Design

Association of American Medical Colleges (AAMC) staff provided de-identified and individually linked data from various sources in 2009 for all 129,860 matriculants in U.S. Liaison Committee on Medical Education (LCME)-accredited medical schools during academic years (AYs) 1993–1994 through 2000–2001. For analysis, we included graduates in calendar years between 1997 and 2012, who were appointed to full-time instructor or assistant professor faculty positions at LCME-accredited medical schools between 2000 and 2020. We obtained follow-up data from the AAMC Faculty Roster data on April 3, 2020, which allowed for up to 19 years of follow-up after initial appointment. This study of de-identified data was reviewed by the Institutional Review Board (IRB) at Washington University School of Medicine and was approved as non-human-subjects research (IRB # 201808151). We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines [18].

Measures

From the American Medical College Application Service® [19], we received data for self-identified sex (male or female) and race/ethnicity (categorized as White, Asian, or URiM [including Black/African American, Hispanic/Latino, American Indian/Native Alaskan, Native Hawaiian/Other Pacific Islander]). At the time of data collection, only one option for race could be selected. We also obtained data from the AAMC for degree program at graduation (MD, including BA/BS-MD; MD-PhD; or MD-other-advanced degrees).

From the AAMC Matriculating Student Questionnaire [20], voluntarily completed at matriculation, we received data for parents’ education (first-generation college graduate [neither parent completed 4-year BA/BS degree], continuing-generation college graduate [at least one parent completed BA/BS degree with or without further graduate or professional education], or both parents’ education was unknown). MSQ response rates for this cohort were 93.5% (Matthew D, AAMC Data Operations and Services, email communication, August 30, 2022).

From the AAMC Graduation Questionnaire [21], voluntarily completed by graduating medical students, we received data for career plans at graduation (full-time university faculty in basic or clinical sciences; non-academic clinical practice/other/non-university research scientist/undecided; and missing), and total debt at graduation (no debt, $1-$99,999, ≥ $100,000, or missing). GQ response rates for this cohort averaged 74.0% (Matthew D, AAMC Data Operations and Services, email communication, August 30, 2022).

From the AAMC’s Graduate Medical Education (GME) Track database [22], we received data about whether a resident participated in at least one year of dedicated research during GME. Accreditation Council for Graduate Medical Education–accredited residency program directors and institutional officials voluntarily complete the National GME Census, with annual response rates averaging 92.6% (Roskovensky L, AAMC Data Operations and Services, email communication, September 2023).

The AAMC acquired information about federal grant awards received from the National Institutes of Health (NIH) Information for Management Planning Analysis and Coordination (IMPAC II) database, linked these data for individuals in our database, and provided the de-identified data to us in August 2014. We created a 3-category variable for analysis of grants awarded to an individual prior to promotion to senior rank or to attrition from academic medicine (never received any grants, received R01 and/or research project grants (RPGs), or received only non-RPG grants [23], including pre- or post-doctoral (e.g., F, K, T, U, or other) grants.

The AAMC also provided data about whether a student attended a research-intensive medical school (yes or no) and whether a faculty member’s appointment was at a research-intensive institution (yes or no). These institutional variables were determined by each school’s annual ranking in the Top 40 for direct research expenditures from federal grants and contracts using data from the LCME Part I-A Annual Financial Questionnaire, which the AAMC administers [24].

The AAMC Faculty Roster data [25] included start and end dates of first and subsequent full-time faculty appointments, initial rank (i.e., instructor or assistant professor) and subsequent faculty ranks, and track at first appointment (nontenure-eligible track, tenure-eligible track without tenure, tenure not available at the institution, or missing). Faculty first appointed with tenure were excluded. The AAMC also provided data for department at first appointment; we created a 9-category variable for analysis (Internal Medicine, Family Medicine, Pediatrics, Obstetrics and Gynecology [OB/GYN], Surgery, Psychiatry, Neurology, Facilities-based specialties [i.e., Anesthesiology, Emergency Medicine, Pathology, Radiology], and all other clinical and nonclinical departments combined). We expected the number of months of follow-up after first appointment to be associated with attrition, as faculty more recently appointed may be less likely to leave academic medicine. We used faculty appointment start and end dates to compute the number of months of follow-up after first appointment until promotion to senior rank, attrition from academic medicine without promotion to senior rank, or date of censorship (February 29, 2020, if retained in academic medicine without promotion to senior rank). We created a dichotomous outcome variable comparing attrition from academic medicine, defined as leaving a full-time faculty position prior to promotion to senior rank, versus retention in a full-time academic-medicine faculty position, with or without promotion to senior rank.

