Faculty Retention at a School of Medicine, 2010–2022

Abstract

Faculty retention at academic health centers is a concern with about one-fifth of physicians reporting intentions to leave. We studied factors affecting faculty at risk for attrition, defined as women, racial/ethnic underrepresented minorities (URM), and clinical faculty. Identification of factors predicting retention of at-risk faculty may help mentors and minority-serving institutions devise novel targeted retention strategies. Our study site was a minority-serving institution in a majority-minority state in the US Southwest where at-risk faculty constitute the majority group. Faculty characteristics and departure dates were extracted from an institutional database maintained by the University of New Mexico (UNM) School of Medicine (SOM) for 2,427 participants employed from July 2009 through June 2022. Annual attrition rates and relative risk (RR) of attrition were estimated by discrete-time hazard rate models assuming a Poisson distribution. The overall annual attrition rate was 11.5%, which projects to 50% attrition in 6.0 years. Time to 50% attrition was 4.6 years for assistant professors, 8.9 years for associate professors 7.2 years for full professors. Faculty with a PhD degree had lower attrition (7.2%, RR=0.69, 95% CI 0.60, 0.79) compared to faculty with an MD degree (10.5%) in adjusted analyses. Clinician educators had a higher attrition rate (8.9%) compared to tenure track (6.4%, RR_{tenure track}=0.72, 95% CI 0.61, 0.85). Black faculty had a higher risk of attrition compared to White faculty (RR=1.56, 95% CI 1.09, 2.25), and non-Hispanic White faculty had a lower risk of attrition (RR=0.83, 95% CI 0.71, 0.98). Annual attrition rates increased over the study period with most of the increase before about 2016. We did not detect significant differences in attrition due to sex or URM status.

Introduction & Literature Review

Retention of School of Medicine (SOM) faculty is of concern with about one-fifth of physicians reporting an intention to leave, with average replacement costs in 2004 for Department of Medicine and Surgery faculty being about $400,000 each (Schloss et al., 2009, Abbasi 2022). The financial cost of replacement is becoming increasingly unsustainable due to the economic environment that academic health centers find themselves in during the post-pandemic era. Not only is attrition costly in dollars but it also causes disruptions to patient care and educational programs and to the work of high-functioning groups. Preventable losses of valued colleagues take a big toll on morale and can increase disengagement and cynicism (Bickel, 2012).

At risk for attrition are groups such as women, racial/ethnic underrepresented minorities (URM), part-time faculty, faculty of international origin, and clinical faculty (Greenberg et al., 2021; Wapman et al., 2022). Factors affecting faculty retention include onboarding, orientation and mentorship processes, recognition and rewards, timely promotions and granting of tenure status, adequate professional development opportunities, protected time for scholarship, faculty communication and engagement, compensation, work-life balance, and the culture and climate of the organization, particularly for URM faculty (DeWitty & Murray, 2020; Skinner et al., 2023; Sood et al., 2018; Zehra et al., 2021).

Faculty retention has been studied using time-to-event type analyses to learn that half of new assistant professors in the United States leave SOMs within seven to eight years of hire (Liu & Morrison, 2014; Myers et al., 2020) and that 57% leave within ten years (Alexander & Lang, 2008). Retention of faculty is, however, not adequately studied in minority-serving institutions, where at-risk faculty constitute the majority group. Improving the retention of URM faculty is a priority at most minority-serving academic institutions to ensure excellence in patient care, research, and health equity. We hypothesize that at-risk faculty have higher attrition rates than those not at-risk at a minority-serving institution in a minority-majority Southwestern state. Identification of factors predicting retention of at-risk faculty may help mentors and minority-serving institutions devise novel targeted retention strategies We take a discrete-time hazard rate modeling approach to make use of time-varying faculty characteristics and to estimate annual attrition rates that are informative to leaders evaluating trends in faculty retention.

Methods

University of New Mexico (UNM) SOM is a minority-serving institution in a majority-minority state in the US Southwest with a robust diversity program. UNM SOM faculty employed during July 2009 through June 2022 were studied to test whether gender, race/ethnicity, rank, academic track, MD degree, department type, and salary were related to separation from the SOM. Faculty demographics and employment status including rank, track, department type, and dates of promotion and termination were extracted from an institutional database maintained by SOM. A longitudinal person-year analytic file was recreated for each faculty member during their employment. In this study, years are fiscal years (FY) beginning July 1 and ending June 30 with FY numbered using the June year. Baseline variables included sex, race/ethnicity URM status, and type of degree. Time-varying variables included department type, rank, years in rank, and appointment type. Faculty are hired in one of the following academic tracks determined by appointment type: tenure track, clinician-educator track for those primarily engaged in patient care and teaching, visiting research track for those primarily engaged in research activities in a non-tenure status, 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 the awarding of tenure, tenure-track faculty appointments are probationary appointments; following the award of tenure, such appointments are tenured. Departments were grouped into medical specialties, surgical specialties, basic research, diagnostic specialties, and others.

