Marginalized identities, mistreatment, discrimination, and burnout among US medical students: cross sectional survey and retrospective cohort study.

OBJECTIVE: To describe the association between mistreatment, burnout, and having multiple marginalized identities during undergraduate medical education.
DESIGN: Cross sectional survey and retrospective cohort study.
SETTING: 140 US medical schools accredited by the Association of American Medical Colleges. PARTICIPANTS: 30 651 graduating medical students in 2016 and 2017. MAIN OUTCOME MEASURES: Self-reported sex, race or ethnicity, and sexual orientation groups were considered, based on the unique combinations of historically marginalized identities held by students. Multivariable linear regression was used to determine the association between unique identity groups and burnout along two dimensions (exhaustion and disengagement) as measured by the Oldenburg Burnout Inventory for Medical Students while accounting for mistreatment and discrimination.
RESULTS: Students with three marginalized identities (female; non-white; lesbian, gay, or bisexual (LGB)) had the largest proportion reporting recurrent experiences of multiple types of mistreatment (88/299, P<0.001) and discrimination (92/299, P<0.001). Students with a higher number of marginalized identities also had higher average scores for exhaustion. Female, non-white, and LGB students had the largest difference in average exhaustion score compared with male, white, and heterosexual students (adjusted mean difference 1.96, 95% confidence interval 1.47 to 2.44). Mistreatment and discrimination mediated exhaustion scores for all identity groups but did not fully explain the association between unique identity group and burnout. Non-white and LGB students had higher average disengagement scores than their white and heterosexual counterparts (0.28, 0.19 to 0.37; and 0.73, 0.52 to 0.94; respectively). Female students, in contrast, had lower average disengagement scores irrespective of the other identities they held. After adjusting for mistreatment and discrimination among female students, the effect among female students became larger, indicating a negative confounding association.
CONCLUSION: In this study population of US medical students, those with multiple marginalized identities reported more mistreatment and discrimination during medical school, which appeared to be associated with burnout. © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ.

Introduction

Mistreatment is a common experience during medical school and has been associated with burnout.1 2 3 Often characterized by cynicism, emotional exhaustion, and a low sense of accomplishment, burnout is related to self-reported unprofessional conduct, reduced help seeking behaviors, and medical errors.4 5 6 7 Prior research has reported disparate experiences of mistreatment and discrimination among students with identities historically marginalized in medicine, placing additional stress on female; non-white; and lesbian, gay, or bisexual (LGB) students.8 9 10 11 Social stress theory as a framework explains that the prejudices (eg, sexism, racism, homophobia) present in the social environment act as sources of stress that might have considerable, negative impact to the mental and physical wellbeing of members of marginalized groups.12 Prior research has also found that microaggressions have been associated with positive depression screening among US medical students.13 Among physicians, workplace discrimination based on race or ethnicity has been associated with lower career satisfaction and increased physician turnover.14 Although studies have begun to explore the relations between various social positions, mistreatment, and burnout, research has yet to take into consideration more than one aspect of an individual’s identity using national data.

This study seeks to review this critical knowledge gap and advance prior methodology by applying an intersectional approach to examine the association between student identity across multiple dimensions of marginalization and the experience of mistreatment and discrimination in the medical school learning environment. With roots in black feminist scholarship, intersectionality provides a framework for understanding how an individual’s social identities combine (or intersect) and reflect larger systems of oppression.15 16 The concept of intersectionality highlights that intersecting patterns of racism, sexism, homophobia, and other prejudices in the social environment offer insight into the experiences of marginalized individuals that are not appreciated when examining their social positions in isolation (eg, race or ethnicity, sex, sexual orientation).16 Such an approach captures the linkages of historical, cultural, and structural factors that shape identities and impact life experiences.

This study investigated the association between a student having multiple marginalized identities and burnout score. Additionally, this study explored how experiences of mistreatment and discrimination influenced burnout scores among medical students with multiple marginalized identities. By exploring mistreatment and discrimination across the intersection of sex, race or ethnicity, and sexual orientation, this study aims to obtain a better understanding of how the convergence of these varying identities contributes to differences in the learning environment. The results of this study could have policy implications to foster a more equitable experience during undergraduate medical education for marginalized students.

