Several reports in the literature detail workforce issues that confront various specialties in medicine. Lyttle and Levey (+1) described demographic characteristics, recruitment practices, postresidency career plans, and their policy implications for internal medicine residents. Kwakwa and Jonasson (+2—+4) have published reports that systematically examined demographics, distribution, and career paths of surgical residents. Kahn et al. (+5) have examined the percentages and characteristics of each U.S. medical school's graduates entering family practice residencies. Similar studies have been conducted in pediatrics (+6), pediatric surgery (+7), neurology (+8), rehabilitation medicine (+9), pathology (+10), and several other medical specialties. Psychiatric literature is marked by studies that have examined various aspects of postresidency psychiatrists (+11—+13). However, relatively few studies have examined the national psychiatric workforce at the graduate medical education (GME) level. Such studies can give policy planners and professional advocates important information on the supply and distribution of future psychiatrists. Accurate quantification in the GME arena is crucial for deliberations on GME funding by governmental agencies and regulatory bodies. This information is also important regarding debates on the health care provided to minorities and concerns about the cultural competence of future practitioners. A complete picture might also help inform the nation's continuing debate on immigration policies for physicians.

This study was conducted to examine the trends in the supply, distribution, and demographics of psychiatry residents during the 1990s, which witnessed widespread attempts to downsize residency training programs in various specialties (+14). The extent to which the predicted downsizing of psychiatric residency training programs had actually occurred and how it affected training programs of different sizes and locations was examined (+15). Finally, the status of international medical graduates (IMGs) vis-à-vis United States Medical Graduates (USMGs) in the GME workforce was compared to detect trends with a view toward understanding what the impact of current policies regarding IMGs might be on subsequent workforce patterns.

Data for this study were obtained from the American Medical Association's (AMA) Annual Survey of GME Programs. Program directors provided information on program characteristics such as clinical and research facilities and the learning environment, while the Census of Graduate Medical Trainees portion of the survey presents data collected on individual residents. The AMA supplied data regarding the composition of psychiatry-training programs from 1992 to 1995, 1997, 1999, and 2000, including the number of international and U.S. medical graduates in each program. This information was supplemented by data from the Green Book (+16) as well as the annual survey of directors of general psychiatry and fellowship programs (psychiatry resident census) conducted by the APA's Division of Education (+17). The APA conducted this survey of both accredited and nonaccredited programs from 1968 through 1996 but discontinued it in 1997 for financial reasons. In 1999, the survey was replaced in a modified form by extracting data from the AMA GME survey. In addition to these sources of information, data from the annual GME surveys of the AMA published in various Fall medical education issues of The Journal of the American Medical Association were used to obtain a more complete picture of psychiatric residency training (+18).

The data set for the years 1992 through 2000 contained several sources of discontinuity: some programs were created, while others closed; there were years that could not be accounted for; and data of interest for this study were missing for a number of years for several programs, particularly for 1997. In order to obtain a complete set of data, information from 1997 was eliminated and a group of 145 programs, for which there were complete data from 1992 to 1995, 1999, and 2000, was used in this analysis. These 145 programs represent 80% of the 182 programs existing in 2000.

In the 1992 survey, programs were sorted into four categories based on the total number of residents in a program, which was, in effect, each program's starting size. With respect to size, programs retained their relative positions over time: the correlation between the total number of residents in each program in 1992 with that of 2000 is 0.795, p < 0.0001. The 1992 data were broken at quartiles yielding four categories of program size as an independent variable: smallest (<24 residents), smaller (24—38), larger (39—51), and largest (>51). The other independent variable, according to the APA classification, was region of the United States.

These two four-level independent variables were analyzed in a factorial analysis of variance (ANOVA) doubly multivariate repeated measurements model, using SPSS v.10.1 Advanced Statistics. A conservative strategy was adopted (i.e., the alpha level for considering significant results was set at 0.02), and contrasts that did not assume homogeneity of variance were used. In this analysis, the 145 individual programs were the subjects.

Also studied was the effect of the medical school setting versus the nonmedical school setting (type) on the trends in the supply, distributions, and demographics of psychiatry residents. However, only 22.1% of the sample programs were situated in nonmedical school settings. Adding this variable to the ANOVA resulted in many cell sizes that were too small or even zero, rendering the ANOVA invalid. To examine the effects of the type of program, we ran two additional, separate factorial ANOVAs, doubly multivariate repeated measurements model, including type as an independent variable in each but excluding program size in one ANOVA and region in the other.

