Graduate and professional programs in the biomedical and health sciences are stressful environments for junior researchers and clinicians. Indeed, the mental health of medical students declines during medical school and is worse than that of the general population by graduation (Dyrbye et al., 2005; Dyrbye et al., 2006; Brazeau et al., 2014). In fact, poor mental health in medical education is evidenced by a robust, decades-old body of literature, including a systematic review of mental health evidence-based research encompassing nearly 200 relevant studies to date (e.g., Slavin et al., 2014; Rotenstein et al., 2016; Quek et al., 2019).

In comparison, evidence-based research on mental health of doctoral students has only begun to emerge over the past five years. Recent reports indicate elevated rates of anxiety and depression among graduate students in all disciplines (UC Berkeley Graduate Assembly, 2014; Evans et al., 2018; Nagy et al., 2019; Levecque et al., 2017). For example, a multi-campus study of undergraduate and graduate students found that 17.3% scored positive for depression, 9.8% scored positive for anxiety, and 6.3% reported suicidal ideation (Eisenberg et al., 2013). One study found that graduate students in the Faculty of Medicine at the University of Calgary have higher anxiety and depression than undergraduates (Toews et al., 1993), and a survey of graduate students at a university in the United States showed that approximately half of respondents reported emotional or stress-related problems (Hyun et al., 2006). In an international sample of Masters’ and PhD students, 39% of respondents indicated moderate to severe depression, and 41% moderate to severe anxiety scores (Evans et al., 2018). In comparison, the Centers for Disease Control and Prevention (CDC) estimate that approximately 16% of adults in the United States experience anxiety, and approximately 5% depression. Moreover, approximately 1 in 6 Americans will experience depression in their lifetime, and more women than men experience anxiety (CDC, 2021; Terlizzi and Villarroel, 2020). However, while these general population incidences can be informative as a benchmark, these rates are not truly comparable to the student population due to multiple potential confounding factors (e.g., age group, socio-economic status, education levels of graduate level students, etc.).

Less is known about the mental health of biomedical doctoral students specifically, as disparate disciplines are often combined when graduate students are included in student mental health studies. Still, these students train in environments which may feature long work hours, pressures to produce, influential and sometimes unsupportive relationships with advisors, as well as financial concerns, uncertainty about future employment, and non-transparent university processes which may all negatively impact doctoral student mental health (Hazell et al., 2020; Mackie and Bates, 2019).

Of note, some studies highlight that Asian, Black, Hispanic, and multi-racial undergraduate students score higher for depression than white students (Eisenberg et al., 2013). Many theories have attempted to account for these racial and ethnic disparities, such as the effects of structural racism on symptoms, diagnosis, treatment and access to care (Kendi, 2019) and the use of white populations as the baseline norm (Legha and Miranda, 2020; for review see Conrad, 2022). Additionally, studies based on data from the national survey on drug use and health have shown ongoing mental health disparities among people identifying as Black, Indigenous, and people of color (BIPOC), women, and lesbian, gay, bisexual, and queer (LGBQ+) people (SAMHSA, 2021). These populations are also particularly at risk in academia due to structural inequities, barriers, and microaggressions among other challenges (for example, academic cultural barriers and stressors for BIPOC, Halsey et al., 2020; barriers for women doctoral students, see Carter et al., 2013; LGBQ+ academics experience toxic environments and microaggressions, see Beagan et al., 2021 and Linley and Nguyen, 2015) that can impact the mental health of these populations. To acknowledge the systemic barriers faced by historically excluded groups, and to recognize that people of certain backgrounds were intentionally excluded from parts of American society for various political, economic, and social reasons (e.g., Berhe et al., 2021; Rollnick, 2015) the terms ‘historically excluded by race/ethnicity, gender, or sexual orientation’ will be used to refer to these groups respectively. While the list of historically excluded groups is not comprehensive, we aim to decenter Whiteness by referring to historically excluded (HE) versus non-historically excluded (NHE) groups.

The COVID-19 pandemic has resulted in globally increased symptoms of anxiety, depression, post-traumatic stress disorder and psychological stress, particularly among healthcare workers, those with pre-existing mental health conditions, women, college students, and individuals under 45 (Xiong et al., 2020). Furthermore, it may also have impacted mental health during academic training (Byrom, 2020). Compounding systemic inequities and racial injustice, COVID-19 disproportionally impacted communities of color in the United States (Li, 2020; Webb Hooper et al., 2020; CDC, 2020). In particular, the highest risk of age-adjusted mortality during COVID was identified for Hawaiian and other Pacific Islander, American Indian or Alaska Native, and Latinx or Hispanic people (Feldman and Bassett, 2021).

While students themselves may be classified as associated with some lower risk groups for direct effects of COVID due to their levels of education, families of students identifying as BIPOC may be disproportionately affected, especially when intersecting with risk factors associated with social class (e.g., Pathak et al., 2021; Feldman and Bassett, 2021). Hence, students historically excluded because of their race and ethnicity (also termed persons excluded due to ethnicity or race, see Asai, 2020) may experience disproportionate impact and heightened concerns about individual, community and family health (Pathak et al., 2021; Feldman and Bassett, 2021; Limas, 2021; Blake et al., 2021). Indeed, a recent longitudinal study on undergraduates at the University of North Carolina at Chapel Hill (UNC-Chapel Hill; same location as the present study) found that the prevalence of anxiety and depression increased among first-year undergraduate students of color, sexual minority students, and women-identifying students during the pandemic (Fruehwirth et al., 2021).

Moreover, the stress caused by the COVID-19 pandemic coincided with heightened responses and protests against the persistent racial injustice found in the United States – notably killings of Black Americans by police and vigilantes. These also took place alongside heightened anti-Asian violence, which may have also affected the mental health of Asian American students within academia.

At UNC-Chapel Hill, a large public university in the United States, the pandemic impacted courses and degrees in different ways. Most biomedical doctoral programs require didactic classes in the first two years only; afterwards students require full-time laboratory access to continue their research for another three to four years, precluding a transition to fully online instruction. Due to health and safety measures, biomedical doctoral student training was severely curtailed at our institution in the spring of 2020, and students returned to labs in June 2020 with strict occupancy limits. In contrast, medical student training is characterized by two years of didactic coursework and frequent test-taking, followed by another two years of sequential clinical experiences in hospitals or other clinical settings. In 2020, medical students could continue their academic progress with online learning options fostered by a switch to remote instruction in place of didactic in-person courses and clinical rotations. Hence, these students were temporarily removed from clinical rotations due to shortages of personal protective equipment; conversely though, graduation dates were not delayed.

Given this background, we sought to examine effects of the concurrent COVID-19 pandemic and heightened community reactions to racial injustice between 2019 and 2020 by exploring the mental health of medical students and biomedical doctoral students at UNC-Chapel Hill. We investigate comparisons between historically excluded (HE) and non-historically excluded (NHE) groups in science, focused on race/ethnicity, gender, and sexual orientation in medical and graduate biomedical students during the intersection of the COVID-19 pandemic and a time of reckoning of persistent racial injustice. Recognizing that an intersectional approach encompasses a multitude of identities (Cho et al., 2013), the facets of identity explored in this study are not intended to be comprehensive, but rather to represent major subpopulations represented in our sample and commonly found in US biomedical graduate and medical education.

The UNC-Chapel Hill School of Medicine annually enrolls approximately 800 medical and 600 biomedical PhD students; part of the student body is representative of diverse identities including trainees from a variety of racial and ethnic groups, genders and LGBQ +identities. For example, a recent report (UNC School of Medicine Office of Diversity Equity and Inclusion, 2021) indicated that the medical student population at UNC-Chapel Hill consisted of 54% women, 58% white, 14% Asian, and 24% underrepresented students (4% not reported); the biomedical graduate student population consisted of 57% white, 15% Asian, and 26% underrepresented students (2% not reported). The present study was reviewed and approved by the Institutional Review Board (#18–0112).

The annual School of Medicine survey used for this study includes self-reported mental health status (examined in this work), as well as assessment of current and desired student-support programs. It is conducted to inform current and future mental health and wellness programming and global satisfaction with services provided (https://doi.org/10.17605/OSF.IO/H9UCX). The same School of Medicine survey was administered to medical and biomedical doctoral students approximately one year apart in 2019 and 2020 (n=431, Fall 2020; n=526, Fall 2019). Each survey was open for four weeks between September and October and distributed using the same mechanisms (e.g., same internal listservs) to be as comparable as possible.

In the primary analysis, mental health data was analyzed by type of training program (MD vs. PhD; that is, medical students vs. biomedical doctoral students), year (2019 vs 2020), and historically excluded vs. non-historically excluded (HE vs NHE) populations based on race/ethnicity (HE-RE vs. NHE-RE) and gender (HE-G vs. NHE-G, that is students identifying as women vs. men). A post hoc analysis included the primary variables as well as historically excluded on the basis of sexual orientation (HE-SO vs. NHE-SO, that is LGBQ+ vs. non-LGBQ+ populations).

