Dozens of factors can influence medical student career choice of psychiatry and other specialties (1—7). Our understanding of these influences is limited, however. Further, as the socio-economics and politics of health care change over time, the extent of a factor's influence may change. Additionally, since 1994, relatively little has been ascertained about influences on medical student selection of psychiatry. Therefore, psychiatrists and psychiatric organizations, with a stake in increasing the numbers of U.S. graduates matching into psychiatry, could benefit from a study of some of the variables that may have influenced recruitment recently.
It is generally accepted that the proportion of students selecting psychiatry at each school results from a complex and not-fully-understood interplay between a) extrinsic factors, such as national trends (5,6), including the increased medical student career choice of psychiatry in 1998 to 2003 (5) and regional trends such as the higher recruitment that took place in southern schools in 1991 and 1992 (4); and b) intrinsic factors (specifically the characteristics of each department), particularly the quality of educational programs (6). The goal of this study was to learn the extent to which extrinsic factors (and one intrinsic factor: the proportions of international medical graduates (IMGs) in the psychiatric residency) influenced psychiatric career choice at U.S. medical schools from 1999 to 2001.
What have we learned from previous studies of influences on student selection of psychiatry? Regarding extrinsic factors, Weissman et al. (4) demonstrated that in 1991 and 1992, southern schools and publicly funded schools had significantly more senior students match into psychiatry than did other schools. In 1993, Kassebaum (1) showed that students who had high medical school debts upon graduation—almost always from private schools—were more apt to choose high-paying specialties, especially surgical subspecialties.
Regarding intrinsic factors, Nielsen (9) reported in 1980 that there was a modest, but significant, positive correlation between proportions of students matching into psychiatry (PMP) and the quality of medical student education programs, as rated by National Institute of Mental Health (NIMH) site visitors. He also found a significant correlation between the PMP of a current year and that of the prior year, demonstrating moderate year-by-year within-school stability in psychiatric specialty choice.
In 1981, Sierles (7) reported that PMP was higher in psychiatry departments a) with high prestige relative to the school's other departments (based on anonymous self-reports by educators at each school) or b) where the director of medical student education in psychiatry was the recipient of a teaching award. Additionally, anecdotal reports from high-recruiting schools in 1981 and 2002 suggest that high recruitment is associated with departments that have a major commitment to medical student education and faculty development (10,11).
Past research about psychiatric career choice, most of which was conducted before 1990, indicated the following: Euro-American Protestant and Asian American students were less apt to choose psychiatry than were Jewish, Catholic, African American, Hispanic, and Native American students (6,12—17). Students from urban and suburban backgrounds were more apt to choose psychiatry than those from rural areas, who tended to prefer family medicine (2,18,19).
Students from schools with psychiatry departments with high NIMH RO1 funding were no more likely to select psychiatry than were other students (20). Studies in 1982 (7) and 2002 (21) demonstrated that psychiatry clerkship length was unrelated to career choice of psychiatry.
To date, studies attempting to link PMP with the extent of local managed care penetration or the proportion of IMGs in the psychiatry residency do not exist. In a survey of senior medical students who were asked to rank internal medicine residencies that were described in vignettes in which the variables—including the proportion of IMGs in the residency—were manipulated, Riley et al. (22) demonstrated that student perceptions of the comparative desirability of each residency were influenced by the proportion of IMGs listed in the vignette. Although this was not a study of recruitment into internal medicine, psychiatry, or any other specialty, it is possible that students' attitudes about the desirability of psychiatry may be affected if a psychiatry department has a larger proportion of IMG residents than other departments at a particular school. It may also be possible that the IMG proportion is associated with the quality of that department's teaching.
To learn more about the factors influencing psychiatric career choice at U.S. medical schools from 1999 to 2001, we tested the following hypotheses: a) Because the many external and internal factors that can affect each school and department tend not to change simultaneously each year (although exceptions exist), significant year-by-year within-school continuity of PMP would occur; namely, that at each school, the prior year's recruitment would tend to predict the present year's recruitment (9). Over long intervals, notable within-school shifts in PMP would occur. b) Southern schools (4), schools that were publicly funded (1,6), and schools with lower tuition (6) would have significantly higher PMP. c) Schools with lower proportions of IMGs in the psychiatry residency would have higher PMP (22).
