Abstract
National data indicate that 50% of assistant professors leave a School of Medicine (SOM) within eight years of hire. At-risk for attrition in some studies are women, racial/ethnic underrepresented minorities (URM), and clinical faculty. Retention of faculty is not adequately studied in the Southwestern US, where at-risk faculty constitute the majority group. The study hypothesized that at-risk faculty have lower retention rates than those not at-risk. Identification of factors predicting retention of at-risk faculty may help institutions devise novel and targeted retention strategies. Prospective time to event analyses studied assistant professors hired at the University of New Mexico’s (UNM) SOM from 2008–2019. Eight factors, measured at the time of hire, included: rank, race/ethnicity, gender, MD degree, academic track, department type, salary, and fiscal year of hire. Univariate analyses included graphical analysis of Kaplan-Meier analysis and Cox proportional hazard ratios with years to departure measuring the main event to resignation. 844 full-time junior faculty included 50% women, 81% physicians, 42% clinician educators, and 18% racial/ethnic URM. Compared to non-Hispanic Whites, Black faculty (HR = 2.24, 1.25–4.03) and faculty with non-US degree (1.53, 1.19–1.94) had a higher risk of leaving. Faculty in clinician educator (2.01, 1.06–3.82) or visiting research tracks (2.41, 1.20–4.84) both had higher risk of leaving than tenure track faculty. Although URM faculty did not have an overall higher risk of departure, male faculty had higher risk of leaving than women when they are URM or unknown-URM status. In our analysis of junior faculty, we showed that faculty who were Black, had an international education, and in clinician educator or visiting research tracks were at greater risk of leaving, but women and Hispanic faculty had similar retention rates as their respective counterparts at UNM SOM. The differential retention rates among several at-risk subgroups of junior faculty may indicate the need to refocus the existing diversity and faculty development programs at UNM SOM.
Background
Faculty are the cornerstones of academic medicine. Yet, half of new assistant professors leave Schools of Medicine (SOMs) within eight years of hire (Liu & Morrison, 2014) and 57% leave within ten years (Alexander & Lang, 2008). Some studies show women and racial/ethnic underrepresented minorities (URM: Hispanic, African-American, American Indian/Alaska Native, and Hawaiian/Pacific Islander), together described as URM in Science (URM-S), and clinical faculty may be ‘at-risk’ for attrition (Brod et al., 2017; Liu & Morrison, 2014; Lowenstein et al., 2007).
The challenges of retaining faculty are great for SOMs in the Southwestern US, which are medium-sized and serve diverse, underfunded, rural, and underserved communities, with significant proportions of racial/ethnic URM faculty (Rodriguez et al., 2014). Based on the University of New Mexico (UNM) SOM faculty database, attrition (defined by voluntary resignation) rates for all non-volunteer faculty was 8.5% for Fiscal Year 2019. The UNM SOM is an ideal place to study faculty retention since 55.3% of its 1,100 faculty are women and/or racial/ethnic URM (49.6% and 15.7% respectively) at a minority-serving institution in a minority-majority state. Faculty turnover can dampen workplace morale and climate, and exacerbate declining interest in pursuing or sustaining careers in academic medicine (Bickel & Brown, 2005), which may disproportionately impact minority serving institutions. Further, given that the minimum cost of clinical staff turnover represents a loss of >5% of the total annual operating budget at the UNM Hospital (Waldman et al., 2010), and that another SOM estimated that the cost to support a cohort of new faculty hires in basic science over seven years was $69 million (Dorsey et al., 2009), UNM SOM can scarcely afford this loss of intellectual capital. Retention is therefore a critical issue for UNM SOM but also for SOMs regionally and nationally (Alexander & Lang, 2008; Krebsbach & Ignelzi, 1999).
The long-term goal of this study is to determine innovative strategies for promoting SOM faculty retention. The objective of this study is to determine factors influencing the retention of assistant professors, a group that has been previously shown to have the highest attrition rate among SOM faculty (Liu & Morrison, 2014). The study hypothesis was that at-risk assistant professors have lower retention rates than those not at-risk. Identification of factors predicting retention of at-risk faculty may help institutions devise novel retention strategies.