The logistic regression model identified significant differences on student mental health outcomes for depression and anxiety by year (interactions), as well as by program (biomedical doctoral students scored worse than medical students), and for those historically on the basis of race/ethnicity (HE-RE) and by gender (HE-G) (Figure 1A).

Figure 1

Download asset Open asset

In the combined population (when medical and biomedical doctoral students are considered together) there was a decrease in depression and anxiety, likely driven by the medical student improvements; see Discussion. In general, HE-RE students exhibited significantly higher rates of depression and anxiety at about 1.5 times the rate of their NHE-RE peers across the combined medical and biomedical doctoral student sample. No differences were evidenced on either drug or alcohol use by year, type of program or historically excluded/non-excluded status.

Overall, across medical and biomedical doctoral students, HE-RE students were approximately twice as likely to say they had thought about ending their life in the last 12months than their NHE-RE peers, and about 1.8 times more likely while enrolled (Figure 1B); such high rates are extremely concerning. Due to different patterns emerging for training type and racial/ethnic identity by year, examining effect solely by year obscured these differences, hence interactions were further explored.

Medical school versus biomedical doctoral training

Prior to 2020, both medical and biomedical doctoral students suffered from depression (46% of medical students, 65% of biomedical doctoral students) and anxiety (47% of medical students, 67% of biomedical doctoral students) at high rates (Figure 2A and B), as defined by no symptoms compared with any symptomatic categories (see Methods for categorical definitions). Doctoral student mental health in 2020 remained very poor (depression: 26% of medical students, 64% of biomedical doctoral students; anxiety: 32% of medical students, 61% of biomedical doctoral students), whereas, surprisingly, the mental health of medical students improved. Suicidal ideation ‘in the last 12 months’ among biomedical doctoral students (compared with medical students) was markedly higher, both before (11% of medical students, 16% of biomedical doctoral students) and during 2020 (6% of medical students, 19% of biomedical doctoral students).

Figure 2

Download asset Open asset

We assessed differences between medical and biomedical doctoral students for depression (PHQ-9), anxiety (GAD), problematic drug use (DAST), and hazardous alcohol use (AUDIT) using nominal outcome variables (Figure 2A and B). We identified significantly higher rates of depression and anxiety for biomedical doctoral students compared with medical students, as well as higher rates of suicidal ideation ‘while enrolled’ and ‘in the last 12 months’. No significant differences emerged for problems with drug use or hazardous alcohol use by training type, year, or historical exclusion by race/ethnicity; furthermore, problems with drug use rates were comparatively low in contrast to depression and anxiety. Known gender effects were consistent with expectations for both problematic substance use categories (greater use found in prior studies is also reflected in our sample for NHE-G vs. HE-G).

Post hoc analysis: historically excluded versus non-historically excluded students based on sexual orientation (HE-SO vs. NHE-SO)

To further examine additional aspects of intersectionality, we completed a post hoc analysis including LGBQ+ identities (HE-SO) into a simplified model (controlling for main effects of the four primary variables and identifying any significant two-way interactions). Due to a smaller sample size, this was included as a post hoc analysis.

The HE-SO variable showed significant main effects across main outcome variables, with HE-SO participants more than twice as likely to experience depression (OR = 2.19, P<.0003) and anxiety (OR = 2.66, P<.0001). In addition, there was a significant interaction with ‘Year’ for depression (P=.03), such that 2020 was significantly worse for HE-SO depression than 2019.

There were also main effects for alcohol use (OR = 2.31, P<.01) and drug use (OR = 2.37, P<.0002), such that HE-SO participants were more at risk for substance use; however, an interaction between gender and sexual orientation (P=.03) indicated that among HE-SO participants, NHE-G (LGBQ +men) were more likely to report drug use than HE-G participants (LGBQ+ women; conditional OR = .29, P<.003). HE-SO NHE-G participants (LGBQ+ men) compared to NHE-SO HE-G participants (LGBQ+ women) were nearly four times as likely to report drug use (conditional OR = 3.91, P<.0002), whereas HE-SO HE-G participants (LGBQ+ women) were not significantly more likely to report drug use (P=.19).

HE-SO participants experienced more suicidal ideation (ever, OR = 4.37, P<.001; while enrolled, OR = 3.66, P<.0001; in the last 12 months, OR = 4.77, P<.0001; see Figure 3 for visualization of comparative percentages). In addition, there was a marginal interaction between ‘HE-SO’ and ‘Program’ (P=.05), such that medical participants were much less likely (OR = .25) than biomedical doctoral students to indicate suicidal ideations while enrolled. Hence, biomedical doctoral HE-SO participants were almost six times more likely to report suicidal ideations than NHE-SO biomedical doctoral participants while enrolled (conditional OR = 5.94, P<.0001) whereas HE-SO medical students where about twice as likely to indicate suicidal ideations while enrolled than NHE-SO medical students (conditional OR = 2.925, P<.02).

Figure 3

Download asset Open asset

Overall, medical student mental health improved from 2019–2020 on measures of both depression and anxiety, whereas biomedical doctoral students showed no change, their scores remaining concernedly high. HE-RE students were worse off than their NHE-RE peers in measures of depression, anxiety, and suicidal ideation – particularly reporting more suicidal ideation than their counterparts both ‘while enrolled’ and in the ‘last 12 months’. All analyses controlled for gender, for which women (HE-G) generally experienced anxiety and depression more than men (NHE-G), whereas men generally reported higher substance use. Post hoc analyses also indicated that those identifying as HE-SO (LGBQ+ respondents) experienced more anxiety and depression than NHE-SO (non-LGBQ+) respondents, as well as greater suicidal ideation.

Underlying data is available on the Open Science Framework at https://doi.org/10.17605/OSF.IO/H9UCX. In accordance with IRB approval, the presentation of data does not include demographic information that could potentially lead to the identification of a student. Furthermore, to protect student anonymity, responses with fewer than ten individuals will not be shared. Accordingly, data sharing is limited to those variables directly relevant to the analyses conducted (e.g., underlying data) and composite categories (e.g., underrepresented vs. well-represented) are shared when applicable rather than granular demographic information to protect the identity of respondents also noted in in the Human Subjects section.

1. 1. Schad A
   2. Layton RL
   3. Ragland D
   4. Cook JG

   (2021) Open Science Framework

   Mental health disparities among biomedical trainees during COVID-19 and racial inequity.

   https://doi.org/10.17605/OSF.IO/H9UCX

1. Website

   1. AAMC Executive Council

   (1992) Recommendations regarding health services for medical students

   Accessed August 8, 2022.

   https://www.aamc.org/professional-development/affinity-groups/gsa/health-services-recommendations
2. 1. Asai D

   (2020) Excluded

   Journal of Microbiology & Biology Education 21:21.1.18.

   https://doi.org/10.1128/jmbe.v21i1.2071

   • PubMed
   • Google Scholar
3. 1. Beagan BL
   2. Mohamed T
   3. Brooks K
   4. Waterfield B
   5. Weinberg M

   (2021) Microaggressions experienced by LGBTQ academics in Canada: “just not fitting in… it does take a toll.”

   International Journal of Qualitative Studies in Education 34:197–212.

   https://doi.org/10.1080/09518398.2020.1735556

   • Google Scholar
4. 1. Berhe AA
   2. Barnes RT
   3. Hastings MG
   4. Mattheis A
   5. Schneider B
   6. Williams BM
   7. Marín-Spiotta E

   (2021) Scientists from historically excluded groups face a hostile obstacle course

   Nature Geoscience 15:2–4.

   https://doi.org/10.1038/s41561-021-00868-0

   • Google Scholar
5. 1. Blake H
   2. Brown N
   3. Follette C
   4. Morgan J
   5. Yu H

   (2021) Black, Indigenous, people of color, and international students: experiences and resolutions beyond COVID-19

   American Journal of Public Health 111:384–386.

   https://doi.org/10.2105/AJPH.2020.306118

   • PubMed
   • Google Scholar
6. Website

   1. Bogel-Burroughs N

   (2020) Antivaccination activists are growing force at virus protests

   The New York Times. Accessed April 5, 2022.

   https://www.nytimes.com/2020/05/02/us/anti-vaxxers-coronavirus-protests.html
7. 1. Brazeau C
   2. Shanafelt T
   3. Durning SJ
   4. Massie FS
   5. Eacker A
   6. Moutier C
   7. Satele DV
   8. Sloan JA
   9. Dyrbye LN

   (2014) Distress among matriculating medical students relative to the general population

   Academic Medicine 89:1520–1525.