Methods

Data source and study sample

This retrospective cohort study used data from the 2016 and 2017 Medical School Graduation Questionnaire. The questionnaire is a national survey administered annually by the Association of American Medical Colleges (AAMC) to graduating medical students across the 140 US medical schools that are accredited by the Liaison Committee on Medical Education. The graduation questionnaire is a critical tool used for program evaluation and improvement of the medical student experience. In addition to questions about general medical education and student services among other areas, the questionnaire contains questions that capture student reports of mistreatment, discrimination, and burnout during their medical education.17 The AAMC linked student responses to the graduation questionnaire to self-reported sex and race or ethnicity from AAMC data applications and services.

Demographics

Sociodemographic variables in the study included self-reported sex (male or female), sexual orientation, and race or ethnicity. Responses to sexual orientation were dichotomized as heterosexual or lesbian, gay, or bisexual (LGB). Data on gender identity were not made available to the study team owing to concerns over the privacy of gender minority students.

Race or ethnicity was also dichotomized as white (non-Hispanic) or non-white. The non-white categories included Asian, black, Hispanic, American Indian, Alaskan Native, Native Hawaiian, Pacific Islander, multiracial (students reporting more than one race or ethnicity), and other. While Asian students were not considered in the AAMC’s historic definition of being under-represented in medicine, prior studies have shown that Asian students experience considerably more mistreatment and discrimination than white students.11 18 A sensitivity analysis was conducted to understand the effects of including Asian respondents in the non-white category by repeating all analyses after excluding students who indicated their race as Asian, and no notable differences were observed.

To examine the simultaneous effect of having multiple marginalized identities, we conducted an intersectional analysis with the eight unique identity combinations that were possible from our three sociodemographic variables of focus (eg, sex, race or ethnicity, and sexual orientation). These unique identity groups provide a more granular and nuanced picture of the mistreatment, discrimination, and symptoms of burnout reported by students of varying identities.

Mistreatment and discrimination measurements

To assess mistreatment and discrimination, the Graduation Questionnaire includes questions designed to measure student experiences of various types of negative behaviors by institutional employees, faculty, supervising residents, and other students (table S1). Of these items, seven asked about experiences of general mistreatment and nine asked about discrimination, specifically negative behaviors targeting one’s gender, race or ethnicity, and sexual orientation. The frequency of experiencing each negative behavior was assessed by a four point scale: “never,” “once,” “occasionally,” and “frequently.” For the analyses, “occasionally” and “frequently” were grouped together over the concern that respondents might have inconsistently distinguished between the two response levels and the small sample sizes of those who reported frequently experiencing the different negative behaviors.

Two summary variables were created to reflect varying degrees (that is, frequencies) of mistreatment (based on seven items) and discrimination (based on nine items). The two summary variables each had four levels: no reported experiences of mistreatment or discrimination; isolated experiences of mistreatment or discrimination (one or more types of negative behaviors experienced once); recurrent experiences of one type of mistreatment or discrimination (one type of negative behavior experienced more than once); or recurrent experiences of multiple types of mistreatment or discrimination (more than one type of negative behavior experienced more than once).

Burnout measurements

Burnout was assessed by the 16 item Oldenburg Burnout Inventory for Medical Students, which is a shortened version of the validated and reliable Oldenburg Burnout Inventory.19 20 The Oldenburg Burnout Inventory for Medical Students includes subscales for exhaustion and disengagement dimensions of burnout. Exhaustion refers to the cognitive and physical fatigue that result from the prolonged demands of medical school; disengagement describes the distancing from and negative attitudes toward the objects or contents of medical school.21 22 Each dimension is assessed through eight questions on a scale from 0 to 3, and responses were transformed so that higher scores corresponded with increased burnout. Thus, each dimension has a score ranging from 0 to 24.