We also hypothesized the following: Schools with admissions policies preferring candidates who stated an interest in primary care would have neither a higher nor lower PMP because, although the admissions committees would seemingly select against psychiatrists, students interested in primary care are likely (23) to have a major interest in the doctor-patient relationship, which is a major "draw" for psychiatrists. The proportion of students matching into psychiatry would be lower in schools where admissions committees had formal preferences for students from rural backgrounds (2,18,19). At schools where the ethnic compositions exceeded the national average for African American, Hispanic, and Native American students (data about student bodies' religious compositions were unavailable) or where an admissions preference for underserved minorities was stated, PMP would be higher (6,12—17). Schools where proportions of Asian Americans exceeded the national average would show lower PMP (6,17).
Finally, we made these hypotheses: Based on perceptions that psychiatry has been harder hit by managed care than other specialties (24), schools with more local managed care penetration would have lower PMP. Clerkship length and PMP would be unrelated (7,21). Presuming that schools with a psychiatrist dean would regard psychiatry highly, the 10 schools with psychiatrist deans would have higher PMP. Additionally because no prior study has demonstrated—and there is no intuitive reason to predict—an association between the year a school was founded and PMP, school age would have no association with PMP.
For most analyses, the dependent variable was the PMP at each of the 125 U.S. medical schools during the 1998 to 1999, 1999 to 2000, and 2000 to 2001 academic years. We obtained the PMP for all 125 schools as follows: In July 2001, we e-mailed deans of student affairs at each school and asked them what the PMP was in their school in 1999, 2000, and 2001. Deans who did not respond the first time were e-mailed again. Simultaneously, we obtained the Electronic Residency Admissions System (ERAS) Match results from the NRMP, promising to keep individual schools' PMP confidential.
Independent variables were obtained thus: The PMP at each school from 1975 to 1978 was published in the Journal of Psychiatric Education (25). In 1993, the PMP from each school from 1987 to 1992 was presented at the APA Council on Education and Career Development (26), which one of us (F.S.S.) attended. The following were obtained from the AMA and the Association of American Medical College's Annual Medical School Questionnaires for 1999 to 2000 and 2000 to 2001: age of the school; admissions preference for applicants from disadvantaged or rural backgrounds or preferring primary care; tuition; and proportion of African American, Asian American, Euro-American, Hispanic and Native American students (27; Barbara Barzansky, personal communication, 2002).
We obtained the proportion of psychiatry IMGs at each school from the APA's 1997 Directory of Psychiatry Residency Programs (28), which was the most recent data available. We also obtained the most recent data (also from 1997) on the extent of local managed care penetration from Harvard University's Health System Consortium (29; Eric Campbell, personal communication), as measured by market stage. The components of market stage include: health maintenance organization (HMO) penetration, HMOs with enrollees earning $100,000 or more, percentage of employees in the top three HMOs, hospital occupancy days/1,000 beds in the system, percentage of group specialists capitated, percentage of Medicare population in HMOs, percentage of Medicaid population in HMOs, and commercial HMO premiums (29). We ascertained clerkship lengths from the AAMC's 2000 to 2001 curriculum directory (30). We obtained information on deans from a listing that was published in Psychiatric News (31).
Geographic region was as such: Northeast—Connecticut, Maine, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island and Vermont. Southeast—Alabama, Arkansas, Delaware, District of Columbia, Florida, Georgia, Kentucky, Louisiana, Maryland, Mississippi, North Carolina, South Carolina, Tennessee, Virginia, and West Virginia. Midwest—Illinois, Indiana, Iowa, Kansas, Michigan, Missouri, Minnesota, Nebraska, North Dakota, Ohio, South Dakota, and Wisconsin. Southwest: Arizona, New Mexico, Oklahoma, and Texas. West—Colorado, Nevada, and Utah. West Coast—California, Hawaii, Oregon and Washington. The Puerto Rican (Cayey, Ponce, San Juan) schools comprised another group. Alaska, Montana, Idaho, and Wyoming have no medical schools.
We used an SPSSx 11.0 system (32,33). The dependent variable, PMP, is interval-ratio. When an independent variable was categorical, we used independent samples t tests or analysis of variance (ANOVA). When an independent variable was interval or ratio and the variables were not normally distributed, we used Spearman product-moment correlations. When we conducted multiple correlation analyses for year-by-year and decade-by-decade variation in PMP (t2), a Bonferroni correction revised the cutoff level for significance to p < 0.007. For geographic region, we used Tukey's post hoc analysis to identify significant pairwise mean regional PMP differences whenever we discovered a significant main effect of region.