   https://doi.org/10.1097/ACM.0000000000000482

   • PubMed
   • Google Scholar
8. 1. Byrom N

   (2020) COVID-19 and the research community: the challenges of lockdown for early-career researchers

   eLife 9:e59634.

   https://doi.org/10.7554/eLife.59634

   • Google Scholar
9. 1. Carter S
   2. Blumenstein M
   3. Cook C

   (2013) Different for women? The challenges of doctoral studies

   Teaching in Higher Education 18:339–351.

   https://doi.org/10.1080/13562517.2012.719159

   • Google Scholar
10. Website

    1. CDC

    (2020) Health equity considerations & racial & ethnic minority groups

    Accessed April 7, 2022.

    https://www.cdc.gov/coronavirus/2019-ncov/community/health-equity/race-ethnicity.html
11. Website

    1. CDC

    (2021) Depression

    Accessed April 7, 2022.

    https://www.cdc.gov/nchs/fastats/depression.htm
12. 1. Chen JH
    2. Li Y
    3. Wu AMS
    4. Tong KK

    (2020) The overlooked minority: mental health of international students worldwide under the COVID-19 pandemic and beyond

    Asian Journal of Psychiatry 54:102333.

    https://doi.org/10.1016/j.ajp.2020.102333

    • PubMed
    • Google Scholar
13. 1. Cho S
    2. Crenshaw KW
    3. McCall L

    (2013) Toward a field of intersectionality studies: theory, applications, and praxis

    Signs 38:785–810.

    https://doi.org/10.1086/669608

    • Google Scholar
14. 1. Clay RA

    (2011)

    Muslims in America, post 9/11

    American Psychological Association 42:8.

    • Google Scholar
15. 1. Conrad JA

    (2022) A black and white history of psychiatry in the United States

    The Journal of Medical Humanities 43:247–266.

    https://doi.org/10.1007/s10912-020-09650-6

    • PubMed
    • Google Scholar
16. 1. Couée I

    (2020) Restoring public trust in science with the help of the humanities

    EMBO Reports 21:e50796.

    https://doi.org/10.15252/embr.202050796

    • PubMed
    • Google Scholar
17. Website

    1. Crawford NC
    2. Graves L
    3. Katzenstein J

    (2021) Costs of war: racial profiling and islamophobia

    Brown University Watson Institute of International Affairs. Accessed August 8, 2022.

    https://watson.brown.edu/costsofwar/costs/social/rights/profiling
18. 1. Daraiseh I

    (2012)

    Effects of Arab American discrimination post 9/11 in the contexts of the workplace and education

    McNair Scholars Research Journal 4:3.

    • Google Scholar
19. 1. Dyrbye LN
    2. Thomas MR
    3. Shanafelt TD

    (2005) Medical student distress: causes, consequences, and proposed solutions

    Mayo Clinic Proceedings 80:1613–1622.

    https://doi.org/10.4065/80.12.1613

    • PubMed
    • Google Scholar
20. 1. Dyrbye LN
    2. Thomas MR
    3. Shanafelt TD

    (2006) Systematic review of depression, anxiety, and other indicators of psychological distress among U.S. and Canadian medical students

    Academic Medicine 81:354–373.

    https://doi.org/10.1097/00001888-200604000-00009

    • PubMed
    • Google Scholar
21. 1. Eisenberg D
    2. Hunt J
    3. Speer N

    (2013) Mental health in American colleges and universities: variation across student subgroups and across campuses

    The Journal of Nervous and Mental Disease 201:60–67.

    https://doi.org/10.1097/NMD.0b013e31827ab077

    • PubMed
    • Google Scholar
22. 1. Evans TM
    2. Bira L
    3. Gastelum JB
    4. Weiss LT
    5. Vanderford NL

    (2018) Evidence for a mental health crisis in graduate education

    Nature Biotechnology 36:282–284.

    https://doi.org/10.1038/nbt.4089

    • PubMed
    • Google Scholar
23. 1. Feldman JM
    2. Bassett MT

    (2021) Variation in COVID-19 mortality in the US by race and ethnicity and educational attainment

    JAMA Network Open 4:e2135967.

    https://doi.org/10.1001/jamanetworkopen.2021.35967

    • PubMed
    • Google Scholar
24. 1. Fruehwirth JC
    2. Biswas S
    3. Perreira KM

    (2021) The COVID-19 pandemic and mental health of first-year college students: examining the effect of COVID-19 stressors using longitudinal data

    PLOS ONE 16:e0247999.

    https://doi.org/10.1371/journal.pone.0247999

    • PubMed
    • Google Scholar
25. 1. Goggin G
    2. Ellis K

    (2020) Disability, communication, and life itself in the COVID-19 pandemic

    Health Sociology Review 29:168–176.

    https://doi.org/10.1080/14461242.2020.1784020

    • PubMed
    • Google Scholar
26. Website

    1. Gordon J

    (2021) One year in: COVID-19 and mental health

    National Institutue of Mental Health. Accessed August 8, 2022.

    https://www.nimh.nih.gov/about/director/messages/2021/one-year-in-covid-19-and-mental-health
27. 1. Gray DM
    2. Anyane-Yeboa A
    3. Balzora S
    4. Issaka RB
    5. May FP

    (2020) COVID-19 and the other pandemic: populations made vulnerable by systemic inequity

    Nature Reviews. Gastroenterology & Hepatology 17:520–522.

    https://doi.org/10.1038/s41575-020-0330-8

    • PubMed
    • Google Scholar
28. 1. Gross M

    (2020) Communicating science in a crisis

    Current Biology 30:R737–R739.

    https://doi.org/10.1016/j.cub.2020.06.052

    • Google Scholar
29. 1. Halsey SJ
    2. Strickland LR
    3. Scott-Richardson M
    4. Perrin-Stowe T
    5. Massenburg L

    (2020) Elevate, don’t assimilate, to revolutionize the experience of scientists who are Black, Indigenous and people of colour

    Nature Ecology & Evolution 4:1291–1293.

    https://doi.org/10.1038/s41559-020-01297-9

    • PubMed
    • Google Scholar
30. 1. Hazell CM
    2. Chapman L
    3. Valeix SF
    4. Roberts P
    5. Niven JE
    6. Berry C

    (2020) Understanding the mental health of doctoral researchers: a mixed methods systematic review with meta-analysis and meta-synthesis

    Systematic Reviews 9:1–30.

    https://doi.org/10.1186/s13643-020-01443-1

    • PubMed
    • Google Scholar
31. 1. Hunt JE

    (2020) COVID-19 and the research community: being vulnerable

    eLife 9:e59285.

    https://doi.org/10.7554/eLife.59285

    • Google Scholar
32. 1. Hyun JK
    2. Quinn BC
    3. Madon T
    4. Lustig S

    (2006) Graduate student mental health: needs assessment and utilization of counseling services

    Journal of College Student Development 47:247–266.

    https://doi.org/10.1353/csd.2006.0030

    • Google Scholar
33. Website

    1. Kauer H

    (2020) Fauci wants people to know that one of lead scientists who developed the COVID-19 vaccine is a Black woman

    CNN. Accessed April 7, 2022.

    https://www.cnn.com/2020/12/09/us/african-american-scientists-vaccine-development-trnd/index.html
34. Book

    1. Kendi IX

    (2019)

    How To Be An Antiracist

    Random House.

    • Google Scholar
35. Website

    1. Kolata G

    (2021) Kati Kariko helped shield the world from the coronavirus

    New York Times. Accessed April 7, 2022.

    https://www.nytimes.com/2021/04/08/health/coronavirus-mrna-kariko.html
36. 1. Krause LA
    2. Harris SL

    (2019) Mental health in academia: get online to support wellbeing of graduate students

    eLife 8:e53178.

    https://doi.org/10.7554/eLife.53178

    • Google Scholar
37. 1. Kreps SE
    2. Kriner DL

    (2020) Model uncertainty, political contestation, and public trust in science: evidence from the COVID-19 pandemic

    Science Advances 6:eabd4563.

    https://doi.org/10.1126/sciadv.abd4563

    • PubMed
    • Google Scholar
38. 1. Kroenke K
    2. Spitzer RL
    3. Williams JB

    (2001) The PHQ-9: validity of a brief depression severity measure

    Journal of General Internal Medicine 16:606–613.

    https://doi.org/10.1046/j.1525-1497.2001.016009606.x

    • PubMed
    • Google Scholar
39. 1. Legha RK
    2. Miranda J

    (2020) An anti-racist approach to achieving mental health equity in clinical care

    The Psychiatric Clinics of North America 43:451–469.

    https://doi.org/10.1016/j.psc.2020.05.002

    • PubMed
    • Google Scholar
40. 1. Levecque K
    2. Anseel F
    3. De Beuckelaer A
    4. Van der Heyden J
    5. Gisle L