Statistical analyses

We used simple descriptive statistics to summarize the characteristics of the sample. Then, χ2 tests were used to examine the distribution of mistreatment and discrimination by sociodemographic characteristics. We then used linear regression to measure the association between students having multiple marginalized identities and their score for exhaustion and disengagement. To determine the extent to which experiences of mistreatment and discrimination mediated or explained the effects of having marginalized identities on burnout, we first examined the main effect by including each of the three individual identity variables (that is, sex, race or ethnicity, sexual orientation) as well as the intersectional variable that included the eight unique identity groups as predictors for the burnout outcome. To test whether experience of mistreatment and discrimination mediated the observed main effect, we sequentially adjusted for these variables in the models. Individual school identifiers were not made available over school privacy concerns, so we could not perform analyses clustered by school; however, all models included adjustment for average burnout score by school, which were rounded to the nearest whole number by the AAMC. Separate analyses were performed for the exhaustion and disengagement dimensions of burnout.

We performed multiple imputation using a fully conditional specification (FCS) method for all missing data assuming a joint distribution for all variables used for the imputation.23 24 Specifically, missing values of categorical variables were imputed using an FCS discriminant function, and those of continuous variables were imputed using an FCS regression method. We included all sociodemographic variables, mistreatment variables, and burnout outcome variables in the imputation model, and created 20 imputed datasets using SAS PROC MI. Analyses were conducted for each imputed dataset. We used PROC MIANALYZE to summarize the results across all 20 datasets. Since the results of imputed data were consistent with those from observed data, only results from multiple imputation were presented. We used P<0.05 to determine statistical significance for all tests. All analyses were performed using SAS 9.4 (SAS Institute, NC). Findings from this study were reported according to STROBE (strengthening the reporting of observational studies in epidemiology) guidelines.25

Patient and public involvement

Owing to the deidentified nature of the study’s dataset, no members of the public were directly involved in the design or analysis of the reported data.

Results

Respondent characteristics

About half of the respondents (48.5%) reported their sex as female, 10 181 (39.2%) students identified as non-white, and 1419 (5.5%) reported their sexual orientation as LGB (table 1). Male, white, and heterosexual students comprised the largest unique identity group (31.0%). Female, non-white, and LGB students (that is, students with all three marginalized identities) comprised 0.8% of the sample.

Table 1

Characteristics of respondents to the 2016 and 2017 Medical School Graduation Questionnaire

Characteristic	Frequency (No (%) of respondents)
	Sample with imputation (n=30 651)	Sample without imputation (n=25 965)
Sex
Male	15 866 (51.8)	13 379 (51.5)
Female	14 785 (48.2)	12 586 (48.5)
Sexual orientation
Heterosexual	28 843 (94.1)	24 546 (94.5)
Lesbian, gay, or bisexual	1808 (5.9)	1419 (5.5)
Race and ethnicity
White	18 335 (59.8)	15 784 (60.8)
Non-white	12 315 (40.2)	10 181 (39.2)
Asian	6480 (21.1)	5379 (20.7)
Black/African American	1651 (5.4)	1299 (5.0)
Hispanic	1090 (3.6)	906 (3.5)
American Indian, Alaskan Native, Native Hawaiian, or Pacific Islander	94 (0.3)	74 (0.3)
Multiracial	2693 (8.8)	2269 (8.7)
Other	307 (1.0)	254 (1.0)
Unique identity group
Male; white; heterosexual	9392 (30.6)	8058 (31.0)
Male; non-white; heterosexual	5424 (17.7)	4490 (17.3)
Male; white; lesbian, gay, or bisexual	626 (2.0)	523 (2.0)
Female; white; heterosexual	7857 (25.6)	6827 (26.3)
Male; non-white; lesbian, gay, or bisexual	424 (1.4)	308 (1.2)
Female; non-white; heterosexual	6170 (20.1)	5171 (19.9)
Female; white; lesbian, gay, or bisexual	459 (1.5)	376 (1.5)
Female; non-white; lesbian, gay, or bisexual	299 (1.0)	212 (0.8)