There was a significant decline in the number of residents during the years studied (+Table 1; F = 75.246, df = 5, p < 0.001). Pairwise comparisons yielded significant differences between each year (p < 0.01) except for that between 1994 and 1995. IMG numbers showed a different pattern over time when compared to that of the USMGs (interaction of IMG/USMG status by years F = 46.990, df = 5, p < 0.001): the number of IMGs increased slightly and gradually from 1992 to 1994, during which time the number of USMGs declined markedly. Both IMG and USMG numbers leveled off between 1994 and 1995. These numbers declined at approximately equal slopes between 1999 and 2000 for both groups. Between 1995 and 2000, the total number of residents declined by 26%, the number of USMGs by 30.6%, and the number of IMGs by 18.4%.

Between 1995 and 2000, according to the APA census, general psychiatry residency training programs consisting of PGY-I to PGY-IV classes showed an 11.5% reduction in IMG numbers, a 20.4% reduction in USMG numbers, and a 16.8% reduction in the total number of residents, and the total number of programs decreased from 197 to 182. The APA census demographics also show the resident workforce becoming more diverse, more heavily represented by females, and older than ever before.

The median training program size was 38 residents. Programs of dissimilar sizes fared differently over the years (+Table 2; interaction of program size by Years F = 13.879, df = 15, p < 0.001). While, in 1992, the larger and largest programs declined from an average of 45.1 and 69.0 residents, respectively, to 34.2 and 56.1 residents, respectively, in 1995, the smallest programs increased from 18.1 to 19.9 from 1992 to 1995, and the smaller programs showed only slight decline during these years. The larger and largest programs suffered the steepest decline from 1999 to 2000 (down 23% to 25%, respectively). In contrast, the smaller programs declined at only 12.5%, and the smallest programs nearly held steady with a 2.4% decline.

While the smallest and the smaller programs nearly offset the decline in USMGs over the years by corresponding increases in IMGs, the larger and largest programs did not. From 1992 to 1994, the larger programs experienced a sharp decline in USMGs, but the average number of IMGs per program remained nearly the same. The number of USMGs declined at a somewhat steeper rate from 1999 to 2000 than did the number of IMGs. Thus, while IMGs comprised 21.1% of residents in larger programs in 1992, they comprised 38.5% in 2000. The largest programs display a similar pattern, though the initial decrease in USMGs is not as steep as in the larger programs. The largest programs were comprised of 17.0% IMGs in 1992, and 22.1% in 2000.

The Northeast had the greatest proportion (37.9%) of training programs, with the South (24.1%) a distant second, followed by the Midwest (22.8%) and the West (15.2%). There were significant inter actions between IMG/USMG status and region (+Table 3; F = 7.775, df = 3/129, p < 0.001) and between IMG/USMG status by region by years. While the mean number of residents per program, averaged for all years, was nearly equal (from 16.6 in the South to 17.8 in the Northeast), the regions showed differing mixes of IMGs and USMGs and marginally significantly changes in their resident mix (F = 1.894, df = 15, p = 0.021). The trend in all regions is that there was a convergence of the IMG and USMG proportions during the years 1992 to 1995: the decreases in USMG numbers was partially compensated by increases in IMG numbers. From 1999 to 2000, there was a nearly parallel decline in the numbers of IMGs and USMGs.

In 1995, the 1992 to 1995 pattern resulted in equal numbers of IMGs and USMGs in the Northeast region. For the western region, the 1992 to 1995 years resulted in the extremely imbalanced USMG/IMG ratio of 7:1 in 1992. The still disproportionate ratio of 3.6:1 was observed in 1995. The Midwest and the South showed similar but less extreme patterns during the period from 1992 to 1995. In 1999, the Northeast region maintained approximate parity of IMGs and USMGs, whose numbers show a parallel decline in 2000. The western region retained a large disparity between the number of IMGs and the number of USMGs and a parallel decline of both from 1999 to 2000. In the Midwest, the 1995 USMG/IMG ratio of 2:1 showed balance by 1999, with a ratio of 1.2:1. Although not as disproportionate, the USMG/IMG profile for programs in the southern region was similar to that of the programs in the western region, with a ratio in 1992 of 4.5:1, which tapered down to 1.8:1 in the year 2000.