    (2017) Work organization and mental health problems in PhD students

    Research Policy 46:868–879.

    https://doi.org/10.1016/j.respol.2017.02.008

    • Google Scholar
41. Website

    1. Li W

    (2020) Racial disparities in COVID-19. Harvard University Special Edition Science Policy and Social Justice Blog

    Accessed April 7, 2022.

    http://sitn.hms.harvard.edu/flash/2020/racial-disparities-in-covid-19/
42. Book

    1. Limas JC

    (2021) Adaptation to Overexpression of Cyclin E in Epithelial Cells Doctoral Dissertation

    University of North Carolina.

    https://doi.org/10.17615/6cpy-mx59

    • Google Scholar
43. 1. Limas JC
    2. Corcoran LC
    3. Baker AN
    4. Cartaya AE
    5. Ayres ZJ

    (2022) The impact of research culture on mental health & diversity in STEM

    Chemistry 28:e202102957.

    https://doi.org/10.1002/chem.202102957

    • PubMed
    • Google Scholar
44. 1. Linley JL
    2. Nguyen DJ

    (2015) LGBTQ experiences in curricular contexts

    New Directions for Student Services 2015:41–53.

    https://doi.org/10.1002/ss.20144

    • Google Scholar
45. 1. Loissel E

    (2019) Mental health in academia: a question of support

    eLife 8:e52881.

    https://doi.org/10.7554/eLife.52881

    • Google Scholar
46. 1. Loissel E

    (2020) Mental health in academia: shedding light on those who provide support

    eLife 9:e64739.

    https://doi.org/10.7554/eLife.64739

    • Google Scholar
47. 1. Mackie SA
    2. Bates GW

    (2019) Contribution of the doctoral education environment to PhD candidates’ mental health problems: a scoping review

    Higher Education Research & Development 38:565–578.

    https://doi.org/10.1080/07294360.2018.1556620

    • Google Scholar
48. 1. McHugh RK
    2. Weiss RD

    (2019) Alcohol use disorder and depressive disorders

    Alcohol Research 40:arcr.v40.1.01.

    https://doi.org/10.35946/arcr.v40.1.01

    • PubMed
    • Google Scholar
49. Report

    1. Medley G
    2. Lipari RN
    3. Bose J
    4. Cribb DS
    5. Kroutil LS
    6. McHenry G

    (2016) Sexual orientation and estimates of adult substance use and mental health: Results from the 2015 National Survey on Drug Use and Health

    NSDUH Data Review.

    https://www.samhsa.gov/data/sites/default/files/NSDUH-SexualOrientation-2015/NSDUH-SexualOrientation-2015/NSDUH-SexualOrientation-2015.htm

    • Google Scholar
50. 1. Nagy GA
    2. Fang CM
    3. Hish AJ
    4. Kelly L
    5. Nicchitta CV
    6. Dzirasa K
    7. Rosenthal MZ

    (2019) Burnout and mental health problems in biomedical doctoral students

    CBE Life Sciences Education 18:ar27.

    https://doi.org/10.1187/cbe.18-09-0198

    • PubMed
    • Google Scholar
51. Website

    1. National Center for Transgender Equality

    (2016) The report of the 2015 U.S. Transgender Survey

    Accessed April 7, 2022.

    https://transequality.org/sites/default/files/docs/usts/USTS-Full-Report-Dec17.pdf
52. Website

    1. National Institute on Drug Abuse

    (2010) Comorbidity: addiction and other mental illnesses (NIH Publication Number 10-5771)

    NIDA Research Report Series. Accessed April 7, 2022.

    https://www.drugabuse.gov/sites/default/files/rrcomorbidity.pdf
53. Website

    1. National Institutes of Health

    (2018) RESCINDED - updated notice of NIH’s interest in diversity - notice number: NOT-OD-18-210

    Accessed July 16, 2018.

    https://grants.nih.gov/grants/guide/notice-files/NOT-OD-18-210.html
54. Website

    1. National Institutes of Health

    (2019) Notice of NIH’s interest in diversity (Notice Number: NOT-OD-20-031)

    Accessed November 22, 2019.

    https://grants.nih.gov/grants/guide/notice-files/NOT-OD-20-031.html
55. Website

    1. National Institutes of Health

    (2020) Populations underrepresented in the extramural scientific workforce

    Accessed April 7, 2020.

    https://diversity.nih.gov/about-us/population-underrepresented
56. Preprint

    1. Pathak EB
    2. Menard JM
    3. Garcia RB
    4. Salemi JL

    (2021) Social Class, Race/Ethnicity, and COVID-19 Mortality among Working Age Adults in the United States

    medRxiv.

    https://doi.org/10.1101/2021.11.23.21266759

    • Google Scholar
57. 1. Quek TT-C
    2. Tam WW-S
    3. Tran BX
    4. Zhang M
    5. Zhang Z
    6. Ho CS-H
    7. Ho RC-M

    (2019) The global prevalence of anxiety among medical students: a meta-analysis

    International Journal of Environmental Research and Public Health 16:E2735.

    https://doi.org/10.3390/ijerph16152735

    • PubMed
    • Google Scholar
58. Book

    1. Rollnick M

    (2015) Access of historically excluded groups to tertiary STEM education

    In: Gunstone R, editors. Encyclopedia of Science Education. Dordrecht: Springer. pp. 1–6.

    https://doi.org/10.1007/978-94-007-2150-0_425

    • Google Scholar
59. 1. Rotenstein LS
    2. Ramos MA
    3. Torre M
    4. Segal JB
    5. Peluso MJ
    6. Guille C
    7. Sen S
    8. Mata DA

    (2016) Prevalence of depression, depressive symptoms, and suicidal ideation among medical students: a systematic review and meta-analysis

    JAMA 316:2214–2236.

    https://doi.org/10.1001/jama.2016.17324

    • PubMed
    • Google Scholar
60. 1. Rothman S
    2. Gunturu S
    3. Korenis P

    (2020) The mental health impact of the COVID-19 epidemic on immigrants and racial and ethnic minorities

    QJM : Monthly Journal of the Association of Physicians 113:779–782.

    https://doi.org/10.1093/qjmed/hcaa203

    • PubMed
    • Google Scholar
61. 1. Salerno JP
    2. Devadas J
    3. Pease M
    4. Nketia B
    5. Fish JN

    (2020) Sexual and gender minority stress amid the COVID-19 pandemic: implications for LGBTQ young persons’ mental health and well-being

    Public Health Reports 135:721–727.

    https://doi.org/10.1177/0033354920954511

    • PubMed
    • Google Scholar
62. Report

    1. SAMHSA

    (2020) 2019 national survey on drug use and health: lesbian, gay, and bisexual (LGB) adults

    SAMHSA.

    https://www.samhsa.gov/data/report/2019-nsduh-lesbian-gay-bisexual-lgb-adults

    • Google Scholar
63. Report

    1. SAMHSA

    (2021) Key substance use and mental health indicators in the United States: results from the 2020 national survey on drug use and health

    SAMHSA.

    https://www.samhsa.gov/data/sites/default/files/reports/rpt35325/NSDUHFFRPDFWHTMLFiles2020/2020NSDUHFFR1PDFW102121.pdf

    • Google Scholar
64. 1. Seppälä EM
    2. Bradley C
    3. Moeller J
    4. Harouni L
    5. Nandamudi D
    6. Brackett MA

    (2020) Promoting mental health and psychological thriving in university students: a randomized controlled trial of three well-being interventions

    Frontiers in Psychiatry 11:590.

    https://doi.org/10.3389/fpsyt.2020.00590

    • PubMed
    • Google Scholar
65. 1. Skinner HA

    (1982) The drug abuse screening test

    Addictive Behaviors 7:363–371.

    https://doi.org/10.1016/0306-4603(82)90005-3

    • PubMed
    • Google Scholar
66. 1. Slavin SJ
    2. Schindler DL
    3. Chibnall JT

    (2014) Medical student mental health 3.0: improving student wellness through curricular changes

    Academic Medicine 89:573–577.

    https://doi.org/10.1097/ACM.0000000000000166

    • PubMed
    • Google Scholar
67. 1. Smith JP
    2. Randall CL

    (2012)

    Anxiety and alcohol use disorders: comorbidity and treatment considerations

    Alcohol Research 34:414–431.