Methods
Study Design
This longitudinal study analyzed the UNM SOM faculty database for all full-time assistant professors hired at UNM SOM from 2008–2019. The SOM faculty database monitors faculty during their entire life cycle at the SOM, providing reliable estimates of termination dates and demographic characteristics. All assistant professors are hired in one of the following academic tracks - tenure track, clinician educator track for those primarily engaged in patient care and teaching, visiting research track for those primarily engaged in research activities, and visiting flex track faculty who may choose to become either clinician educators or tenure-track faculty within three years of hire. Faculty appointments may be probationary, tenured, continuing non-tenure-track, or temporary/non-continuing. Prior to awarding of tenure, tenure-track faculty appointments are probationary appointments; following the award of tenure, such appointments are tenured.
This study included full-time assistant professors who were employed for at least six months from their appointment to their first departure from the institution, regardless of whether individuals had appointments in different departments. Attrition was defined by resignation by any non-volunteer faculty from the SOM. The 25th percentile and median years until attrition occurred (i.e., the number of years elapsed when 25 percent and 50 percent of the faculty from the original cohort left UNM SOM) were examined. Eight factors, measured at the time of hire, included: rank, race/ethnicity, gender, MD degree, academic track, department type, salary, and fiscal year.
Analysis
Years to resignation was the event of interest with censoring at the end of the study (March 3, 2020), death and retirement. Univariate analyses included descriptive frequencies and percentages, Kaplan-Meier analyses, Cox proportional hazards models and Fine-Gray competing risk analyses (Austin & Fine, 2017). Multivariable models were used to investigate joint effects of URM-status, gender and type of appointment. Point estimates and 95% confidence intervals (CI) for years until 25% and 50% attrition were calculated. Effect sizes from time-to-event analyses are hazard ratios (HR) and 95% CI. When HR > 1.0 there is a greater risk of early resignation compared to the referent, and risk of resignation is lower when HR < 1.0. Two approaches evaluated whether results were sensitive to censoring rules and promotion. First, we employed Fine-Gray competing risks models where promotion was a competing risk. The second approach was to also censor at promotion, at a change in academic track, and after seven years of follow up, the typical time for promotion at UNMHSC. We used α = 0.05 for testing of main effects and use α = 0.10 in exploratory analyses for effect modification.
The UNM Human Research Protection Office, which served as the Institutional Review Board of record (HRPO 17–347), approved this study.
Results
UNM SOM hired 844 assistant professors, mostly physicians (80.7%) during 2008–2019 (Figure 1). Most assistant professors were hired in either the visiting flex-track (44.6%) or clinician educator track (41.6%). The majority of the assistant professors in the tenure-track (80.5%) and the visiting research track (96.0%) had PhD or other degrees. The study included 49.8% women, 58.6% non-His-panic Whites, 18.3% racial-ethnic URM, 53.5% URM-S, and 60.9% in medical specialties, defined as internal medicine, pediatrics, emergency medicine, and family and community medicine (Table 1).
Figure 1.
Sample size distribution of assistant professors hired into UNM SOM, 2008 – 2019, by contract type, appointment and academic degree.
Table 1.
Characteristics of Assistant Professors hired at UNM SOM (N = 844), 2008 – 2019.