Of the programs studied, 77.9% were medical school based or medical school affiliated. The ANOVA that included the type of program and the region yielded significant effects for type by region by years (F = 3.331, df = 5, p < 0.01). In 1992, medical school programs had a USMG/IMG ratio of 4:1, and by the year 2000, the ratio was 2:1. The USMG/IMG ratio of nonmedical school programs in 1992 was 2:1. By the year 2000, the number of USMGs was nearly equal to that of IMGs.

In 1992, programs that were totally comprised of IMGs did not exist. In the year 2000, however, there were 11 programs entirely populated by IMGs. In 1992, few programs had more than 40% IMGs. In 2000, IMGs constituted more than 40% of more than one-half of all programs. The median percentage of IMGs in 1992 was 22.1%, while in 2000, it was 51%.

One should keep in mind that there are a number of limitations to the study. Loopholes in the data on the psychiatric GME workforce exist. The AMA data were not available for every year between the years 1992 to 1999. The response rate to the AMA GME survey was not always 100%, several respondents omitted various items, and data are missing for several years. Additionally, the AMA only surveys accredited programs. Consequently, data from nonaccredited psychiatric fellowship programs in research, consultation, and liaison and nonaccredited programs approved by the American Osteopathic Association (AOA) that used to be reported in the former APA Resident Census are not available. As a result, there were differences noted in the numbers reported by the APA Resident Census and the AMA survey of GME 1996. Future AMA surveys could be improved by including information that was available in former APA surveys.

Future survey information pertaining to country of origin, race, and ethnicity would be helpful in addressing the current concern regarding the decrease in the number of minority students entering medicine: To what extent are foreign medical schools providing a pathway for U.S. minorities to enter the U.S. physician workforce? Additionally, information regarding the visa status and the postresidency career plans of graduating psychiatry residents would be useful in monitoring the Exchange Visitor Waiver Program. Finally, data on organizations that sponsor training programs would be useful in identifying the so-called "fault lines" in residency training—in this day and age of constant turmoil in GME financing—and providing information beneficial for future planning.

Following the release of a report of the Graduate Medical Education National Advisory Committee (GMENAC) (+19) as well as several Council on Graduate Medical Education (COGME) reports during the 1990s (+20,+21), there was considerable concern among policymakers that the nation was heading toward a significant physician surplus and that a generalist/specialist imbalance that favors specialists was occurring. Consequently, several influential organizations recommended reductions in the total number of GME positions and decreases in the number of specialists, but increases in the number of primary care physicians (+22—+24). This effort was bolstered by significant reductions in GME funding by Medicare that were incorporated in the 1997 Balanced Budget Act (BBA) (+25). Several large academic institutions downsized, and smaller programs merged. Medicare supported the Greater New York Hospital Association's attempts to help its member hospitals reduce their GME positions by 25% over a 5-year period (the New York Demonstration Project) (+26). It appears that psychiatry has responded to these initiatives positively and that between 1992 and the year 2000, the total number of general psychiatry residency programs diminished by 7.6%, and the total number of residents in these programs dropped by 16.8% (+17,+18).

In addition to the GME financing the push to downsize, there were other forces at work, resulting in the reduction in residency positions. In the 1990s, secondary to the managed care impact on medical practice, the locus of clinical care shifted from the inpatient to the outpatient area, which resulted in the closure of excess inpatient beds (+27,+28). Consequently, hospitals closed residency lines that had been allocated for the care of inpatients. In psychiatry, the reduction in resident numbers was not uniform among the four categories of program size. The larger and the largest training programs (>51 and between 39 and 51 residents) declined modestly during the first 4 years and then more steeply during the last 2 years when the full impact of downsizing efforts was felt. It is not surprising that the larger programs suffered greater declines because they could afford to close residency lines without jeopardizing their survival. In contrast, it is likely that the smaller programs, because of their proximity in numbers to the critical mass requirement (12 residents) of the residency review committee (RRC), could not afford to downsize any further.