    • PubMed
    • Google Scholar
68. 1. Spitzer RL
    2. Kroenke K
    3. Williams JBW
    4. Löwe B

    (2006) A brief measure for assessing generalized anxiety disorder: the GAD-7

    Archives of Internal Medicine 166:1092–1097.

    https://doi.org/10.1001/archinte.166.10.1092

    • PubMed
    • Google Scholar
69. Website

    1. Terlizzi EP
    2. Villarroel MA

    (2020) Symptoms of generalized anxiety disorder among adults: United States

    Accessed April 5, 2022.

    https://www.cdc.gov/nchs/products/databriefs/db378.htm
70. 1. Toews JA
    2. Lockyer JM
    3. Dobson DJ
    4. Brownell AK

    (1993) Stress among residents, medical students, and graduate science (MSc/PhD) students

    Academic Medicine 68:S46–S48.

    https://doi.org/10.1097/00001888-199310000-00042

    • PubMed
    • Google Scholar
71. Website

    1. UC Berkeley Graduate Assembly

    (2014) Graduate student happiness and well-being report

    Accessed April 7, 2022.

    http://ga.berkeley.edu/wp-content/uploads/2015/04/wellbeingreport_2014.pdf
72. Website

    1. UNC Health and UNC School of Medicine

    (2020) Teaching COVID-19’s real-time lessons

    Accessed April 7, 2022.

    https://news.unchealthcare.org/2020/06/teaching-covid-192019s-real-time-lessons/
73. Website

    1. UNC School of Medicine Office of Diversity Equity and Inclusion

    (2021) ODEI Annual Report

    Accessed August 8, 2022.

    https://www.med.unc.edu/inclusion/about/odei-reports/
74. 1. Webb Hooper M
    2. Nápoles AM
    3. Pérez-Stable EJ

    (2020) COVID-19 and racial/ethnic disparities

    JAMA 323:2466–2467.

    https://doi.org/10.1001/jama.2020.8598

    • PubMed
    • Google Scholar
75. Website

    1. World Health Organization

    (2001) AUDIT: The alcohol use disorders identification test: guidelines for use in primary health care (no.WHO/MSD/MSB/01.6 a)

    Accessed April 7, 2022.

    http://whqlibdoc.who.int/hq/2001/WHO_MSD_MSB_01.6a.pdf
76. 1. Xiong J
    2. Lipsitz O
    3. Nasri F
    4. Lui LMW
    5. Gill H
    6. Phan L
    7. Chen-Li D
    8. Iacobucci M
    9. Ho R
    10. Majeed A
    11. McIntyre RS

    (2020) Impact of COVID-19 pandemic on mental health in the general population: a systematic review

    Journal of Affective Disorders 277:55–64.

    https://doi.org/10.1016/j.jad.2020.08.001

    • PubMed
    • Google Scholar
77. Website

    1. Yam K

    (2021a) Anti-Asian hate crimes increased by nearly 150% in 2020, mostly in NY and LA, new report says

    NBC News. Accessed April 7, 2022.

    https://www.nbcnews.com/news/asian-america/anti-asian-hate-crimes-increased-nearly-150-2020-mostly-n-n1260264
78. Website

    1. Yam K

    (2021b) Anti-Asian hate incident reports nearly doubled in March, new data says

    NBC News. Accessed April 7, 2022.

    https://www.nbcnews.com/news/asian-america/anti-asian-hate-incident-reports-nearly-doubled-march-new-data-n1266980
79. 1. Yudko E
    2. Lozhkina O
    3. Fouts A

    (2007) A comprehensive review of the psychometric properties of the drug abuse screening test

    Journal of Substance Abuse Treatment 32:189–198.

    https://doi.org/10.1016/j.jsat.2006.08.002

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Mental health disparities among biomedical trainees during COVID-19 and racial inequity" to eLife for consideration as a Feature Article.

Four peer reviewers have reviewed your article, and the evaluation has been overseen by two members of the eLife Features Team (Elsa Loissel and Peter Rodgers). The following individuals involved in reviewing your submission have agreed to reveal their identity: Marcus Lambert and Amanda Haage.

The reviewers and editors have discussed the reviews and we have drafted this decision letter to help you prepare a revised submission.

Please also note that if your manuscript is accepted, we would like to publish some of the relevant code with the article: please let me know if this will be a problem.

In addition, the eLife Features Editor may also contact you separately about some editorial issues that you will need to address.

Summary:

The article aims to explore the wellbeing of medical students and biomedical PhD students at an R1 institution using self-reported markers of mental health across race/ethnicity. This work was conducted over two years, one of which was marked by the COVID-19 pandemic and a wave of racial and civil unrest. This manuscript is important and timely for a broad biomedical audience, highlighting major concerns around the prevalence of mental health challenges, especially in biomedical PhD and underrepresented students.

However, several points need to be addressed to make the article suitable for publication. In particular, the reviewers were concerned with some aspects of the methodology and the discussion.

Essential revisions:

Comparing medical and PhD students:

1 – Please consider discussing in the introduction (and the discussion) why there may be a difference between medical and biomedical PhD students. In particular, please expand on:

a. The level of institutional support provided by the respective programs: For instance, what are the mental health resources provided to medical and PhD students? Are they the same and are they easily accessible e.g. paid for? facilitated/triaged? on-campus?

b. The awareness of the two populations of their mental health (one might argue that medical students are trained to recognize certain signs and symptoms).

c. The differences that are intrinsic to their training: for example, those familiar with the medical student training process might suggest more acute stressors but more structured support due to LCME requirements.

d. The response to the COVID pandemic from the two different programs: for instance, the curriculum and/or level of support changed for the medical students in 2020, which is highlighted in the discussion. However, this is a big limitation in interpreting the data (comparatively between MD and PhD students). This can be viewed as an intervention that the PhD students did not have.

2 – Following that thought: medical and graduate students have extremely different training paradigms and, in this institution, medical students received an intervention that could have mitigated the effects of the pandemic on their mental health. Therefore, please consider whether directly comparing medical to graduate students is relevant, and if so, how to do so: for example, the bar graphs side by side are fine, but the chi-sq tests may be an unfair comparison. Or at least, please discuss the caveat of the different training paradigms and levels of support provided throughout the 2020 year.

3 – The reason why PhD students have worse PHQ9 outcomes over time comparatively is because the medical students show a decrease from 2019 to 2020. Could you please report whether you have looked at the trends across the medical student years (1 through 4), and justify if not.

Discussing the sample, and how it was coded:

4 – Line 158: Please detail the criteria for sample invitations (e.g. are there any inclusion or exclusion criteria for the study, and if so, are they the same for the 2019 and 2020 surveys?).

5 – Line 158: Please clarify whether the respondents (n=431, Fall 2020; n=526, Fall 2019) are a representative sample that closely matches the characteristics of your population. If not, please address this selection bias in the discussion.

6 – Line 166-167. "Middle Eastern", while an important subcategory, is not a part of the NIH definition of underrepresented groups. If you would like to use the NIH guidelines, please remove this subcategory from the UR coding. Alternatively, please directly justify the coding of "Middle Eastern" as underrepresented. Please see revisions #13 and #14 for further discussion about NIH guidelines.

7 – Please ensure that the in-text reported percentages and N values are easy to understand, as they are quite unclear in the current version. This is particularly the case in the "participants" section of the methods and figure legends, where it is unclear what year each number represents.

Methodological and statistical issues:

8 – The data collected in 2019 and 2020 was combined for data analysis, suggesting that some participants may have participated in both years. Logistic regression requires the observations to be independent of each other: that is, the observations should not come from repeated measurements. This could bias/inflate the results – the larger number of repeated respondents, the higher chance to obtain biased results – and it therefore needs to be carefully justified and addressed. If you can identify repeated respondents, please report how many, and justify why you think you can include them in the analysis: in addition, please consider exploring whether your findings stay the same if you take out those repeated respondents, and/or use a repeated measures design to compare the difference in mental health between 2019 and 2020. If you are not able to identify repeated respondents, please discuss this limitation.

9 – The mental health outcomes are sensitive to recent events and affected by many factors/variables (e.g., variables related to the pandemic, years in the degree, disciplines, etc.): it is possible that the mental health disparities you report might be no longer statistically significant once these covariates are included in the analysis. Please either include these covariates in your logistic regression, or explain in the manuscript why they were not included.

10 – The collected mental health measures could be treated as continuous variables, which could offer better variabilities for statistical modelling. Please clarify why you decided to simplify the scale (0 and 1) to run logistic regression analyses (binary outcome) rather than multiple regressions (continuous outcome). If you are not able to provide the reason, please make sure that both approaches (logistic regression and multiple regression) yield consistent findings.

Additional analyses and discussion points pertaining to intersectionality approaches:

11 – Please consider conducting additional analyses to explore whether they were any differences across gender identity. The N values seem similar to the other populations you have chosen to compare, and this would significantly increase the impact of this study. Otherwise, please justify why this variable was not explored.

12 – If feasible, please adopt an intersectionality approach, in particular by examining how gender AND race/ethnicity impact your findings. Otherwise, please justify why these analyses were not performed, and strengthen your discussion around intersectionality.