Mistreatment

Overall, 46.0% of all 30 651 medical school graduates reported experiencing mistreatment, including 9.5% who reported recurrent experiences of multiple types of mistreatment (table 2). Larger proportions of male (59.1%) and heterosexual (54.9%) students reported no experiences of mistreatment than did their female (48.5%, P<0.001) and LGB (38.9%, P<0.001) counterparts. In the intersectional analysis of the unique identity groups, reports of recurrent experiences of multiple types of mistreatment generally increased as the number of marginalized identities held by a student increased. Compared with students who were male, white, and heterosexual, those students who identified as LGB had the highest proportions of recurrent experiences of multiple types of mistreatment (7.8% v male, white, LGB: 16.5%, P<0.001; male, non-white, LGB: 21.9%, P<0.001; female, white, LGB: 19.6%, P<0.001). Female, non-white, and LGB students had the highest proportion of students reporting recurrent experiences of multiple types of mistreatment (29.3%, P<0.001).

Table 2

Degree of general mistreatment reported by US medical school graduates, according to their sex, race or ethnicity, sexual orientation, and unique identity group (n=30 651)*

Characteristic	No reported experiences of mistreatment(n=16 544, 54.0%)	Isolated experiences of mistreatment(n=6925, 22.6%)	Recurrent experiences of one type of mistreatment(n=4277, 14.0%)	Recurrent experiences of multiple types of mistreatment(n=2904, 9.5%)	P value†
Sex
Male (n=15 866)	9379 (59.1)	3090 (19.5)	1974 (12.4)	1422 (9.0)	P<0.001
Female (n=14 785)	7165 (48.5)	3835 (25.9)	2303 (15.6)	1482 (10.0)
Race and ethnicity
White (n=18 335)	9876 (53.9)	4013 (21.9)	2821 (15.4)	1625 (8.9)	P<0.001
Non-white (n=12 315)	6668 (54.1)	2912 (23.6)	1456 (11.8)	1279 (10.4)
Sexual orientation
Heterosexual (n=28 843)	15 841 (54.9)	6509 (22.6)	3963 (13.7)	2530 (8.8)	P<0.001
Lesbian, gay, or bisexual (n=1808)	703 (38.9)	416 (23.0)	314 (17.4)	375 (20.7)
Unique identity group
Male; white; heterosexual (n=9392)	5651 (60.2)	1743 (18.6)	1270 (13.5)	729 (7.8)	P<0.001
Male; non-white; heterosexual (n=5424)	3288 (60.6)	1104 (20.4)	534 (9.9)	497 (9.2)
Male; white; lesbian, gay, or bisexual (n=626)	281 (44.9)	133 (21.2)	109 (17.4)	104 (16.5)
Female; white; heterosexual (n=7857)	3787 (48.2)	2028 (25.8)	1340 (17.1)	702 (8.9)
Male; non-white; lesbian, gay, or bisexual (n=424)	159 (37.7)	110 (26.1)	61 (14.4)	93 (21.9)
Female; non-white; heterosexual (n=6170)	3115 (50.5)	1634 (26.5)	819 (13.3)	602 (9.8)
Female; white; lesbian, gay, or bisexual (n=459)	158 (34.4)	109 (23.8)	102 (22.2)	90 (19.6)
Female; non-white; lesbian, gay, or bisexual (n=299)	105 (35.1)	64 (21.4)	42 (14.1)	88 (29.3)

Data are number (%) of students unless stated otherwise.

Row percentages might not add to 100% owing to rounding.

P value for χ2 test of the association between each sociodemographic variable and the four level mistreatment variable. P values for unique identity groups are compared to the reference group of male, white, and heterosexual medical graduates.

Discrimination

More male (84.3%), white (78.3%), and heterosexual (77.8%) students reported no experiences of discrimination than their counterparts who were female (67.5%, P<0.001), non-white (73.0%, P<0.001), and LGB (50.1%, P<0.001; table 3). In the intersectional analysis, the male, white, and heterosexual group had the largest proportion of students with no reported experiences of discrimination (89.3%, P<0.001); only 1.8% (P<0.001) of this group reported recurrent experiences of multiple types of discrimination. In contrast, those with all three marginalized identities (female, non-white, LGB) had the largest proportion of students experiencing recurrent experiences of multiple types of discrimination (30.6%, P<0.001).