Because we used one dependent variable and multiple independent variables, we used multiple regression analysis with simultaneous entry of predictor variables to control for associations among independent variables. For this analysis, we chose those independent variables for which the association with PMP was statistically significant or showed a statistical trend.
Characteristics of the independent variables are displayed in t1 and t2.
Between-year correlations are displayed as a matrix in t2. PMPs for any two successive years were significantly correlated, even after Bonferroni correction. As between-year intervals increased, correlations diminished and became nonsignificant.
Other Factors Affecting PMP
t3 summarizes the other independent variables and their associations with PMP. There was a modest but significant negative correlation between PMP and the proportion of IMGs in the psychiatry residency. When IMG percents are displayed as quartiles from 0% to 24% IMGs to 75% to 100% IMGs (t1), there appears to be a "break point" between 50% to 74% IMGs and 75% to 100% IMGs. However, when we computed ANOVA by quartile, the between-quartile significance was only a trend (N=113, df=3, F=2.194, p=.093).
From highest to lowest, the PMP for geographic regions was 1) Puerto Rico, 2) West Coast, 3) Southwest, 4) Southeast, 5) Midwest, 6) Northeast and 7) West. Although the omnibus ANOVA was significant, subsequent Tukey post hoc comparisons revealed no significant group differences.
There was no significant association between PMP and tuition, although a trend in the predicted direction was detected, and there was no significant association between PMP and the extent of managed care penetration, although a trend in the opposite direction was detected.
Additionally, there was no association between PMP and the following factors: whether a school was publicly or privately funded; whether an admissions policy favored rural applicants, underserved minority applicants, or applicants stating an interest in primary care; whether a school had a psychiatrist dean; the length of the psychiatry clerkship; the age of the school; or the ethnic composition of the student body (i.e., whether the proportions of Euro-American, underserved minority, or Asian American students exceeded the national average).
Using simultaneous entry of predictor variables, we conducted multiple regression analysis for those independent variables that were significantly associated with PMP (e.g., PMP during the preceding year, proportion of IMGs in the psychiatry residency, and geographic region) and for those that showed a statistical trend in the correlation between the variable and PMP (e.g., tuition, extent of managed care penetration). For the first analysis, we used PMP in 2001 as the dependent variable. PMP in 2000, percentage of IMGs in the psychiatry residency, tuition, extent of local managed care penetration, and geographic region were used as independent variables. To do this, we converted geographic region from a polytomous variable to a dichotomous variable in which the West Coast was one possibility and the other regions combined constituted the other possibility. In this first analysis, PMP in 2000 (B=.374, SE=.104, beta=.368, t=3.612, p=.001) was the only predictor variable that remained significant.
For the second analysis, we used mean PMP for 1999 to 2001 as the dependent variable. Percentage of IMGs in the residency, tuition, extent of local managed care penetration, and geographic region were used as independent variables. In this second analysis, we found that IMG percentage (B=−.013, SE=.006, beta=−.223, t=2.143, p=.035) was the only predictor variable that remained significant.
A school's PMP is not random. National trends (an extrinsic factor), which change from decade to decade and are discussed elsewhere (5,6), influence all schools. In this study, the best predictor of a school's PMP is its PMP from the previous year.
None of the regional or school-related extrinsic factors in this study, including southern location and public funding, predicted PMP. Departments cannot expect to "luck out" by their students choosing psychiatry based on regional or school-related extrinsic factors.
By the process of elimination, intrinsic factors must be most influential in determining a school's PMP (over and above national trends), and departments wishing to improve their recruitment must consider them. The only intrinsic factor that we studied, which was proportion of IMGs in the psychiatry residency, was significantly correlated with PMP. But the correlation is only modest. Although we predicted this finding, we did not determine its meaning. For example, we could not ascertain whether having a higher proportion of IMGs in the psychiatry residency a) means that medical students are less well taught or b) is a proxy for the department's reputation and the desirability of its programs (22).