13 – While Asians are included as well-represented according to NIH standards, the percentage of Asian-identifying individuals in the sample (at 16% and 9%) clearly demonstrates that they are not. This is discussed at the end, but please also include these comments in the Results section and strengthen the language. In addition, please consider comparing the experiences of Asian and White students, or, if you are concerned with anonymity/statistical power, please consider comparing WR vs. UR where "Asian" is moved from WR to UR, as this would allow the reader to understand if coding "Asian" as WR vs. UR makes a difference (this does not necessarily need to be in the main part of the paper). Alternatively, please justify why these analyses were not performed.

14 – On that note: although discussed at the end, please provide a more nuanced discussion of race and ethnicity, and strengthen your stance. In particular, while it is understandable that you want to use NIH standards, please acknowledge more overtly that race is a social construct (so as to not reinforce the notion that these categories may be biological); and please discuss in more depth how the NIH categories may be lacking and not best represent a marginalized identity in science.

Strengthening certain claims

15 – Please provide evidence for the following statements:

a – Line 40 ("biomedical graduate programs are exposed to unique stressors by the nature of their training"): please consider moving up certain citations (e.g. Hazell et al., 2020 and Mackie and Bates 2019) to support this affirmation. Please also consider replacing the word "unique" as long hours, pressure to produce, and unsupportive mentors may not necessarily support a "uniquely" stressful situation.

b – Line 93 ("Trainees of color, especially people identifying as Black, Indigenous, and people of color (BIPOC), may also be affected by heightened concerns about individual, community, and family health."): While covid disparities are certainly real, and likely at play, please provide supporting evidence for this affirmation.

c – Line 340 ("We also speculate that being in a medical training program with obvious direct impact during a time when medical professionals were publicly celebrated had a positive effect on MD student mental health.") This discussion of MDs being "publicly celebrated" and "heroic" and able to connect their training to the crisis compared to PhD lacks evidence, particularly as we see protests outside of hospitals continue.

d – Line 343 ("Furthermore, medical students were able to continue some form of their training remotely, maintain some social contact and peer support networks through academic training, and avoid graduation delays.") Please provide citations for this statement, and clarify whether this is institution-specific.

e – Line 356 ("Biomedical PhD student training was severely curtailed in the 2020 spring months…"). Please clarify if this refers to specific measures at your institution.

f – Line 410 ("the mental health crisis in graduate biomedical education has only recently received considerable attention from the laboratory research community."): Please provide evidence for this statement.

Essential revisions:

Comparing medical and PhD students:

1 – Please consider discussing in the introduction (and the discussion) why there may be a difference between medical and biomedical PhD students. In particular, please expand on:

a. The level of institutional support provided by the respective programs: For instance, what are the mental health resources provided to medical and PhD students? Are they the same and are they easily accessible e.g. paid for? facilitated/triaged? on-campus?

b. The awareness of the two populations of their mental health (one might argue that medical students are trained to recognize certain signs and symptoms).

c. The differences that are intrinsic to their training: for example, those familiar with the medical student training process might suggest more acute stressors but more structured support due to LCME requirements.

d. The response to the COVID pandemic from the two different programs: for instance, the curriculum and/or level of support changed for the medical students in 2020, which is highlighted in the discussion. However, this is a big limitation in interpreting the data (comparatively between MD and PhD students). This can be viewed as an intervention that the PhD students did not have.

We thank the reviewers for this important point and have added the following to the discussion to address these valid critiques:

“In contrast, PhD students continued work at best, or experienced forced absence from the lab, with no specific intervention addressing emerging COVID knowledge or mental health coping strategies.”

“Furthermore, accreditation requirements for MD training provide a more structured support system for medical student mental health and wellness compared with PhD student supports. Given the difference in support offered due to enforced standards, this makes MD support services more easily accessible and freely available (no cost, and/or available funding to cover any out-of-pocket mental health medical expenses). Whereas both MD and PhD have availability of limited initial contact of facilitated and triaged on-campus resources, MDs have access to 1.5 FTE dedicated support for 800 students (e.g., 533:1 student-counselor ratio) whereas PhDs have only.5 FTE dedicated support for 600 students (e.g., 1200:1 student-counselor ratio). The combination of the more robust support for MD students with the removal of exposure to acute stressors often encountered in clinical rotations may have magnified this effect.”

2 – Following that thought: medical and graduate students have extremely different training paradigms and, in this institution, medical students received an intervention that could have mitigated the effects of the pandemic on their mental health. Therefore, please consider whether directly comparing medical to graduate students is relevant, and if so, how to do so: for example, the bar graphs side by side are fine, but the chi-sq tests may be an unfair comparison. Or at least, please discuss the caveat of the different training paradigms and levels of support provided throughout the 2020 year.

We agree that the nuanced discussion of differences between the two populations is very important, and brings to attention the point that adjustment of the training environment (e.g., MD intervention) can in fact positively impact mental health even amidst difficult times. We have replaced the post hoc chi-squared tests with a more complex logistic regression model that allows for statistical control of differences between the two populations, including marginal means and conditional odds ratios to expresses in terms of differences in each population, while simultaneously controlling for the other variables’ impact.

3 – The reason why PhD students have worse PHQ9 outcomes over time comparatively is because the medical students show a decrease from 2019 to 2020. Could you please report whether you have looked at the trends across the medical student years (1 through 4), and justify if not.

We agree that analysis by year in program would be important to include in future work; however, in our sample this was a demographic question that was not completed by a sizeable portion of respondents, which would reduce our ability to complete the main analysis if we included it. We also have concerns about including variables with a significant amount of missing data as this would potentially not be representative of the sample or population.

Discussing the sample, and how it was coded:

4 – Line 158: Please detail the criteria for sample invitations (e.g. are there any inclusion or exclusion criteria for the study, and if so, are they the same for the 2019 and 2020 surveys?).

The same listservs were used each year to recruit School of Medicine trainees in Medical Education (MD students) and Graduate Education (biomedical PhDs), as described in Methods. A standard IRB-approved invitation included a brief paragraph explaining the survey, signed by the appropriate administrator (e.g., SOM Wellness Coach). There were no formal inclusion or exclusion criteria in either year other than being enrolled in either the MD or PhD programs, and anyone included on the SOM trainee listservs was invited to respond.

5 – Line 158: Please clarify whether the respondents (n=431, Fall 2020; n=526, Fall 2019) are a representative sample that closely matches the characteristics of your population. If not, please address this selection bias in the discussion.

Unfortunately, we can’t directly assess how well our samples were representative of their respective populations, the following has been added in limitations to address this:

“Due to anonymous data collection and optional questions to protect participant identities, we cannot assess granular response rates by demographic characteristics. Future studies could be completed with a controlled sample matched with participant identifiers to know who in the sample is and isn’t responding, as this could impact findings.”

6 – Line 166-167. "Middle Eastern", while an important subcategory, is not a part of the NIH definition of underrepresented groups. If you would like to use the NIH guidelines, please remove this subcategory from the UR coding. Alternatively, please directly justify the coding of "Middle Eastern" as underrepresented. Please see revisions #13 and #14 for further discussion about NIH guidelines.

Thank you for pointing out the limitations of NIH definition, here and elsewhere – we have further explained the reasoning for inclusion of Middle Eastern and expanded the discussion of why this category was included. In addition, we adjusted the terminology to better reflect the categories included in our analyses to include identities that have been historically excluded. This category has been retitled accordingly to, “historically excluded (HE)” as reflected in science by race/ethnicity and gender (HE gender discussed elsewhere). Please also see other author responses for a summary; as well as expanded responses about marginalized identity, gender, intersectionality, Asian identity, and an expanded discussion of NIH definition along with limitations thereof.

The following has been added:

“Further, the stress experienced by Asian Americans as scapegoats of the pandemic and targets of anti-Asian violence may also affect the mental health of Asian American students within academia.” (Introduction)

“In the primary analysis, mental health data was analyzed by type of training program (MD vs. PhD), year (2019 vs. 2020), and historically excluded (HE) populations by race/ethnicity (HE-RE vs. NHE-RE) and by gender (HE-G women vs. NHE-G men). A post hoc analysis included the primary variables as wells as historically excluded by sexual orientation (HE-SO LGBQ+ vs. NHE-SO non-LGBQ+).” (Methods)

“in accordance with National Institutes of Health definitions of underrepresented in the biomedical, clinical, behavioral, and social sciences (NIH, 2020). Historically, Asian Americans have also faced exclusion from American culture as evidenced by Japanese Internment and the Chinese Exclusion Act. However, Asian Americans have been well-represented in the sciences and thus are included as NHE-RE for analysis. Additionally, Middle Eastern is not included in the US Census or NIH definitions as underrepresented; however, this identity group is included as HE-RE due to the marginalization especially over the past two decades in the United States following 9/11/2001.” (Methods-Participants)

7 – Please ensure that the in-text reported percentages and N values are easy to understand, as they are quite unclear in the current version. This is particularly the case in the "participants" section of the methods and figure legends, where it is unclear what year each number represents.