In addition, during the 1990s, IMGs made significant inroads into the type of programs (i.e., medical schools) and geographical areas (i.e., outside of the Northeast) that were not open to them in previous decades. This broad acceptance of IMGs challenges the two-tier professional situation that existed for USMGs and IMGs in psychiatry during the 1970s and the 1980s: USMGs were trained mainly in medical school programs while IMGs were trained mainly in large nonmedical school programs (+29). By the year 2000, IMGs constituted 100% of the residencies in 11 training programs. This is in contrast to 1992, when no training programs were comprised entirely of IMGs. These findings underscore the fact that psychiatry residency programs developed substantial dependence on IMGs for filling residency classes during the 1990s.

Another important finding of this study is that the supply dynamics of USMGs and IMGs as residents in psychiatry changed during the late 1990s. Historically, whenever the number of USMGs entering psychiatry decreased (+30—+33), IMGs ameliorated the shortfall (+29,+34—+35), thus keeping the total number of residency positions constant. However, in 1999 and 2000, the numbers of both USMGs and IMGs entering psychiatry decreased. Whelan et al. (+36,+37) reported that consequential to the introduction of the Clinical Skills Assessment (CSA) Examination in 1998 by the Educational Commission for Foreign Medical Graduates (ECFMG), the number of IMGs taking Step I of the United States Medical Licensing Examination (USMLE) decreased by 45.5%, and those registered to take Step II decreased by 38.1%. Additionally, the total number of ECFMG certificates issued annually decreased from 9,000 to 12,000 (1995—1998) to fewer than 6,000 (1999—2001). The introduction of this exam most likely influenced reduction in the number of foreign IMG entries, which is cause for concern because of the different roles played by USMGs and Foreign IMGs in the U.S. health care system.

One has to wonder about the potential impact of the reduction of IMGs on the quality and types of services that are generally rendered by IMGs. In a study conducted to assess the progress of IMGs over an 8-year period, Gary et al. (+38) found that when compared to U.S. citizen IMGs (USIMGs), foreign IMGs passed the ECFMG examination sooner, but they took longer to find residency positions. In addition, some reports show that foreign IMGs score higher on Step I and have equal scores on Step II, when compared to USIMGs (+39). In a study that examined the participation of IMGs in GME and hospital care of the poor, Whitcomb and Miller (+40) pointed out that among 106 hospitals characterized as IMG programs (having more than 50% of residents IMGs), 77 (73%) could be considered dependent on IMG residents to deliver care to the poor. Fifteen psychiatric institutions were among these 77 hospitals. With participation of IMGs being more widespread in current training programs in psychiatry (+Figure 1), one may predict that their absence will have a far greater impact on hospital care for the poor. Other studies point out that more frequently than USMGs, postresidency IMGs work in the public sector, treating sicker and poorer patients. They have different practice patterns than USMGs (+41) and are a major source of physician supply in physician shortage areas (+42).

Concerns have already been expressed that the physician surplus that was forecast earlier has failed to materialize, and there may soon be a shortage of specialists (+43,+44). Similarly, some reports suggest a net replacement deficit of practicing psychiatrists of more than 100 per year (+45). This deficit has occurred before the full impact of the recent downsizing efforts will be felt. It is also important to note that these developments are occurring against a backdrop of the expansion of the nonphysician mental health workforce, including psychologists, nurse practitioners, and social workers (+46), and this may not be a good sign for the continued viability of our discipline.

In conclusion, psychiatry-training programs have been both downsized and reduced. This will inevitably result in fewer psychiatrists entering the field in the future, perhaps worsening both the shortage and maldistribution of psychiatrists. Presently, the number of USMGs entering the field is increasing, but if their past specialty preference patterns are any indication, psychiatry will not be able to depend on them as a guaranteed supply of future practitioners. Although psychiatry turned to IMGs to fill positions by default in the past, it may not be able to maintain an adequate supply of them in the future. The field will have to decide whether it can afford any more residency downsizing in light of emerging evidence of a shortage of psychiatrists. If the answer is negative, the current downsizing and reduction in training programs must cease. Simultaneously, the field must demand that the restrictive immigration and examination policies be eased in order to expand the pool of IMGs, while redoubling its recent successful efforts to make psychiatry more attractive to USMGs. In the final analysis, the field is better served by having an adequate number of competent psychiatrists, regardless of their medical school of origin, than to have its role usurped by other disciplines due to lack of sufficient numbers of practitioners.

The author thanks Joel Yager, M.D. and Larry Faulkner, M.D. for reading the paper and making suggestions and Arthur Meinzer, Ph.D. for his statistical analysis.