Table 3

Degree of discrimination reported by US medical school graduates, according to their sex, race or ethnicity, sexual orientation, and unique identity group (n=30 651)*

Characteristic	No reported experiences of discrimination(n=19 921, 76.7%)	Isolated experiences of discrimination(n=3005, 11.6%)	Recurrent experiences of one type of discrimination(n=1928, 7.4%)	Recurrent experiences of multiple types of discrimination(n=1111, 4.3%)	P value†
Sex
Male (n=15 866)	13 371 (84.3)	1234 (7.8)	719 (4.5)	542 (3.4)	P<0.001
Female (n=14 785)	9978 (67.5)	2371 (16.0)	1561 (10.6)	875 (5.9)
Race and ethnicity
White (n=18 335)	14 355 (78.3)	1938 (10.6)	1422 (7.8)	620 (3.4)	P<0.001
Non-white (n=12 315)	8995 (73.0)	1667 (13.5)	858 (7.0)	797 (6.5)
Sexual orientation
Heterosexual (n=28 843)	22 444 (77.8)	3253 (11.3)	2059 (7.1)	1086 (3.8)	P<0.001
Lesbian, gay, or bisexual (n=1808)	906 (50.1)	352 (19.5)	220 (12.2)	331 (18.3)
Unique identity group
Male; white; heterosexual (n=9392)	8387 (89.3)	484 (5.2)	355 (3.8)	167 (1.8)	P<0.001
Male; non-white; heterosexual (n=5424)	4403 (81.2)	538 (9.9)	258 (4.8)	224 (4.1)
Male; white; lesbian, gay, or bisexual (n=626)	357 (56.9)	132 (21.1)	67 (10.7)	71 (11.3)
Female; white; heterosexual (n=7857)	5409 (68.8)	1231 (15.7)	922 (11.7)	295 (3.8)
Male; non-white; lesbian, gay, or bisexual (n=424)	224 (53.0)	80 (18.8)	39 (9.2)	81 (19.0)
Female; non-white; heterosexual (n=6170)	4245 (68.8)	1000 (16.2)	525 (8.5)	400 (6.5)
Female; white; lesbian, gay, or bisexual (n=459)	203 (44.1)	91 (19.7)	79 (17.1)	88 (19.1)
Female; non-white; lesbian, gay, or bisexual (n=299)	122 (40.9)	49 (16.5)	36 (12.0)	92 (30.6)

Data are number (%) of students unless stated otherwise.

Row percentages might not add to 100% owing to rounding.

P value for χ2 test of the association between each sociodemographic variable and the four level mistreatment variable. P values for unique identity groups are compared to the reference group of male, white, and heterosexual medical graduates.

Burnout—exhaustion

Generally, female students had significantly higher exhaustion scores than male students (adjusted mean difference 0.80, 95% confidence interval 0.71 to 0.88; table 4). Non-white students had higher exhaustion scores than white students (0.65, 0.56 to 0.73), and LGB students had higher average exhaustion scores than heterosexual students (0.89, 0.69 to 1.10). Adjusting for mistreatment and discrimination partially explained this observed main effect. In the intersectional analysis, as the number of marginalized identities increased from 0 to 3, so did the average exhaustion score. Students who were female, were non-white, and identified as LGB (that is, who had all three marginalized identities) had the largest difference in average exhaustion score compared to male, white, and heterosexual students (1.96, 1.47 to 2.44). With the addition of mistreatment and discrimination to the model, differences in average exhaustion scores (that is, the main effect) decreased for all unique identity groups, although the effects were not fully explained.