| N | (%) | |
|---|---|---|
| Gender | ||
| Female | 420 | (49.8) |
| Male | 424 | (50.2) |
| Race/Ethnicity | ||
| American Indian | 6 | (0.7) |
| Asian | 126 | (14.9) |
| Black | 18 | (2.1) |
| Hispanic | 131 | (15.5) |
| Non-Res Alien | 12 | (1.4) |
| Two or More Races | 17 | (2.0) |
| Unknown | 40 | (4.8) |
| White | 494 | (58.6) |
| Under Represented Minorities (URM, Hispanic, Black, American Indian, Hawaiian/Pacific Islander) | ||
| White and Asian | 620 | (73.5) |
| Hispanic, Black, and American Indian | 155 | (18.3) |
| Other / No answer | 69 | (8.2) |
| Under Represented Minorities in Science (URM-S) | ||
| White or Asian Males | 323 | (38.3) |
| All Females and Hispanic, Black, or American Indian Males | 452 | (53.5) |
| Unknown | 69 | (8.2) |
| Annual Contract Salary at Hire | ||
| < $120K | 172 | (20.4) |
| $120K - $159K | 237 | (28.1) |
| $160K - $199K | 195 | (23.1) |
| $200K+ | 240 | (28.4) |
| Contract type | ||
| Continuing | 769 | (91.1) |
| Non-continuing | 75 | (8.9) |
| Appointment type | ||
| Clinician Educator | 351 | (41.6) |
| Tenure track | 41 | (4.9) |
| Flex track | 377 | (44.6) |
| Visiting/Research track | 75 | (8.9) |
| MD degree | ||
| No | 161 | (19.1) |
| Yes | 681 | (80.7) |
| Undetermined | 2 | (0.2) |
| USA degree | ||
| USA | 640 | (75.8) |
| Non-USA | 193 | (22.9) |
| Undetermined | 11 | (1.3) |
| Department Type | ||
| Medical Specialties | 514 | (60.9) |
| Surgical Specialties | 204 | (24.2) |
| Basic Research | 31 | (3.7) |
| Diagnostic Specialties | 84 | (9.9) |
| Others | 11 | (1.3) |
New assistant professor faculty accrued a total of 3,538 person-years of follow up with 335 resignations during the study timeframe. There were two censored faculty times due to death, three retirements and 500 censored at the end of the study. In the competing risks analysis 171 faculty were promoted. Time to 25% attrition was 2.7 years (95% CI 2.4 – 3.0), and time to 50% attrition was 7.4 years (95% CI 6.1 – 8.6, Figure 2).
Figure 2.
Retention function for assistant professors hired at UNMHSC, 2008 – 2019.
Factors associated with differential risk of departure are shown in Table 2. Our unadjusted analysis of faculty show that race/ethnicity, international education, and appointment type are associated with retention (Table 2). Compared to non-Hispanic White counterparts, Black faculty (HR = 2.24, 1.25–4.03) and faculty of international origin (1.53, 1.19–1.94) have a higher risk of leaving. Faculty in clinician educator (2.01, 1.06–3.82) and visiting research tracks (2.41, 1.20–4.84) both have a higher risk of leaving than tenure track faculty.
Table 2.
Factors associated with faculty retention, unadjusted analyses (HR > 1 equal higher risk of leaving and HR < 1 equal lower risk of leaving). Several subgroups had insufficient number of faculty to estimate percentiles (indicated by ‘−’).
| Years until % attrition | Retention | ||||
|---|---|---|---|---|---|
| 25% | 50% | HR | (95% CI) | P-value | |
| All | 2.7 | 7.4 | |||
| Gender | 0.26 | ||||
| Female | 3.0 | 7.4 | 1.00 | ||
| Male | 2.4 | 7.2 | 1.13 | (0.91–1.40) | |
| Race/Ethnicity | 0.01 | ||||
| Asian | 3.2 | 8.0 | 0.89 | (0.65–1.23) | |
| Black | 1.6 | 2.5 | 2.24 | (1.25–4.03) | |
| Hispanic | 3.0 | -- | 0.85 | (0.62–1.16) | |
| Other/Unknown | 1.9 | 4.4 | 1.55 | (1.03–2.32) | |
| Two or More Races | -- | -- | 0.39 | (0.13–1.23) | |
| White | 2.7 | 7.7 | 1.00 | ||
| Under Represented Minorities | 0.36 | ||||
| White and Asian | 2.8 | 7.7 | 1.00 | ||
| Hispanic, Black, and American Indian | 2.7 | 7.4 | 0.97 | (0.73–1.28) | |
| Other/Unknown | 2.1 | 4.6 | 1.31 | (0.89–1.92) | |
| Under Represented Minorities in Science | 0.37 | ||||
| White or Asian Males | 2.7 | 8.0 | 1.00 | ||
| All Females and Hispanic, Black, or American Indian Males | 2.7 | 7.3 | 1.01 | (0.80–1.27) | |