Resignation and retirement status were encoded for each faculty and FY with an event indicator variable coded 1 if the faculty resigned/retired during that FY and coded 0 if they did not resign/retire. This coding enabled us to obtain annual attrition rates for each FY as the number resign/retire by summing the event indicator and dividing by the number of faculty at risk during the FY. The number of at-risk faculty is the total number of faculty employed on the first date of each FY. Faculty contracts typically begin on July 1^st and are for one year. With a retirement/resignation, the data record ends for that faculty member unless they return later in a different role, such as a working retiree, which has a unique appointment type code. Therefore, a faculty member could have a second employment period that is captured in the data and modeled using time-varying appointment type and other variables. All SOM faculty were included in analyses to estimate attrition except for models to assess factors affecting attrition of assistant professors based on time in rank.

Retention and attrition rates were estimated using discrete-time hazard rate models assuming a Poisson distribution. Models estimate the hazard probability of leaving, h, the one-year probability of attrition in the FY among those at-risk in the FY, as a function of covariates in the model. When covarites are binary indicator variables one can take the exponential of the estimated coefficient to obtain an estimate of relative risk (RR) of attrition. In addition to estimating RR, we used the h and assumed an exponential attrition model to estimate the time to to 50% attrition using the equation.

Model covariates included baseline demographics and time-varying covariates. We estimated hazard probabilities using Poisson regression instead of logistic regression, so the exponentiated regression coefficients have a relative risk interpretation. Participant factors, such as gender, were used in factor ´ FY interaction terms added to models to test whether risk varied over time. Restricted cubic splines were used to test whether temporal trends were nonlinear. Factors affecting the retention of assistant professors were tested by restricting analyses to faculty beginning their employment as assistant professors.

Attrition rates by year in rank were combined for year seven and higher as many faculty had either been promoted or left by this time, causing rate estimates to be unstable. Estimates for relative risk of attrition and 95% confidence intervals were computed. Time-varying estimates of attrition are displayed graphically. A two-sided pvalue <0.05 is considered statistically significant. These statistical tests were performed using SAS Base software version 9.4 (SAS, Cary, NC, USA). The institutional review board permitted a consent waiver (University of New Mexico Human Research Protection Office 17–347).

Results

A total of 2,427 contract faculty were studied that had 13,816 person-years of follow up (Table 1). When they began SOM employment, the faculty distribution was 50% female, 15% racial/ethnic URM, 70% assistant professors, 14% associate professors, 15% professors (Table 1). Clinician educators made up 40%, tenure track or tenured were 14% and medical specialties were 63%. As a fraction of person-years of follow up, assistant professors were 43% and clinician educators were 53% of the sample. Males and females had about 11.5% annual attrition, the overall average (Table 1).

Table 1.

Faculty Characteristics and Attrition Rates

	Baseline			Attrition Rate (FY10-FY22)
	N	(%)	Person-years	Annual % Attrition	RR (95% CI)	p-value	Time x group p-value
Overall	2427	100	13,816	11.5
Sex
Female	1211	(49.9)	6634	11.4	0.98 (0.88, 1.08)	0.689	0.933
Male	1216	(50.1)	7182	11.6	1.00
URM						<.001	0.042
Not URM	1875	(77.3)	11,042	11.2	1.00
URM	373	(15.4)	2036	10.2	0.91 (0.79, 1.05)
Two or more races	43	(1.8)	187	13.9	1.24 (0.84, 1.83)
Unknown	136	(5.6)	551	21.8	1.94 (1.61, 2.34)
Rank						<.001	0.108
Assistant	1710	(70.5)	5936	15.2	1.00
Associate	350	(14.4)	3594	7.8	0.52 (0.45, 0.59)
Full/Senior	367	(15.1)	4286	9.6	0.63 (0.56, 0.71)
MD degree	1701	(70.1)	10,033	11.0	0.84 (0.76, 0.93)	0.001	0.228
Non-MD	726	(29.9)	3783	13.1	1.00
Appointment type						<.001	0.411
Clinician educator	966	(39.8)	7294	8.9	1.00
Tenure Track/Tenured	338	(13.9)	2876	6.4	0.72 (0.61, 0.85)
Visiting-Flex	475	(19.6)	1051	9.3	1.05 (0.85, 1.30)
Visiting-Research	159	(6.6)	757	13.1	1.47 (1.19, 1.82)
Visiting-Working retiree	58	(2.4)	860	25.7	2.89 (2.48, 3.36)
Visiting-Other	308	(12.7)	378	71.7	8.06 (6.99, 9.28)
Non-probationary	123	(5.1)	595	11.4	1.28 (1.00, 1.65)
Department type
Medical specialties	1519	(62.6)	8474	11.7	1.00	0.033	0.907
Surgical specialties	487	(20.1)	2810	10.9	0.93 (0.82, 1.06)
Basic research	109	(4.5)	752	8.9	0.76 (0.59, 0.97)
Diagnostic specialties	252	(10.4)	1461	12.3	1.05 (0.90, 1.23)
Others	60	(2.5)	319	15.4	1.31 (0.98, 1.75)

There was significant variation among URM categories with 22% unknown leaving compared to 14% for two or more races, 10% for URM and 11% for remaining non-URM. When URM and non-URM characteristics were modeled as race/ethnicity categories we found that compared to White faculty (11.4% attrition), a significantly higher risk of attrition was found for Black faculty (17.9%, RR=1.56, 95% CI 1.09, 2.25, p=.016), and a lower risk of attrition was found for non-Hispanic White faculty (9.5%, RR=0.83, 95% CI 0.71, 0.98, p=.024).

