% respectively.

Discussion

The results of our survey have great relevance for the emphasis placed by the current health care reform movement on primary care physicians. These data highlight trends that could both support and impede a shift toward an increased number of general internists, show that many internists do not enter subspecialty training directly from residency training, and identify a set of programs that are likely to be particularly vulnerable to a reduction in the number of FNIMGs in graduate medical education.

First, the number of women training in internal medicine has increased steadily since the NaSIMM first collected data in 1976-1977. Although these increases have not been as dramatic as those for FNIMGs, they represent a long-term trend that may ultimately have a greater effect. Earlier NaSIMM studies showed that women subspecialize at a lower rate than men, and other studies have corroborated this finding by emphasizing that women prefer primary care specialties [5, 11]. However, this sex-related preference appears to be diminishing [12], and female physicians regret their decision to enter medicine more frequently than men [13]. Thus, whereas the number of women entering internal medicine is increasing, this trend will only support an increased number of general internists if women continue to subspecialize at a lower rate than men.

Second, our data show an extraordinary demographic shift in internal medicine residency training programs that began in 1987-1988. The number of USMGs entering internal medicine residencies through the NRMP decreased from 4994 in 1986-1987 to 4428 in 1993-1994, while at the same time, the number of R1 positions has increased to an all-time high of 8936 in 1992-1993. The number of FNIMG trainees has thus dramatically increased from 830 residents (11%) in 1986-1987 to 2870 residents (32%) in 1992-1993. Currently, FNIMGs and USIMGs compose 36% of first-year internal medicine residents. These trends are even stronger when preliminary programs are excluded.

Because FNIMGs subspecialize to a greater degree than do USMGs, the increase in the number of FNIMGs poses a significant challenge to the proposed goal of having 50% of trainees enter primary care [6, 7, 14]. Both the Council on Graduate Medical Education and the Physician Payment Review Commission have recommended that residency positions be capped at 110% of students graduating from U.S. allopathic and osteopathic medical schools [15, 16]. This recommendation was formalized in the Primary Care Workforce Act (S1315) sponsored by Senator Jay Rockefeller. If this restriction is implemented, the demographic characteristics of persons entering internal medicine residency training programs will be profoundly affected because USMGs, USIMGs, and FNIMGs will compete for fewer positions.

The training of FNIMGs does not affect only the United States. Harrington and colleagues [17] found that the reduction in the number of FNIMGs who were trained in the United States between 1973 and 1983 adversely affected the six Latin American cities they studied. Should increased competition for residency positions drastically reduce the number of FNIMGs in U.S. graduate medical education, we will have to decide if we have a responsibility to improve the practice of medicine in foreign countries through the training of FNIMGs and how that responsibility relates to the 110% cap.

Third, the disparity between the number of R3s choosing to immediately enter subspecialty training and the number who ultimately subspecialize is particularly relevant to our urgent need to decrease the subspecialization rate and produce more primary care physicians. That about one third of those entering subspecialty training do not do so directly from residency training suggests that many causes of this phenomenon may exist. For example, R3s may deliberately choose to take a hiatus in training before they undertake a subspecialty fellowship, or they may apply unsuccessfully for a fellowship program after completing their residency training. In addition, we cannot ignore the possibility that some practicing general internists are returning to graduate medical education for subspecialty training. Whatever the case, it is likely that many of these internists may be more easily influenced to pursue or continue pursuing a generalist career than their colleagues who enter subspecialty training directly after residency training. Additional research is required before this opportunity can be adequately exploited.

Fourth, we are faced with significant issues regarding the programs and hospitals that train FNIMGs. The primary care emphasis of reform proposals means that internal medicine is unlikely to see large cuts in R1 positions in categorical or primary care tracks; however, many preliminary programs may be reduced or discontinued, and the number of fellowship positions will almost certainly be decreased. Where these reductions will be made will depend, in part, on the perceived quality of the residency training programs.

Our analyses indicate that more than half of the FNIMGs training in internal medicine are in the one quarter of the programs classified as group A and that these programs exhibit some characteristics that make them vulnerable on the issue of perceived quality of training. For example, a higher percentage of R1s transfer from these programs to continue internal medicine training in other programs. These transfers may be the result of residents seeking better training experiences or may reflect attempts by the training programs to rid themselves of unsatisfactory residents. In either case, the image produced by so many residents transferring is inconsistent with a high-quality training experience. As another indicator of quality, these programs offer only 60% of their positions through the NRMP and fill only 62% of those offered. This difficulty suggests that many applicants may not perceive these programs as particularly attractive. However, because these programs contain fewer residents in preliminary tracks and because their residents have a relatively low subspecialization rate despite the high concentration of FNIMGs, these programs can make important contributions toward the desired goal of having 50% of internal medicine residents pursue careers in primary care.

Fifth, implementing the 110% rule or alternative methods to decrease the number of internal medicine programs will necessarily stress health care services in urban, underserved areas, including many of the hospitals in groups A and B. Those hospitals, facing a reduced pool of FNIMGs, will have to discover ways of attracting USMGs to their training programs or alternative methods of delivering medical care to their respective populations. To compensate for lost service from trainees, program directors have indicated that their hospitals will probably rely on a combination of faculty, full-time practicing physicians, physician assistants, and nurse practitioners. We have yet to investigate the strategies these programs will use to attract USMGs. However they are accomplished, meeting the service needs of the hospitals and attracting USMGs will be most challenging for these programs.

In conclusion, data from the National Study of Internal Medicine Manpower show the continued increase of the number of women and FNIMGs in internal medicine. Because in the past women have subspecialized at a lower rate than men and FNIMGs have subspecialized at a higher rate than USMGs, these trends will affect efforts to increase the number of general internists. We have identified a large proportion of internists who do not enter subspecialty training directly from residency training and who thus may be encouraged to pursue general medicine. Finally, we have identified a group of programs with large proportions of FNIMG residents that are likely to be particularly vulnerable to reductions under a reformed system of graduate medical education. Although the sources for concern should be recognized, we should not lose sight of the contributions these programs can make to producing general internists of high quality.

Abbreviations

FNIMG: Foreign national international medical school graduate

IMG: International medical school graduate

NaSIMM: National Study of Internal Medicine Manpower

NRMP: National Residency Matching Program

R1: First-year resident

R2: Second-year resident

R3: Third-year resident

USIMG: U.S. citizen international medical school graduate

USMG: U.S. medical school graduate