| Other/Unknown | 2.1 | 4.6 | 1.32 | (0.88–1.98) | |
| Annual Contract Salary at Hire | 0.77 | ||||
| < $120K | 3.6 | 8.0 | 1.00 | ||
| $120K - $159K | 2.4 | 6.5 | 1.16 | (0.85–1.57) | |
| $160K - $199K | 3.0 | 7.3 | 1.12 | (0.81–1.55) | |
| $200K+ | 2.7 | 7.0 | 1.04 | (0.76–1.44) | |
| Contract type | 0.10 | ||||
| Continuing | 2.7 | 7.4 | 1.00 | ||
| Non-continuing | 2.4 | 5.2 | 1.33 | (0.95–1.87) | |
| Appointment type | 0.05 | ||||
| Tenure track | 6.1 | 8.8 | 1.00 | ||
| Clinician Educator | 2.4 | 7.0 | 2.01 | (1.06–3.82) | |
| Flex track | 3.1 | 7.4 | 1.75 | (0.92–3.32) | |
| Visiting/Research track | 2.4 | 5.2 | 2.41 | (1.20–4.84) | |
| Has an MD degree | 0.42 | ||||
| No | 3.7 | 8.6 | 1.00 | ||
| Yes | 2.7 | 7.2 | 1.12 | (0.85–1.48) | |
| USA degree | .01 | ||||
| USA | 3.0 | 8.6 | 1.00 | ||
| Non-USA | 2.3 | 4.9 | 1.53 | (1.19–1.94) | |
| Department Type | 0.66 | ||||
| Medical Specialties | 2.7 | 6.9 | 1.00 | ||
| Surgical Specialties | 2.9 | 7.7 | 0.93 | (0.72–1.21) | |
| Basic Research | 6.1 | 8.8 | 0.66 | (0.34–1.28) | |
| Diagnostic Specialties | 2.6 | 5.0 | 1.13 | (0.79–1.61) | |
| Others | 3.0 | 7.0 | 0.92 | (0.34–2.46) | |
Although URM faculty did not have an overall higher risk of departure, the association may not be consistent across gender categories (P = 0.09 URM x Gender effect modification, Table 3). Unlike non-URM, male faculty have higher risk of leaving than female faculty when they are URM (HR = 1.80, 1.01–3.22) or unknown-URM (HR = 1.72, 1.04–2.84) but not when they are non-URM (HR = 0.98, 0.76–1.25). Adjusting for appointment type (P = 0.09 URM x Gender effect modification) and for competing risks (P = 0.11 URM x Gender effect modification, Table 3) did not substantially change risk of departure. Using modified censoring and follow up rules also did not have substantial effects on point estimates for conclusions.
Table 3.
Risk of departure stratified by gender and URM status estimated by models with URM x Gender effect modification: unadjusted, adjusted for appointment type (HRAdj) and adjusted for appointment type and competing risks (HRAdj-CR).
| Factor tested | Subgroup | HR (95% CI) | HRAdj (95% CI) | HRAdj-CR (95% CI) |
|---|---|---|---|---|
| Male vs. female (referent) | Non-URM | 0.98 (0.76–1.25) | 1.00 (0.78–1.29) | 1.00 (0.77–1.29) |
| URM | 1.72 (1.04–2.84) | 1.75 (1.05–2.90) | 1.64 (0.96–2.79) | |
| Unknown-URM | 1.61 (0.76–3.41) | 1.69 (0.80–3.58) | 1.85 (0.89–3.85) | |
| URM vs non-URM (referent) | Female | 0.75 (0.51 – 1.13) | 0.76 (0.51 – 1.14) | 0.78 (0.52 – 1.18) |
| Male | 1.33 (0.85 – 2.07) | 1.33 (0.89 – 1.97) | 1.29 (0.84 – 1.97) | |
| Unknown URM vs non-URM (referent) | Female | 0.99 (0.53 – 1.83) | 1.01 (0.54 – 1.88) | 0.95 (0.54 – 1.68) |
| Male | 1.63 (0.99 – 2.67) | 1.70 (1.04 – 2.79) | 1.76 (1.03 – 3.00) |
Discussion
We found that 50% of our junior faculty departed with 7.4 years (95% CI 6.1 – 8.6), which is consistent with eight years found previously in a national study (Liu & Morrison, 2014). In our analyses of junior faculty, we showed that assistant professors who are Black or who have had an international education, and in clinician educator or visiting research tracks are at greater risk of leaving, but women and Hispanic faculty have similar retention rates, than their respective counterparts at UNM SOM. Although URM faculty do not have an overall higher risk of departure, male faculty have higher risk of leaving than women when they are URM. The reasons why faculty leave UNM SOM are multi-factorial, the top five reasons include personal/family matters, work environment, leadership, better career opportunity, and salary (Sood et al., 2018). We have also previously demonstrated that male and racial/ethnic URM faculty leave UNM SOM for the same set of reasons as female and non-URM faculty respectively (Sood et al., 2018).
