Mental health and substance use among women and men at the intersections of identities and experiences of discrimination: insights from the intersectionality framework

Intersectionality theory highlights how one’s human experiences are constituted by mutually reinforcing interactions between different aspects of one’s identities, such as race, class, gender, and sexuality [1]. Intersectionality as an analytical concept for the social and health sciences has its roots in black feminist scholarship. In the seminal article “Mapping the margins: Intersectionality, identity politics, and violence against women of color,” Crenshaw pointed out that “the intersection of racism and sexism factors into Black women’s lives in ways that cannot be captured wholly by looking at the race or gender dimensions of those experiences separately” [2]. While Crenshaw focused mainly on the intersection of sex and race, she also highlighted the need to investigate aspects of one’s identity that extend beyond these two categories, as well as the ways in which these intersecting aspects create lived experiences embedded in structural systems of opportunities and oppression.

Crenshaw’s arguments are further extended in works by others such as Bowleg (2008; 2012) and Bauer (2014). For example, Bowleg asserted that intersectionality should be applied as an analytical theoretical framework as opposed to a traditional theory with operationalized and empirically testable variables [3, 4]. Additionally, the core tenet and starting point of intersectionality should be “multiple socially disadvantaged statuses,” or in other words, historically oppressed and marginalized populations such as racial or sexual minority populations [4]. In regard to modeling intersectionality through statistical methods, Bauer emphasized the need to distinguish between variables capturing intersecting identities (e.g., race or sexual orientation) and positions (e.g., racism or homophobia), as well as the need to structure statistical models properly to make differential effects visible across strata or groups [5]. While intersectionality arises from the more historical, interpretative, and qualitative perspectives in the social sciences [6], the abovementioned works have been fundamental in providing more concrete theoretical and methodological directions for quantitatively modeling intersectionality in public health research.

Recent years have seen a proliferation in the number of health studies using intersectionality as a guiding framework along with increasing appreciation of its potential for investigating social determinants of health [7]. Intersectionality has been explicitly applied in research examining disparities or differences in a variety of health beliefs, behaviors, and outcomes, including mental health [6, 8] and substance use [9]. In addition to examining the intersection of identity categories typically seen in public health, such as age, sex, race, sexual orientation, and socioeconomic status, recent studies have also incorporated other measures accounting for social phenomena that reflect power and oppression [10]. Intersections of different experiences of discrimination have also received increasing attention [11, 12].

Central to the intersectionality theory is the idea that multiple social identities at the micro-level (e.g., race, sex, or social class) are linked to macro- and structural-level inequalities (e.g., racism, sexism, and poverty) [3, 4]. This idea is shared with the minority stress theory. Minority stress theory posits that individuals with membership in a minority group will experience unfair treatment due to their group membership [13‐15]. Similar to intersectionality theory, minority stress theory hypothesizes that discrimination plays an important role in explaining health disparities between dominant and minority groups.

Several pathways have been proposed for how interpersonal discrimination negatively affects health outcomes and increases health risk behaviors. In regard to mental health, interpersonal and direct experiences with discrimination can lead to heightened vigilance, challenge one’s beliefs about fairness and justice, create internalized stigma towards oneself, and exacerbate physiological and psychological stress, all of which contribute to poorer mental health outcomes [14, 16‐18]. A meta-analysis of 110 studies found a significant negative correlation between experiences of perceived discrimination and mental health status [19].

Regarding substance use (e.g., alcohol, tobacco, and marijuana), experiences of discrimination can prompt adolescents and young adults to increase their affiliation with drug-using peers, which subsequently may lead to higher substance use [20]. In addition, individuals exposed to discrimination are more likely to endorse substance use as a coping mechanism [21, 22]. Moreover, psychological distress as a result of experiences of discrimination has also been shown to play a mediating role in the relationship between discrimination and substance use [23‐25].