Data analysis

We used chi-square tests to measure associations between our binary outcome (attrition vs. retention, with or without promotion to senior rank) and each categorical variable of interest, reporting frequencies (%). Analysis of variance was used to test the difference in mean (SD) number of months of follow-up after initial appointment by our outcome. A multivariable logistic regression model, controlling for number of months of follow-up after first appointment, was run to identify independent predictors of attrition from academic medicine (between-subjects variables) before promotion to senior rank (vs. faculty retention). We report adjusted odds ratios (aORs) and 95% confidence intervals (CIs); two-tailed p-values < 0.05 were considered statistically significant. Analyses were performed using IBM SPSS Statistics, version 29.0.2.0 (IBM Corp., Armonk, NY).

Demographics

In this study, we observed a greater likelihood of attrition before reaching senior rank among faculty who were female, Asian, URiM, and first-generation college graduates. Although increasing numbers of women have received faculty appointments over the years, there remain substantial sex disparities in career-advancement opportunities and salary compensation [26–28], which may lead women to leave academic medicine. Others have also reported Asian and URiM faculty are at greater risk of attrition compared with White faculty [4,12], and reasons for attrition may differ between Asian and URiM groups. Asian American medical students have reported exclusion from campus diversity efforts [29], and such experiences may persist as junior faculty. Therefore, supportive resources designed to promote retention of Asian faculty may be warranted [30], with some suggesting tailoring interventions, like mentoring, for different groups of faculty whose needs might vary [31]. Mentoring skills can be taught and can foster faculty retention and promotion [32,33]. Mentorship is perhaps especially important for groups known to be at risk for attrition. Training a diverse cadre of mentors, tailoring mentoring programs to meet specific faculty needs, and providing resources to promote retention were each reported to support the success of Asian and URiM faculty in academic medicine [34]. Furthermore, first-generation college graduates, many of whom were also from URiM groups [35], have reported a lack of institutional support during medical school, including lack of career advice and educational and psychological support [36], which also may be experienced after faculty appointment. Several publications of successful, national, research-mentoring programs, such as the NIH National Heart, Lung, and Blood Institute-funded PRIDE initiative, demonstrated promising results. These programs helped promote success in receipt of research grant awards, publications, faculty promotion and retention of a diverse cadre of faculty in academic medicine as well as reporting satisfaction with mentoring and increases in research self-efficacy [37–40]. Rigorous evaluations of faculty-mentoring programs designed to enhance faculty promotion and retention are needed to highlight institutions’ return on investment to justify ongoing support for such programs [41]. Faculty-mentoring programs that train faculty to become mentors also require evaluation [42].

Career plans and total debt at graduation

Faculty who planned any non-academic career upon medical-school graduation were more likely to leave academic medicine before reaching senior rank, similar to an earlier study of initial promotion and attrition with shorter follow-up [4]. Experiences during medical school can contribute to students’ consideration of academic-medicine careers [43–46]. For example, medical students’ engagement in research/authorship was associated with faculty receipt of career-development and other grant awards [47,48], and graduates who attended research-intensive (vs. non-research-intensive) medical schools were more likely to be appointed to full-time faculty positions in academic medicine [49,50]. Similarly, MD-PhD graduates who attended medical schools with federal Medical Scientist Training Program (MSTP) funding (vs. non-MSTP-funded) were more likely to be appointed to full-time faculty positions in academic medicine [51]. Thus, interventions during medical school that encourage and prepare students to pursue academic careers not only increase the likelihood of faculty appointment [50], but they also can have lasting impact thereby promoting retention.