Previous studies related to attrition have focused on assessing long-term retention rates by taking snapshots of faculty members after five or 10 years have elapsed (Alexander & Lang, 2008; Corrice et al., 2011). These snapshots facilitate institutional benchmarking against national norms because the method is straightforward and easy to replicate. However, such snapshots are not ideal for comparing change in retention rates over time or across distinct groups. Instead, labor market analysts interested in learning about change in retention over time, or differences across groups, employ life-table and time-to-event analysis methods for studying departure (or attrition) (Blossfeld & Rohwer, 2002). Such methods facilitate a more complete picture of longitudinal outcomes and allow mediating factors to be controlled. As such, this study presents faculty attrition statistics and operationalizes retention rates in terms of time from initial appointment to time of departure in order to draw inferences about factors associated with retention. Although the study accounts for potential competing risks on attrition, a time-varying approach would be a natural next step towards understanding attrition throughout a faculty member’s career, especially as they choose to stay or depart.
Although previous studies show that women faculty had lower retention rates than men (Alexander & Lang, 2008; Yamagata, 2002), a longitudinal national study published in 2014 showed that women left their institutions at similar rates as men at the assistant professor level (Liu & Morrison, 2014). This study confirms the 2014 findings by Liu and Morrison that women assistant professors have currently closed the retention gap with their male counterparts. However, we continue to demonstrate that certain non-White groups leave more quickly than White assistant professors (Liu & Morrison, 2014). Their attrition, particularly of male racial/ethnic URM assistant professors, points towards challenges for recruitment, promotion, and mentoring programs for diverse junior faculty at SOMs.
Research suggests that faculty in clinical roles are less likely to be at higher academic ranks, less likely to be satisfied with their progress towards academic promotion, and more likely to leave or express intent to leave academic medicine (Buckley et al., 2000; Lowenstein et al., 2007; Thomas et al., 2004). Clinician educators may face challenges in juggling multiple responsibilities, navigating competing demands, and fulfilling expectations for promotion and advancement – pressures that could contribute to their intent to leave (Shanafelt et al., 2009). They may be most prone to feeling that their work is not valued, recognized, or rewarded by the institution (Liu & Morrison, 2014). In addition, with tightening patient care budgets and the resulting pressure to increase clinical workload, clinically-focused faculty may find themselves lacking time for scholarly activities, which can disadvantage their career advancement and satisfaction (Lowenstein et al., 2007). On the other hand, faculty on visiting research tracks are more likely to have part-time appointments or non-continuing contracts, factors that may increase their attrition. Studies have not previously evaluated the attrition risk for faculty with international education, an at-risk group identified by our study. The risk for the members of this group may be related to their identity as racial/ethnic minorities, religious minorities, weak immigration status, and inadequate network ties or enculturation in the community.
The academic medicine community has maintained high interest in retaining faculty to effectively achieve their clinical, education, and research missions. These analyses showed that assistant professors who were Black, of international education, and in clinician educator or visiting research tracks had greater risk of leaving, but women and Hispanic faculty had similar retention rates, than their respective counterparts at UNM SOM. The differential retention rates among several at-risk subgroups of junior faculty may indicate the need to refocus the existing diversity and faculty development programs at UNM SOM.