Two interrelated challenges appear in the application of intersectionality in public health. The first issue is the tension between the intersectionality paradox and the multiple jeopardy perspective (also known as multiple disadvantages or multiple-hierarchy stratification approach). Bowleg emphasized the “intersectionality paradox” as one of the most noteworthy contributions of intersectionality theory to public health [4]. According to Bowleg, “low”-status social identities (e.g., racial or sexual minority) do not automatically equate disadvantages; rather, they intersect with “high”-status identities (e.g., high socioeconomic status) to produce differences in outcomes. An example of the “intersectionality paradox” could be seen in Rosenfield’s study on mental health at the intersection of sex, race, and class [8]. The study demonstrated that, among women with higher than a high school education, Black women have lower rates of depression than White; among women with lower education levels, there were no significant differences between the two groups. These different patterns for women were not indicated in Black versus White men in the study.

While the multiple jeopardy approach also investigates the impacts of multiple social identities and statuses, this perspective assumes that disadvantages accumulate in an added-burden or additive fashion. With additional minority statuses, individuals are assumed to be at risk for incrementally poorer health [14, 26‐28]. For example, at the intersections of race, sex, and sexual orientation, a woman who is Black and sexual minority is assumed to have worst health outcomes (i.e., three “low”-status identities), whereas a man who is White and heterosexual would have best health outcomes (i.e., three “high”-status identities). In recent years, scholars have identified discrepancies in findings for the multiple jeopardy and criticized this approach for oversimplifying social realities [8, 29], though there still remains a need for more empirical data to address these issues.

The tension between the intersectionality and multiple jeopardy approaches also points to a need to reevaluate measurements in public health and, in particular, the need to distinguish between identity and experience. Frequently, studies in this research area do not include measures for discrimination, and many studies still treat disadvantaged statuses or minority identities as identical to, or an approximation for, experiences of discrimination. For example, a commonly seen approach is when studies report a poorer health outcome or increased health risks for a minority group, they often hypothesize that the differences may be due to impacts of interpersonal discrimination, without actually including a measure for discrimination [30, 31]. While it is logical and consistent with minority stress theory to posit that individuals with minority statuses will face stigma due to their membership in the minority groups, it is problematic to assume that these two domains (identity and experience) are interchangeable or synonymous.

In light of the tensions of how intersectionality has been developed and applied in public health, this paper seeks to contribute to the field by investigating the following two questions: 1) Do associations between intersecting identities (i.e. race and sexual orientation) and mental health (depressive symptoms) and substance use (alcohol, tobacco, and marijuana) differ between men and women? and 2) How do single or intersecting self-reports of perceived racial and/or sexual orientation discrimination influence mental health and substance use outcomes for men and women? In answering these two questions, we compare results of assessing identities versus experiences of discrimination on health behaviors and outcomes. The goal of this research is to provide empirical evidence supporting the need to: 1) incorporate understanding of differing mental health and substance use outcomes based on intersections of identities and intersections of experiences of discrimination, and 2) distinguish between the two domains of identity and experience in public health research.

Data for the current study came from Project DECOY (Documenting Experiences with Cigarettes and Other Tobacco in Young Adults). The methods employed for sampling and recruitment for Project DECOY have been described elsewhere [32]. Briefly, this is a two-year, six-wave longitudinal cohort study that involved 3418 racially/ethnically diverse students (ages 18 to 25 years) from seven colleges and universities in Georgia. Schools are located in both rural and urban settings and include two public universities/colleges, two private universities, two community/technical colleges, and one historically black university. Our project was approved by the Emory University and ICF Institutional Review Boards (IRBs) as well as the IRBs of the participating colleges and universities. Data collection began in Fall 2014 and consisted of self-report assessments via an online survey every four months for two years (during Fall, Spring, and Summer).