That faculty who reported debt at graduation were more likely to leave academic medicine before reaching senior rank is not surprising, given the notable salary differences between academic physicians and physicians in private practice [52]. Medical-education-related debt is burdensome and has been associated with a lower likelihood of choosing an academic-medicine career among family medicine trainees [53]. Even before GME, the effect of debt was observed among medical students’ plans at graduation to pursue academic-medicine careers, with lower levels of debt < $150,000 showing lower odds of diminished interest and greater odds of sustained interest in pursuing academic-medicine career [43]. URiM residents across all specialties were more likely to have debt, in general, than their White and Asian counterparts [54]. Participating in loan-repayment programs could help reduce debt burden, although physician applications for loan repayment programs had declined by more than 30% between 2011–2020 [55]. Multiple interventions, such as educational scholarships, financial guidance/planning for trainees and early-career faculty, and loan-forgiveness programs, may be necessary to reduce debt-related stress and perhaps promote initial interest in an academic-medicine career and retention of physicians on the faculty and in the biomedical-research workforce [4,54].

Research-intensive institutions

Attending a research-intensive medical school and having a first faculty appointment at a research-intensive institution were independently associated with a lower likelihood of attrition before reaching senior rank. Attending and being employed at research-intensive medical schools may be beneficial to junior faculty given the broad availability of resources, including research mentoring and faculty-development programs, which are intended to promote retention in biomedical-research and academic careers. Many research-intensive medical schools, especially those with MSTP funding, have clear mentoring program objectives and processes to support MD/PhD students in order to achieve educational program success [56,57], and graduates with dual MD/PhD degrees have been observed to be more likely to hold full-time faculty appointments [58]. All medical students, trainees, and faculty at schools with an MSTP-funded MD/PhD training program also likely benefit from the strong institutional support for research at research-intensive medical schools.

Research experiences and grant funding

A recent review about surgeon-scientist trainees’ experiences noted that GME-research years often support trainees’ professional development via education and mentoring in grant writing, applying for funding, structured research experiences, and career mentoring [59]. Experiences that trainees obtain during their research year(s) likely set them ahead of their peers in navigating hurdles of an academic career. Indeed, we observed faculty who had participated in ≥1 year of research during GME were less likely to leave academic medicine. It may be that GME programs and institutions offering research opportunities to trainees can better mentor trainees for successful academic careers more generally.

Given that grant awards and publications are known “traditional criteria” for promotion in academia [60], our findings that faculty who received RPGs and other types of grant funding were less likely to leave academic medicine were not surprising. URiM faculty are reported to have fewer publications [2] and lower scholarly productivity in higher-impact journals [61], receive fewer R01 grants [61–63], and have lower rates of promotion to rank of professor compared to White faculty [2]. Racial, ethnic, and sex differences in research funding may be at least partially responsible for the underrepresentation of certain groups of academic-medicine faculty at senior rank. Having fewer refereed publications also was reported to be associated with women’s lower likelihood of academic-medicine promotion and retention [64]. Institutions might consider research-focused faculty-development programs to help strengthen faculty grant-writing skills [40], foster productive mentor-mentee relationships [40,65], and offer funding opportunities to allow early-career faculty protected time for research. At one institution, modest one-year intramural grants awarded to junior faculty were associated with higher numbers of subsequent external grant awards and peer-reviewed publications among recipients [66], although potential bias introduced by the program’s competitive selection process cannot be discounted.