References
- Alexander H, & Lang F (2008). The long-term retention and attrition of US medical school faculty. AAMC Analysis in Brief, 8, 1–2. [Google Scholar]
- Austin PC, & Fine JP (2017). Practical recommendations for reporting Fine-Gray model analyses for competing risk data. Stat Med, 36(27), 4391–4400. 10.1002/sim.7501 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bickel J, & Brown AJ (2005). Generation X: Implications for faculty recruitment and development in academic health centers. Acad Med, 80(3), 205–210. 10.1097/00001888-200503000-00003 [DOI] [PubMed] [Google Scholar]
- Blossfeld HP, & Rohwer G (2002). Techniques of event history modeling. New approaches to causal analysis (2nd ed.). Lawrence Erlbaum Associates. [Google Scholar]
- Brod HC, Lemeshow S, & Binkley PF (2017). Determinants of faculty departure in an academic medical center: A time to event analysis. Am J Med, 130(4), 488–493. 10.1016/j.amjmed.2016.12.011 [DOI] [PubMed] [Google Scholar]
- Buckley LM, Sanders K, Shih M, & Hampton CL (2000). Attitudes of clinical faculty about career progress, career success and recognition, and commitment to academic medicine. Results of a survey. Arch Intern Med, 160(17), 2625–2629. 10.1001/archinte.160.17.2625 [DOI] [PubMed] [Google Scholar]
- Corrice AM, Fox S, & Bunton SA (2011). Retention of full-time clinical M.D. faculty at U.S. medical schools. Analysis in Brief, AAMC, 11(2), 1–2. [Google Scholar]
- Dorsey ER, Van Wuyckhuyse BC, Beck CA, Passalacqua WP, & Guzick DS (2009). The economics of new faculty hires in basic science. Acad Med, 84(1), 26–31. 10.1097/ACM.0b013e3181904633 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Liu CQ, & Morrison E (2014). U.S. medical school full-time faculty attrition. Analysis in Brief, AAMC, 14(2). [Google Scholar]
- Lowenstein SR, Fernandez G, & Crane LA (2007). Medical school faculty discontent: prevalence and predictors of intent to leave academic careers. BMC Med Educ, 7, 37. 10.1186/1472-6920-7-37 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Rodriguez JE, Campbell KM, Fogarty JP, & Williams RL (2014). Underrepresented minority faculty in academic medicine: A systematic review of URM faculty development [Review]. Fam Med, 46(2), 100–104. http://www.ncbi.nlm.nih.gov/pubmed/24573516 [PubMed] [Google Scholar]
- Shanafelt TD, West CP, Sloan JA, Novotny PJ, Poland GA, Menaker R, Rummans TA, & Dyrbye LN (2009). Career fit and burnout among academic faculty. Arch Intern Med, 169(10), 990–995. 10.1001/archinternmed.2009.70 [DOI] [PubMed] [Google Scholar]
- Sood A, Parker T, Nieto N, Maloney J, Greenberg N, Rishel Brakey H, Myers O, Sigl D, & Wilson B (2018). Medicine faculty consider mentoring as the critical issue for development and retention (2018). 11th Annual Mentoring Conference Proceedings (11th ed.). The Chronicle of Mentoring & Coaching, 2(Special Issue 1), 377–383. [Google Scholar]
- Thomas PA, Diener-West M, Canto MI, Martin DR, Post WS, & Streiff MB (2004). Results of an academic promotion and career path survey of faculty at the Johns Hopkins University School of Medicine. Acad Med, 79(3), 258–264. 10.1097/00001888-200403000-00013 [DOI] [PubMed] [Google Scholar]
- Waldman JD, Kelly F, Arora S, & Smith HL (2010). The shocking cost of turnover in health care. Health Care Manage Rev, 35(3), 206–211. 10.1097/HMR.0b013e3181e3940e [DOI] [PubMed] [Google Scholar]
- Yamagata H (2002). Trends in faculty attrition at U.S. medical schools, 1980–1999. Analysis in Brief. AAMC, 2(2), 1–2. [Google Scholar]