Table 4

Mediating effects of mistreatment and discrimination on associations between sociodemographic variables and experiences of exhaustion*

Sociodemographic variable	Adjusted mean difference β (95% CI)
	Main effect	Adjusted for degree of general mistreatment	Adjusted for degrees of general mistreatment and discrimination
Model 1: sex
Male	Reference	Reference	Reference
Female	0.80 (0.71 to 0.88)	0.64 (0.56 to 0.72)	0.56 (0.48 to 0.65)
Model 2: race and ethnicity
White	Reference	Reference	Reference
Non-White	0.65 (0.56 to 0.73)	0.65 (0.57 to 0.74)	0.59 (0.50 to 0.68)
Model 3: sexual orientation
Heterosexual	Reference	Reference	Reference
Lesbian, gay, or bisexual	0.89 (0.69 to 1.10)	0.46 (0.27 to 0.65)	0.31 (0.12 to 0.50)
Model 4: unique identity group
Male; white; heterosexual	Reference	Reference	Reference
Male; non-white; heterosexual	0.53 (0.41 to 0.66)	0.55 (0.43 to 0.67)	0.50 (0.38 to 0.62)
Male; white; lesbian, gay, or bisexual	1.11 (0.80 to 1.41)	0.75 (0.46 to 1.04)	0.60 (0.31 to 0.89)
Female; white; heterosexual	0.74 (0.63 to 0.85)	0.57 (0.47 to 0.68)	0.48 (0.37 to 0.59)
Male; non-white; lesbian, gay, or bisexual	1.69 (1.28 to 2.10)	1.18 (0.78 to 1.58)	1.00 (0.61 to 1.39)
Female; non-white; heterosexual	1.40 (1.28 to 1.52)	1.27 (1.16 to 1.39)	1.17 (1.06 to 1.29)
Female; white; lesbian, gay, or bisexual	1.56 (1.18 to 1.93)	1.00 (0.64 to 1.35)	0.76 (0.42 to 1.11)
Female; non-white; lesbian, gay, or bisexual	1.96 (1.47 to 2.44)	1.28 (0.83 to 1.74)	1.01 (0.56 to 1.47)

Table displays results of the linear regression models run in the imputed dataset (n=30 651). All models included adjustment for average burnout score.

Burnout—disengagement

On average, students who were non-white and identified as LGB had higher disengagement scores than students who were white (adjusted mean difference 0.28, 95% confidence interval 0.19 to 0.37) and heterosexual (0.73, 0.52 to 0.94; table 5). Adjusting for mistreatment and discrimination partially explained the main effect. In contrast, female students had significantly lower disengagement scores than male students (−0.76, −0.85 to −0.68). After adjusting for mistreatment and discrimination, the difference in average disengagement scores between male and female students increased, with female students being even less disengaged than their male counterparts. Additionally, in our intersectional analysis, we observed that any unique identity group that included female sex had significantly lower average disengagement scores than the reference group of male, white, and heterosexual students, after adjusting for mistreatment and discrimination (female, white, heterosexual: adjusted mean difference −1.13, 95% confidence interval −1.24 to −1.02; female, non-white, heterosexual: −0.65, −0.78 to −0.53; female, white, LGB: −1.06, −1.42 to −0.70; female, non-white, LGB: −0.72, −1.17 to −0.26).

Table 5

Mediating effects of mistreatment and discrimination on associations between sociodemographic variables and disengagement*

Sociodemographic variable	Adjusted mean difference β (95% CI)
	Main effect	Adjusted for degree of general mistreatment	Adjusted for degrees of general mistreatment and discrimination
Model 1: sex
Male	Reference	Reference	Reference
Female	−0.76 (−0.85 to −0.68)	−0.89 (−0.98 to −0.81)	−1.00 (−1.09 to −0.92)
Model 2: race and ethnicity
White	Reference	Reference	Reference
Non-White	0.28 (0.19 to 0.37)	0.29 (0.20 to 0.37)	0.20 (0.11 to 0.29)
Model 3: sexual orientation
Heterosexual	Reference	Reference	Reference
Lesbian, gay, or bisexual	0.73 (0.52 to 0.94)	0.34 (0.14 to 0.53)	0.15 (−0.04 to 0.34)
Model 4: unique identity group
Male; white; heterosexual	Reference	Reference	Reference
Male; non-white; heterosexual	0.22 (0.09 to 0.34)	0.23 (0.10 to 0.35)	0.17 (0.04 to 0.29)
Male; white; lesbian, gay, or bisexual	0.68 (0.36 to 0.99)	0.34 (0.05 to 0.64)	0.15 (−0.15 to 0.45)
Female; white; heterosexual	−0.87 (−0.98 to −0.76)	−1.02 (−1.13 to −0.91)	−1.13 (−1.24 to −1.02)
Male; non-white; lesbian, gay, or bisexual	0.92 (0.52 to 1.32)	0.44 (0.06 to 0.82)	0.20 (−0.18 to 0.57)
Female; non-white; heterosexual	−0.41 (−0.54 to −0.29)	−0.53 (−0.64 to −0.41)	−0.65 (−0.78 to −0.53)
Female; white; lesbian, gay, or bisexual	−0.23 (−0.61 to 0.16)	−0.74 (−1.11 to −0.38)	−1.06 (−1.42 to −0.70)
Female; non-white; lesbian, gay, or bisexual	0.31 (−0.16 to 0.78)	−0.34 (−0.80 to 0.11)	−0.72 (−1.17 to −0.26)