Assistant professors had 15% attrition, which was significantly higher than for associate and full professors (p<.001). In this unadjusted analysis, faculty with an MD degree had significantly lower annual attrition (11%) compared to others, the latter included visiting research track non-MDs (13% for others, RR=0.84 95% CI 0.76, 0.93). After adjusting for visiting faculty track, faculty with a PhD degree had lower attrition (7%, RR=0.69, 95% CI 0.60, 0.79) compared to faculty with an MD degree. Clinician educators had a higher attrition rate (9%) compared to tenure track (6%, RR=0.72, 95% CI 0.61, 0.85), and lower attrition compared to visiting-research (13%) and working retirees (26%). Medical specialties (12%) had a higher risk of leaving relative to basic research (9%, RR_{basic research}=0.76, 95% CI 0.59, 0.97).

Crude attrition rates varied over time (p<.001), starting at 7.5% to 9.2% in FY10-FY12 and ending at 13.0%, 11.0% and 13.3% in FY20, FY21 and FY22, respectively (Figure 1). The increasing trend was nonlinear (p<.001) with more rapid increases before about FY16 and less increase after that. The highest attrition rate was 14.4% in FY19, which was prior to the COVID19 pandemic. Temporal variation for groups formed by sex, URM and rank was relatively consistent (p>0.15 for all). Non-clinician educators had a greater increase in attrition rate over time compared to clinician-educators (p=0.02), and non-tenure track faculty had a greater increase in attrition rate over time compared to tenure track faculty (p=0.04).

Figure 1. Faculty attrition rate (%, unadjusted) by year, sex, URM, clinician educator and tenure track status.

We tested whether attrition in the first years of employment for assistant professors was associated with faculty characteristics (Figure 2). Attrition did not vary significantly over time for sex, URM, or degree status (p>.10 for all).

Discussion

In our study, we found an overall attrition rate of 11.5%, which projects to 50% attrition in 6.0 years. Time to 50% attrition was 4.6 years for assistant professors, 8.9 years for associate professors 7.2 years for full professors. Retention was not associated with sex or with URM status, except when race/ethnicity information did not allow determination of URM. We noted that race/ethnicity status may be missing more often in later years, which may signal a need to redesign the collection of this information. Our finding that males and females had similar attrition rates was not consistent with studies that found lower retention rates in women than in men (Brod et al. 2017, Alexander & Lang, 2008; Yamagata, 2002), however, a nationwide, longitudinal study of retention during 1970 to 2000 did not detect differential retention by sex (Liu & Morrison, 2014). The non-differential retention rates among sex and URM-related at-risk groups of faculty may reflect the strong diversity programs at UNM SOM, which serves a majority-minority state. However Black faculty had a higher risk of attrition than White faculty and a projected 3.9 years to 50% attrition.

Junior faculty (largely assistant professors) have greater attrition than senior faculty. The reasons are multifactorial and may include refusal of tenure, inadequate career advancement and mentoring opportunities, and heavy workloads with poor work-life balance. To address the attrition of junior faculty, institutions can implement retention strategies specifically tailored to their needs, such as mentoring programs, professional development support, transparent promotion and tenure policies, adherence to effort split, protected time, childcare, and work-life balance initiatives.

Faculty with MD degrees versus PhD degrees have differing work requirements and responsibilities, market salaries, academic expectations, and career trajectories. The attrition comparison between faculty with MD degrees and PhD degrees is complex, with a need for different and targeted retention strategies for each group of faculty. After adjusting for visiting faculty track, faculty with a PhD have lower overall attrition compared to faculty with an MD degree. Yet, non-clinician educators have experienced a greater increase in attrition rates over time compared to clinician-educators, indicating that these time trends need to be addressed.

The strengths of the study include its setting in a minority-serving institution in a minority-majority state, and a relatively large sample size with longitudinal records from an institutional database from July 2010 through June 2022. The margin of error for the overall attrition rate was ±2%, but subgroup analyses were limited by small sample sizes. A significant proportion of faculty were women and non-White faculty participants. The weaknesses of the study include the single study site limiting the generalizability of study findings.

Conclusion

Junior faculty (largely assistant professors) have greater attrition than senior faculty. Clinician educators had a higher risk of leaving compared to tenure track/tenured faculty. After adjusting for visiting faculty track, faculty with a PhD have lower overall attrition compared to faculty with an MD degree. Women and racial/ethnic URM faculty departure rates were not, however, significantly different from men and non-URM faculty. Recognizing the unique needs and challenges of at-risk faculty groups, institutions can provide specialized mentorship programs, career guidance, and professional development opportunities.

Figure 2.

Assistant Faculty Attrition by Year in Rank and Gender, URM and Degree