Faculty department, track, and rank

We also observed differences in faculty attrition at the department level, with faculty first appointed in each of Family Medicine and Psychiatry departments more likely to leave academic medicine before reaching senior rank. Family Medicine and Psychiatry departments might offer less support for faculty to engage in research and scholarly productivity than Medicine and other departments offer. Family Medicine department chairs [67] and faculty [68] have described barriers to research and scholarly productivity, including having insufficient numbers of faculty engaged in research [68], serving as principal investigators, in addition to a lack of research-funding sources [67]. Psychiatry residents with greater financial debt have reported lower levels of interest in research [69]. In addition, a faculty-survey study reported an association between faculty members’ perceived lack of recognition of exemplary clinical service during promotion and tenure reviews and their intent to leave academic medicine [70]. We speculate that greater support for clinical service and teaching, especially in non-research-focused clinical departments at research-intensive institutions could help reduce attrition in such departments.

Faculty first appointed as instructors and at institutions where tenure was not available were also more likely to leave academic medicine before promotion to senior rank. Others have reported greater attrition among faculty appointed to the clinician-educator track, in particular [71]; notably, an earlier study of 83 medical schools reported higher proportions of female than male full-time faculty were on the clinician-educator track [72]. The Faculty Roster does not distinguish between faculty on the clinician-educator track and other non-tenure-eligible tracks, although we did not observe a significant difference in attrition between faculty on the tenure-eligible track and those on the non-tenure-eligible track. Notably, tenure tracks are an expensive venture for academic medicine [73] and are decreasing in number across U.S. medical schools [74].

Strengths and limitations

The study’s strengths include having data for an entire national cohort of U.S. medical students who graduated from LCME-accredited medical schools with long-term follow-up after first full-time faculty appointment as instructor or assistant professor. We also had data for receipt of federal grant awards before attrition or promotion to associate professor and for institutional-level data indicating whether faculty were first appointed at research-intensive medical schools. As an individual’s demographic characteristics cannot be changed, we sought to identify other variables associated with attrition that can be targeted for intervention to increase the size and diversity of physicians at senior ranks in the academic-medicine workforce.

Our study also has limitations. Unmeasured variables, such as number of publications, known to be positively associated with faculty promotion and retention [64], characteristics of the institution where faculty completed GME, and availability of departmental support (e.g., protected time for research, mentoring, and other resources) for trainees and faculty, also might be associated with faculty attrition before reaching senior rank. We could not adjust our regression models for reasons why faculty left academic medicine, including specialty-based differences in salary, call schedules, and other factors that are likely to make academic careers challenging. Differences in compensation by sex or other characteristics may be relevant, as previously reported [75]. Physicians’ reasons for leaving the faculty may differ both within and across departments; however, we lacked data for their reasons. Furthermore, we did not have data about the disparate burden of parenting children, usually placed on mothers, nor about faculty perceptions of work-life balance to measure these factors in association with attrition from the faculty, as suggested by the literature [76–78]. We also acknowledge the potential for collinearity in our regression model, noting that several variables included in our model have been shown in other studies to be associated with each other. For example, female and URiM medical students were more likely to report diminished plans to pursue a career in academic-medicine from matriculation to graduation [43]. Nevertheless, we observed independent effects of individual- and institutional-level variables on faculty attrition prior to reaching senior rank, and the 95% CIs were reasonably small, indicating stable and reliable observations. That said, as one might expect, the power of the Hosmer-Lemeshow goodness-of-fit test increases with very large samples [79]. With over 30,000 cases included in this large national cohort, it was not surprising that the goodness-of-fit test was significant. Nevertheless, a significant Hosmer-Lemeshow test, given such a large sample, does not necessarily mean that our findings and variables we identified as significantly associated with attrition are neither sound nor useful [80]. In this study, we aimed to identify factors on which we might intervene to prevent attrition from academic medicine before attaining senior rank and succeeded in doing so. In addition, we cannot infer causation from our regression model identifying factors associated with faculty attrition. Our results may not be generalizable to faculty at non-LCME-accredited, osteopathic, or international medical schools. Our sample was limited to full-time faculty initially appointed as instructor or assistant professor, which represents most first-time faculty appointments. Thus, we cannot generalize our findings to faculty first appointed to part-time or volunteer positions or at the rank of associate or full professor; these faculty would be expected to have quite different experiences prior to appointment and were therefore excluded from sample selection.