Table displays results of the linear regression models run in the imputed dataset (n=30 651). All models included adjustment for average burnout score.

Discussion

Principal findings

This study’s major finding is that a higher proportion of students with multiple marginalized identities report mistreatment and discrimination. Students with all three marginalized identities (female, non-white, LGB) had the largest proportion of individuals who cited recurrent experiences of discrimination during medical school (30.6%, P<0.001). While prior research has indicated greater mistreatment and discrimination among female, non-white, and LGB students during undergraduate medical education when examining these aspects of identity in isolation, in this study we take an intersectional approach using national data.2 26

These results also show a significant association between multiple marginalized identities and burnout in these medical students. These findings are consistent with prior work while adding nuance to the experience of burnout, especially along the dimension of exhaustion. The higher exhaustion scores among historically marginalized students is consistent with prior non-intersectional studies that use the minority tax and racial battle fatigue frameworks to describe experiences among racial or ethnic minority medical trainees and faculty.27 28 The additional pressure marginalized groups face to represent their communities or contribute to diversity efforts at their institutions and experiences of bias and discrimination likely contributes to the effect observed with this dimension of burnout. Further, prior studies point to the lack of mentorship opportunities and working toward recognition and professional advancement in a traditional academic setting as potential factors that contribute to exhaustion among students and physicians from marginalized backgrounds.29 30

Another key finding from this study is that disengagement, unlike exhaustion, was lower among female students regardless of their other identities. Female students, irrespective of their other identities, were even less disengaged than male students after adjusting for mistreatment and discrimination. Considering that mistreatment and discrimination were significantly associated with both female sex and higher disengagement scores, this observed effect of female sex and disengagement was one of negative confounding. This distinction is important because it suggests that interventions to reduce burnout might be different depending on the student’s sex. Although disengagement was lower among female students overall, this study’s intersectional approach did show variation in disengagement scores when accounting for both the student’s race or ethnicity and sexual orientation. Further investigation of this relation might help with understanding the attrition trends observed among female physicians in academic medicine.31

Last, LGB students, irrespective of their other reported identities, were shown to have higher exhaustion and disengagement scores. While the reasons for these associations are multifactorial, previous studies have shown that many LGB students are reluctant to disclose their sexual orientation, and many conceal this facet of their identity during medical school.32 33 34 Such concealment can act as a stressor that contributes to fatigue and impedes on one’s ability to meet educational objectives.35 36 Owing to a pervasive heteronormative climate (that is, the assumption that heterosexuality is the default), LGB students might witness more discrimination against LGB groups by bigoted people who might not regulate their actions as they would with more observable identities such as race or gender.

Finally, although mistreatment and discrimination generally mediated the association between unique identity group and symptoms of exhaustion, these factors did not completely explain the effect. Therefore, other factors in the learning environment might contribute to the disproportionate reports of symptoms of exhaustion by students with multiple marginalized identities.

Implications

By using an intersectional framework, we have been able to obtain a better understanding of the simultaneous influence of sex, race or ethnicity, and sexual orientation on medical students’ experience of mistreatment and on their symptoms of burnout. Our findings suggest that focusing on one aspect of social identity could underestimate the degree of harm students with multiple marginalized identities experience in their learning environment. We provide some considerations for academic medicine leadership to tailor programs to support a diverse set of future physicians.