The registrar’s office from each campus provided e-mail addresses for English-speaking students ages 18–25. We randomly selected 3000 email addresses from each of the three largest campuses, and emailed a census of students at the four smaller campuses with fewer than 3000 students. We met our sampling quota target in a short time interval (24 h at the private schools to seven days at the technical colleges) and enrolled between 12.0 to 59.4% of those approached at different campuses, and overall 22.9% (N = 3574/15,607). Seven days after initial recruitment and completion of the baseline survey, we asked participants to confirm their participation by clicking a “confirm” button included in an email sent to them. The email reiterated the tasks involved in the study and its timeline. Once participants clicked “confirm,” they were enrolled into the study and sent their first incentive in the form of a $30 gift card via email. The confirmation rate was 95.6% (N = 3418/3574). Our intent was to enroll participants who were engaged in email and were potentially more likely to be retained in the subsequent waves of the larger, multi-wave longitudinal project.

The current analyses examined data from Wave 5 of the study. Data were collected between April and May 2016, which took place around 1.5 years from the baseline (Wave 1) data collection. Wave 5 surveys were completed by 2690 participants (the retention rate from Wave 1 to Wave 5 is 78.7%). Since our research questions focused on participants who identified as Black or White, a total of 2315 participants who met this criterion were included in the analyses.

Data were taken from the baseline survey assessments of socio-demographic information and the Wave 5 assessment of experience of discrimination, depressive symptoms, and use of alcohol, tobacco, and marijuana within the past 30 days.

Simple logistic regressions were conducted to assess the association between race and experience of racial discrimination as well as the association between sexual orientation and experience of sexual orientation discrimination. We found that being Black was associated with reporting racial discrimination (odd ratio or OR = 7.46, p < 0.001), and being sexual minority was associated with reporting sexual orientation discrimination (OR = 42.67, p < 0.001). We also conducted bivariate analyses to examine sociodemographic characteristics and experience of discrimination in relation to the four outcomes of interest: 1) number of depressive symptoms; 2) number of days of alcohol use within the past 30 days; 3) use of any tobacco products within the past 30 days; and 4) use of marijuana within the past 30 days.

Multivariable linear and binary logistic regression models were used to identify correlates of the outcomes of interest. Given evidence on sex differences in depression and substance use behaviors (refer to Table 1), for each outcome, we stratified models by sex. We then constructed two different multivariable regression models (for a total of four models per outcome). For each outcome, Model 1 contained variables capturing intersecting identities (White and heterosexual, White and sexual minority, Black and heterosexual, or Black and sexual minority), and Model 2 contained variables capturing experience of discrimination (no experience, only racial discrimination, only sexual orientation discrimination, or both forms of discrimination). Key sociodemographic variables (age, highest parental education, and school type) were also entered into all models. In our data, we observed that 70.91% of Black students reported experiencing racial discrimination compared to 24.63% of White students; moreover, 2.85% of heterosexual students reported experiencing sexual orientation discrimination compared to 55.56% of sexual minority students. We also observed a high correlation between race and experience of racial discrimination as well as between sexual orientation and experience of sexual orientation discrimination (OR = 7.46 and OR = 42.67, respectively, with both p values < 0.001, as noted above). Therefore, we did not include race and experience of racial discrimination in the same model. We also did not include sexual orientation and experience of sexual orientation discrimination in the same model. Analyses were conducted using SAS 9.4.

In bivariate analyses, higher depressive symptoms was associated with being female, sexual minority, reporting racial discrimination, reporting sexual orientation discrimination, younger age, having parents with highest education being below a Bachelor’s degree, and school type.

Multivariable linear regression models (shown in Table 2) indicated that compared to White heterosexual women, White sexual minority women had higher depressive symptoms (beta or B = 3.16, p < 0.001). Compared to women who experienced no discrimination, women who experienced only racial discrimination (B = 1.57, p < 0.001), only sexual orientation discrimination (B = 2.59, p < 0.001), and both forms of discrimination (B = 3.63, p < 0.001) had higher depressive symptoms. For women, older age was associated with lower depressive symptoms.