We have simplified the figures and legends in favor of an expanded unified description in the Methods, as we recognize that the prior version may be confusing. Previously, the chi-squared post hoc analysis relied on specific n and degrees of freedom for each test that impacted interpretation (no longer crucial for the current analyses). Especially given the updated modeling with interaction terms used to replace the previous chi-squared post hoc tests, the exact n per each subgroup is less important and instead the odds ratios and p-values are the primary focus. Accordingly, to reflect this change, Figure 1 has been simplified per the editors and reviewers’ suggestions; Figure 2 and Figure 3 (new Figure 3 added to visualize effects identified in the post hoc analysis) also now include percentages embedded in the figures rather than embedding these details in the legends.

Methodological and statistical issues:

8 – The data collected in 2019 and 2020 was combined for data analysis, suggesting that some participants may have participated in both years. Logistic regression requires the observations to be independent of each other: that is, the observations should not come from repeated measurements. This could bias/inflate the results – the larger number of repeated respondents, the higher chance to obtain biased results – and it therefore needs to be carefully justified and addressed. If you can identify repeated respondents, please report how many, and justify why you think you can include them in the analysis: in addition, please consider exploring whether your findings stay the same if you take out those repeated respondents, and/or use a repeated measures design to compare the difference in mental health between 2019 and 2020. If you are not able to identify repeated respondents, please discuss this limitation.

Due to data being collected anonymously, we cannot confirm if there were any repeated participants across the year timepoints; we recognize that this is a limitation, and have added the following to the discussion to acknowledge this:

“We recognize that if a large number of participants responded in both samples this could be problematic for assumptions of independence for use of parametric statistics and the Logistic Regression analysis used specifically, hence results should be interpreted with caution. Because we collected data anonymously to protect respondents’ privacy, we cannot assess to what extent there may have been respondents who participated in both years. If we did have identifiable data and were able to use a repeated-measures design, this would be preferable to reduce error variance. To the extent that significant findings were achieved even with the increased error variance inherent in a between-subjects design, these results likely had large enough effect sizes to be identified even given the loss of power from using a between-subjects design rather than the more sensitive within-subjects design that would be preferable.”

9 – The mental health outcomes are sensitive to recent events and affected by many factors/variables (e.g., variables related to the pandemic, years in the degree, disciplines, etc.): it is possible that the mental health disparities you report might be no longer statistically significant once these covariates are included in the analysis. Please either include these covariates in your logistic regression, or explain in the manuscript why they were not included.

We did not collect pandemic-specific data as we used the same annual survey questions from the previous year, future evaluations should ask and control for these variables. Limitations for being able to include years in training is discussed elsewhere (see additional author responses to year in training variable). We did not collet specialty or departmental affiliations to protect anonymity, to increase comfort with responding, and to maximize response rates. Furthermore, disciplinary clusters were deemed homogeneous enough to be considered in the two major categories of biomedical PhD training and medical MD training.

“Additional limitations included our lack of ability to control for other possible variables of interest such as pandemic-specific factors, years in training, and departmental affiliation or specialty area. Pandemic-specific questions were not asked because we used the same annual survey questions from the previous year to maintain comparable responses. Department and specialty information were not asked in order to protect anonymity, to increase comfort with responding, and to maximize response rates. Yet, populations were purposefully defined by reasonably homogenous training experiences into the two major clusters of interest: biomedical PhD training and MD training. Optional demographic questions were not completed for number of years in training for a large portion of the sample, limiting our ability to include this in the analysis. Future work should consider controlling for as many of these variables as possible.”

10 – The collected mental health measures could be treated as continuous variables, which could offer better variabilities for statistical modelling. Please clarify why you decided to simplify the scale (0 and 1) to run logistic regression analyses (binary outcome) rather than multiple regressions (continuous outcome). If you are not able to provide the reason, please make sure that both approaches (logistic regression and multiple regression) yield consistent findings.

The research question of interest was best captured by a bivariate designation of symptomatic versus asymptomatic (or problematic use versus non-problematic use). Hence, we approached the question from a bivariate logistic regression approach to answer the question accordingly. We felt this was the most appropriate model to answer the research question, and could be best interpreted to determined when respondents were functionally affected by their mental health; hence our decision to define it as such in the statistical modeling. Severity of symptoms at a more granular level, especially relying on self-report approaches rather than clinical diagnoses, posed several limitations that we felt made it a less appropriate approach to the data. Furthermore, for implications and policy purposes, we feel that the defined research question best illustrates the sweeping problem of mental health in graduate biomedical and medical education.

Additional analyses and discussion points pertaining to intersectionality approaches:

11 – Please consider conducting additional analyses to explore whether they were any differences across gender identity. The N values seem similar to the other populations you have chosen to compare, and this would significantly increase the impact of this study. Otherwise, please justify why this variable was not explored.

We agree that including gender as variable would add to the intersectional applicability of our model, and hence have added it as a control variable for all primary analyses.

We also decided that if evaluating the impact of gender were to be added to the model, we would be remiss to not also include sexual orientation status if we had a large enough sample size, as the intersectionality of gender and sexual identity can interact with race/ethnicity, trainee type, and year as well. However, we were also cautious about inadvertently reducing power to evaluate main effects beyond the primary research questions in order to ensure enough power to evaluate the primary intersectional identities (race/ethnicity, gender, and trainee type) by year. Hence, we added a post hoc analysis that accounted for both gender and LGBQ status using an intersectional approach.

We have added sections in the results and discussion describing the revised primary analysis (adding gender using a more sophisticated Logistic Regression model including interaction terms, conditional odds ratios, and marginal means to replace the more simplistic Chi-Squared comparisons previously displayed) as well as the post hoc model which includes LGBQ identity. Along with these substantive changes to the analyses, we have included justification and precedent for including both gender and sexual orientation based on previous studies. Transgender identity could not be included due to the small sample size, but should be included in future work and this limitation is also addressed in the manuscript.

Relatedly, to avoid reductions in power we also ran post hoc tests to further examine the impact of Asian social identities in a more granular way, and have reported the new results accordingly (using the model that also controls for gender).

12 – If feasible, please adopt an intersectionality approach, in particular by examining how gender AND race/ethnicity impact your findings. Otherwise, please justify why these analyses were not performed, and strengthen your discussion around intersectionality.

We have adopted an intersectional approach, with additional citations to justify this in the introduction as well as a deeper dive in the discussion. Thank you for this suggestion, we believe it has strengthened our approach and the applicability of the findings greatly.

13 – While Asians are included as well-represented according to NIH standards, the percentage of Asian-identifying individuals in the sample (at 16% and 9%) clearly demonstrates that they are not. This is discussed at the end, but please also include these comments in the Results section and strengthen the language. In addition, please consider comparing the experiences of Asian and White students, or, if you are concerned with anonymity/statistical power, please consider comparing WR vs. UR where "Asian" is moved from WR to UR, as this would allow the reader to understand if coding "Asian" as WR vs. UR makes a difference (this does not necessarily need to be in the main part of the paper). Alternatively, please justify why these analyses were not performed.

Thank you for this suggestion. We agree that a more granular investigation of the impacts on Asian participants is useful, despite having a small sample once dividing HE-RE into 3 classifications; hence have included a supplemental post hoc analysis as mentioned in other author responses which still allows us to address the questions while retaining power for the primary analysis. The following has been added:

“People who identified of Asian descent generally scored better than other NHE or HE groups, indicating fewer mental health challenges on all four outcome variables and the three suicidal ideation variables. However, conditional odds rations based on these models indicated some changes by year that may be associated with the negative psychosocial and environmental factors many Asian Americans experienced between 2019-2020. Accounting for the effect of race (3-level) while controlling for year, MD/PhD, and gender in 2020 compared with 2019, Asian participants showed greater mean depression scores (p<.03) and anxiety scores (p<.04); there was no change in either alcohol use (p=.07, though marginal) or drug use (p=.28). There were no significant changes in conditional probabilities for suicidal ideation of Asian participants during that time. Future directions should include examination of the intersectionality of these and other identity groups, and systemic barriers that each may encounter differentially. Follow-up studies should explore the impact of racism on Asian Americans and contrast Asian American experiences with White American experiences when reporting HE-RE data.”

14 – On that note: although discussed at the end, please provide a more nuanced discussion of race and ethnicity, and strengthen your stance. In particular, while it is understandable that you want to use NIH standards, please acknowledge more overtly that race is a social construct (so as to not reinforce the notion that these categories may be biological); and please discuss in more depth how the NIH categories may be lacking and not best represent a marginalized identity in science.