This study’s findings support the need for more integration of students’ support services and for more interaction between student affinity groups. Traditionally, student groups for women, ethnic minorities, and LGBTQI+ groups (lesbian, gay, bisexual, transgender, queer, intersex+) have operated in isolation. Increased interaction between groups that focus on a singular aspect of identity could provide a support structure for students with multiple marginalized identities.

Additionally, the strong association between exhaustion and a student having multiple marginalized identities reveals that the cognitive and physical demands of medical school, rather than generally negative attitudes about medical school and its subject matter, have a larger role in experiences of burnout symptoms among medical students with multiple marginalized identities. Interventions targeting fatigue might be a promising starting point to manage the exhaustion disparately reported by these students. For example, a two category, pass-fail grading structure has been associated with less emotional exhaustion and might lead to improvements in this facet of burnout among marginalized students.37

Despite gains in diversity among medical students by sex and race or ethnicity over the last 10 years, the diversity of faculty has remained stagnant.38 39 Prior research has shown that students from marginalized background consider the lack of concordant mentors as a large barrier to their medical education experience.40 41 42 43 44 45 Additionally, prior work has called for increasing faculty diversity as an evidence based intervention to mitigate bias.46 Institutions should increase efforts for recruitment and retention in this respect.

Our finding that LGB students across unique identity groups had higher exhaustion and disengagement scores suggests a greater need for institutional support for these students. Such support at various institutions has come in the form of financial resources to student and faculty affinity groups and meaningful incorporation of LGBTQI+ health topics in medical school curriculums.45 47

Further, the fact that students with multiple marginalized identities reported greater exposure to recurrent mistreatment and discrimination indicates the need for leaders in academic medicine to improve the existing mechanisms for dealing with reports of negative behaviors and create a climate of respect. A restorative justice approach, which brings together members of a community in a safe and open forum to collaboratively devise solutions to hurtful or disruptive incidents could be an alternative. Moreover, a restorative justice process that uses an intersectional approach can provide a space to show how negative behavior might reflect larger structures of oppression.48 49

Limitations

This study had several limitations. The 2016 and 2017 Medical School Graduation Questionnaire surveys graduating students, and students who experienced burnout during medical school could have left school before graduation. Historically, the national attrition rate of students at US medical schools has remained at 3.3% on average.50 Relatedly, graduating students who experienced burnout might not have completed the questionnaire. In addition, the questionnaire might not capture all types of mistreatment; and because students self-report mistreatment in the questionnaire, the definitions of negative behaviors that constitute mistreatment might differ among students, faculty, and residents.

Our analysis only focused on three aspects of identity. Additional social positions, such as socioeconomic status and disability, influence privilege and discrimination in complex ways. An intersectional perspective asserts that dimensions of identity are interdependent—that is, one identity cannot explain disparate outcomes without the consideration of other social identities.51 52 Although the unique identity combinations in our study allowed us to examine the experiences of students with multiple marginalized identities, these different aspects of identity do not contribute to student exposure to mistreatment and burnout in a uniform fashion.

Conclusion

As the number of marginalized identities that a student holds increases, so do their scores for the exhaustion dimension of burnout. Disparate exposure to mistreatment and discrimination and the resulting differences in burnout scores among students with intersecting marginalized identities indicates a need for institutions to create a more equitable and inclusive learning environment.

Mistreatment, which is associated with burnout, is a common experience for students during medical school

Prior research has reported disparate experiences of mistreatment and discrimination among students with identities historically marginalized in medicine, with additional stress on students who are female; non-white; and lesbian, gay, or bisexual

Although studies have begun to explore the relation between various social identities and reported mistreatment and burnout, research has yet to consider multiple aspects of a student’s identity

This study takes an intersectional approach using national data to examine the associations between multiple facets of a medical student’s identity, reported mistreatment, reported discrimination, and student burnout

The findings suggest that focusing on one aspect of identity might underestimate the degree of harm students with multiple marginalized identities experience in their learning environment