Compared to White heterosexual men, White sexual minority men had higher depressive symptoms (B = 1.90, p = 0.01). Compared to men who experienced no discrimination, men who experienced only racial discrimination (B = 0.96, p = 0.03) and only sexual orientation discrimination (B = 3.78, p < 0.001) had higher depressive symptoms.

In bivariate analyses, higher number of days of alcohol use within the past 30 days was associated with being male, White, and sexual minority, as well as older age, having parents with highest education of a bachelor’s degree or above, and school type.

Multivariable linear regression models (shown in Table 3) indicated that, compared to White heterosexual women, White sexual minority women had higher alcohol use (B = 1.45, p = 0.001), while Black heterosexual women had lower alcohol use (B = − 1.39, p < 0.001). Compared to women experiencing no discrimination, women who experienced both racial and sexual orientation discrimination had higher alcohol use (B = 1.65, p = 0.003). For both men and women, older age was associated with higher alcohol use. Compared to attending a private school, attending a technical college was associated with lower alcohol use; for women, attending a state university was associated with lower alcohol use. For both men and women, having parents with highest education of less than a Bachelor’s degree was associated with lower alcohol use.

In bivariate analyses, use of tobacco products within the past 30 days was associated with being male, sexual minority, reporting racial discrimination, reporting sexual orientation discrimination, and school type.

Multivariable logistic regression models (shown in Table 4) indicated that, compared to White heterosexual women, White sexual minority women (OR = 2.21, p = 0.003) and Black sexual minority women (OR = 2.64, p = 0.003) were more likely to use tobacco products. Compared to women who experienced no discrimination, women who experienced only racial discrimination (OR = 1.42, p = 0.04) and both racial and sexual orientation discrimination (OR = 3.45, p < 0.001) were all more likely to use tobacco products. Additionally, for both men and women, compared to attending a private school, students attending a state university or a technical college were more likely to use tobacco products. Women attending a HBCU also were more likely to use tobacco products.

In bivariate analyses, use of marijuana within the past 30 days was associated with being Black, sexual minority, reporting racial discrimination, reporting sexual orientation discrimination, and school type.

Multivariable logistic regression models (shown in Table 5) indicated that, compared to White heterosexual women, Black heterosexual women (OR = 1.72, p = 0.02), Black sexual minority women (OR = 2.81, p = 0.007), and White sexual minority women (OR = 3.01, p < 0.001) were all more likely to use marijuana. Compared to women who experienced no discrimination, women who experienced racial (OR = 1.48, p = 0.03) or sexual orientation discrimination (OR = 3.07, p = 0.001) or both forms of discrimination (OR = 3.38, p < 0.001) were more likely to use marijuana. Compared to White heterosexual men, White sexual minority men (OR = 2.37, p = 0.009) were more likely to use marijuana. Finally, compared to attending a private school, women attending a state university were more likely to use marijuana. In men, however, this effect was reversed, such that men attending a state university were less likely to use marijuana.

The generalizability of findings is limited because our sample was drawn from young adult college students in Georgia. However, it is important to note that our sample was drawn from diverse schools, including private, public, technical, and historically black colleges and universities in both rural and urban settings. We also limited our analysis to only Black and White students and did not consider non-White Hispanic and Asian students due to the small size of these groups in our sample. Future studies with larger sample size of these populations should continue to investigate the interplay between intersecting identities and experiences of discrimination.

Additionally, in our study, we grouped students identifying as homosexual and students identifying as bisexual into one (“sexual minority”) due to the small sample size of each category. Some research has shown differences in substance use between homosexual versus bisexual college students [53, 54]. Future studies may want to over sample these different categories in order to further understand whether differences exist in regards to mental health and substance use between different sexual minority categories.

While our data was limited by being self-reported, cross-sectional data, the data included key measures, including the assessment of use of different substance (alcohol, diverse alternative tobacco products, and marijuana). Most importantly, as highlighted throughout this paper, the inclusion of items capturing intersecting identities and experience of discrimination allowed us to separate influences of minority statuses versus discrimination on health behaviors and outcomes.