Thank you for these valuable suggestions, which we have leaned on heavily to reframe much of our discussion and additional analyses. We believe this has strengthened our findings and our ability to interpret implications more broadly. We have: (a) expanded our discussion of intersectionality; (b) added more nuanced analysis controlling for a wider range of identities (including the addition of gender to all of our primary models; c) included a post hoc analysis and discussion of Asian identity as well as sexual orientation); (d) re-envisioned the terminology used to Historically Excluded (HE) in science (HE-RE and HE-G); (e) we have expanded our discussion of limitations to the NIH definitions of WR/UR; and f) we have explicitly noted the social derivation of race.

The following has been added to the discussion (see NIH definition expanded discussion in other author responses):

“We recognize [both that race is a social construct], and that other social identity groups beyond race and ethnicity (Rothman, Gunturu, Korenis, 2020) may also experience inequitable impacts of COVID-19 and similar structural biases may negatively impact mental health for other reasons (Hunt, 2020). (Discussion)

“In fact, the NIH has recently expanded its definition of underrepresented in science to include women, people with disabilities, first generation college students, and those from disadvantaged backgrounds, including people who have experienced or are experiencing homelessness, foster care participation, recipients of free and reduced lunch, Pell grants, SNAP or WIC, and those who grew up in a low-income or rural areas (NIH, 2020; NIH, 2018, rescinded and replaced in 2019; NIH, 2019). This expanded definition is still flawed as some groups are still excluded from recognition and inclusion as societal norms and practices shift over time. For instance, people of middle eastern descent who have faced heightened discrimination, especially following 9/11/2001.” (Discussion)

NIH definition limitations are discussed in a number of other points in the response as well (please see also other author responses regardig Middle Eastern and Asian identities, gender, intersectionality, and NIH definition limitations).

Furthermore, particularly in recent American history, it is noteworthy that over the past 20 years many groups have been systematically marginalized during different periods of time. Future definitions of historically excluded groups may also want to consider the impact of current events with politics and their intersection with discrimination against specific social identities. We must acknowledge that science is not immune from such effects, which can drive out a talented, diverse pool of future scientists in training.

Strengthening certain claims

15 – Please provide evidence for the following statements:

a – Line 40 ("biomedical graduate programs are exposed to unique stressors by the nature of their training"): please consider moving up certain citations (e.g. Hazell et al., 2020 and Mackie and Bates 2019) to support this affirmation. Please also consider replacing the word "unique" as long hours, pressure to produce, and unsupportive mentors may not necessarily support a "uniquely" stressful situation.

b – Line 93 ("Trainees of color, especially people identifying as Black, Indigenous, and people of color (BIPOC), may also be affected by heightened concerns about individual, community, and family health."): While covid disparities are certainly real, and likely at play, please provide supporting evidence for this affirmation.

c – Line 340 ("We also speculate that being in a medical training program with obvious direct impact during a time when medical professionals were publicly celebrated had a positive effect on MD student mental health.") This discussion of MDs being "publicly celebrated" and "heroic" and able to connect their training to the crisis compared to PhD lacks evidence, particularly as we see protests outside of hospitals continue.

d – Line 343 ("Furthermore, medical students were able to continue some form of their training remotely, maintain some social contact and peer support networks through academic training, and avoid graduation delays.") Please provide citations for this statement, and clarify whether this is institution-specific.

e – Line 356 ("Biomedical PhD student training was severely curtailed in the 2020 spring months…"). Please clarify if this refers to specific measures at your institution.

f – Line 410 ("the mental health crisis in graduate biomedical education has only recently received considerable attention from the laboratory research community."): Please provide evidence for this statement.

Unique has been replaced as follows:

“Considering that students in medical school and in biomedical graduate programs are exposed to a multitude of stressors by the nature of their training including long hours, pressure to produce, and unsupportive mentors…”

B – We have reframed the terminology to center on Historically Excluded instead of BIPOC; nonetheless, we have also included an additional reference to support the extension of the COVID-19 impacts to disproportionately impact BIPOC individuals, communities, and family health.

We have also added the following reference, examining HE-RE working class inequities due to COVID-19:

Pathak, E. B., Menard, J. M., Garcia, R. B., and Salemi, J. L. (2021). Social Class, Race/Ethnicity, and COVID-19 Mortality Among Working Age Adults in the United States. medRxiv.

Feldman, J. M., and Bassett, M. T. (2021). Variation in COVID-19 Mortality in the US by Race and Ethnicity and Educational Attainment. JAMA network open, 4(11), e2135967-e2135967.

Limas, J.C. (2021). Adaptation to Overexpression of Cyclin E in Epithelial Cells. (Publication No. 2616907910) [Doctoral dissertation, University of North Carolina at Chapel Hill]. ProQuest Dissertations and Theses Global.

Limas, J. C., Corcoran, L. C., Baker, A. N., Cartaya, A. E., and Ayres, Z. J. (2022). The Impact of Research Culture on Mental Health and Diversity in STEM. Chemistry–A European Journal, e202102957.

“While our students themselves may be classified as associated with some lower risk groups for direct effects COVID due to their levels of education, BIPOC students disproportionately hail from working-class families and identify as first-generation college graduates. Hence, students historically excluded by race and ethnicity (HE-RE; e.g., those identifying as BIPOC) may experience disproportionate impact to their families, in that their families have been most at risk for mortality from COVID during 2020 within each social class (Pathak et al., 2021). Similarly, highest risk of age-adjusted mortality during COVID was identified for Hawaiian and Other Pacific Islander, American Indian or Alaska Native, and Latinx or Hispanic people (Black, Feldman and Bassett, 2021). An illustrative example includes a PhD sharing instances of her own experience as a Mexican-American woman PhD trainee during this time (Limas, 2021).”

C – We acknowledge the very real impact of anti-medical establishment influences as well, and have added the following to address this oversight: “Even so, this was not a ubiquitous experience, as simultaneously continued protests and divisiveness regarding vaccination were also pervasive across the country, which could ameliorate this effect (e.g., Bogel-Burroughs, 2020).”

Bogel-Burroughs, N. (2020). Antivaccination activists are growing force at virus protests. The New York Times. May, 2.

D and E – We have clarified these statements, and qualified them by specifying, “at our institution.”

F – The following has been added to clarify:

“Whereas medical schools have had mental health recommendations in place since 1992 (AAMC, 1992; enforced by Liaison Committee on Medical Education accreditation standards), biomedical PhD programs have had no such parallel policies regarding mental health provisions nationwide. Similarly, while there have been frequent calls for mental health research in higher education and there is already a robust body of literature decades old already in medical education (e.g., Slavin, 2016; a systematic review of medical education mental health evidence-based research included 195 relevant studies, Rotenstein et al., 2016) , evidence-based research on mental health research of PhDs has begun to emerge over the past five years (e.g., Levecque et al., 2017; Evans et al., 2018). Resources have been mobilized to evaluate and respond to the needs of medical students (e.g., 28 studies evaluating interventions, Wasson et al., 2016), which indeed still merit additional attention given high rates of depression, anxiety, and suicidality.”

The following additional references have also been added:

AAMC Executive Council (1992). Recommendations Regarding Health Services for Medical Students. Washington, DC: Association of American Medical Colleges

Available: https://www.aamc.org/professional-development/affinity-groups/gsa/health-services-recommendations

Wasson, L. T., Cusmano, A., Meli, L., Louh, I., Falzon, L., Hampsey, M., … and Davidson, K. W. (2016). Association between learning environment interventions and medical student well-being: a systematic review. Jama, 316(21), 2237-2252.

Author details

1. Allison Schad

   Allison Schad is in the Office of Medical Education, University of North Carolina at Chapel Hill School of Medicine, Chapel Hill, United States

   Contribution

   Conceptualization, Resources, Data curation, Formal analysis, Funding acquisition, Investigation, Visualization, Methodology, Writing – original draft, Project administration, Writing – review and editing

   Contributed equally with

   Rebekah L Layton

   For correspondence

   allison_schad@med.unc.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-6672-5202

2. Rebekah L Layton

   Rebekah L Layton is in the Office of Graduate Education, University of North Carolina at Chapel Hill School of Medicine, Chapel Hill, United States

   Contribution

   Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Investigation, Visualization, Methodology, Writing – original draft, Project administration, Writing – review and editing

   Contributed equally with

   Allison Schad

   For correspondence

   rlayton@unc.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-7113-1348

3. Debra Ragland

   Debra Ragland is in the Office of Graduate Education, University of North Carolina at Chapel Hill School of Medicine, Chapel Hill, United States

   Contribution

   Validation, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-9512-1988

4. Jeanette Gowen Cook

   Jeanette Gowen Cook is in the Office of Graduate Education and the Department of Biochemistry and Biophysics, University of North Carolina at Chapel Hill School of Medicine, Chapel Hill, United States

   Contribution

   Resources, Funding acquisition, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-0849-7405

• 1,849

  Page views

• 233

  Downloads

• 5

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.