To answer the question pertaining to the association between PMP and IMGs in the residency, further research using multivariate methods—similar to those used by Sierles in 1982 (7) and in this study—should address the relative impact on PMP of intrinsic factors, including the resources devoted to the department's medical student education programs, the quality and characteristics of these programs, the proportion of IMGs in the psychiatry residency, characteristics and in-house prestige of the department, and characteristics of the director of medical student education.
Despite the significant inverse correlation between PMP and IMG proportion in the residency, residents must be selected based on their overall competency, not their country of origin or where they obtained their medical degree (5). Regardless of the IMG or American medical graduate (AMG) status of the faculty or residents, students should be exposed to the best attending and resident teachers, residents should be taught how to teach students, and residencies should have acculturation programs for their IMG residents (34).
Although the data from this study provide relatively little guidance about what intrinsic factors a department should address to improve recruitment, it is reasonable to speculate, based on the literature (6,7,9—11). The most important intrinsic factor that we can control is the quality of psychiatric education programs, in the interest of education as well as recruitment. As Scully (35) wrote, "Excellence in teaching is the critical factor in recruitment. If we convey to our students the importance and the excitement of psychiatry, there are two good possibilities. Either the students will choose a career in psychiatry or they won't. If they do, we can select from the best qualified; if they don't, they will have an appreciation of the importance of psychiatry for their patients, leading to good referrals and addressing the still real problem of stigma." With good education as the primary goal and recruitment as a secondary objective, departments should be able to provide the resources in personnel, time, and mentorship (36—38) to produce the best possible psychiatric education programs.
Recommendations for providing the best possible programs, presented in detail elsewhere (36—39), include: a) producing a well-organized preclinical behavioral science course(s) that highlights the department's strengths and has a highly visible course director who teaches a sizeable portion of the classes and provides feedback for course faculty (37); and b) producing a well-organized clerkship with a highly visible director who has adequate protected time and staff support, a dedicated faculty with sufficient teaching time, and an environment where students have a thoughtful combination of patient care responsibility, faculty supervision, and didactic activities (36,38).
If resources are limited, perhaps students could still be influenced to choose psychiatry by a small number of charismatic faculty members who teach considerably in an efficiently run preclinical course and clerkship. This fits Sierles' 1982 finding of a significant association between PMP and the winning of a teaching award (perhaps an indicator of "charisma") by the director of medical student education (7).
It is also tempting to recommend that psychiatrists should play an active role on medical school admissions committees for recruitment's sake as well as for service to the school. Unfortunately, the literature on psychiatrists' participation on admissions committees is too sparse to support an evidence-based recommendation. The only study of PMP and admissions committee participation by psychiatrists (7) showed a modest correlation between the two that became nonsignificant following regression analysis. Additionally, Owen's study (40) on admissions committee members' ability to predict which candidates will become primary care physicians showed that the members often made inaccurate predictions about applicants' career plans.
It would have been useful to determine each school's interest-change ratio, as Weissman did in 1994 (4). Interest-change ratio is the PMP divided by the proportion of students at each school who expressed an interest in psychiatry at or before matriculation (4). We did not have access to data about students' interest in psychiatry at or before matriculation. Although interest-change ratio and PMP are closely related, interest-change ratio is probably more sensitive to intrinsic factors, and PMP is probably more sensitive to extrinsic ones. To confirm the close relationship between these two outcome measures, we assessed PMP alone for 1992, and our results replicated Weissman's finding of the strong association in 1992 between geographic region, public funding, and PMP.
From this study, we conclude that recruitment at each school from 1999 to 2001 could not be accounted for by regional or local extrinsic factors, and we infer that the psychiatry departments' programs (i.e., intrinsic factors) are most important. In the interest of psychiatric education as well as recruitment, the intrinsic factors deserve the most attention and should be the subject of more research.
The authors thank Barbara Barzansky, Ph.D., Eric G. Campbell, Ph.D., Elizabeth Lostumbo, J. Jon Veloski, M.A., and H. Jonathan Polan, M.D., for their assistance. Earlier versions of this paper were presented at the annual meeting of the American Association of Directors of Psychiatric Residency Training in San Juan, PR on March 6, 2003; the American Psychiatric Association Meeting in San Francisco, CA on May 21, 2003; and the annual meeting of the Association of Directors of Medical Student Education in Psychiatry in Jackson Hole, WY on June 13, 2003.