Trajectories of Depressive Symptoms Among Young Adults in Texas 2014–2018: A Multilevel Growth Curve Analysis Using an Intersectional Lens

Abstract

Introduction:

Research has demonstrated disparities in depressive symptoms among people who are marginalized. However, more work should examine depressive symptoms through an intersectional lens, recognizing that multiple systems of privilege and oppression interlock to create unique struggles where multiple marginalized identities meet. Recent methodological developments have advanced quantitative intersectionality research using multilevel modeling to partition variance in depressive symptoms to person-level sociodemographic variables and intersectional-level social strata. The purpose of this study is to leverage these methods to examine trajectories of depressive symptoms among young adults in Texas through an intersectional lens.

Methods:

Multilevel modeling was used to examine the longitudinal trajectories of depressive symptoms among 3,575 young adults from 24 colleges in Texas assessed seven times between Fall 2014 and Spring 2018. Intersectional identities included sex, race/ethnicity, and sexual and gender minority identities. The model examined time nested within individuals and individuals nested within intersectional social strata.

Results:

Young adults in Texas experienced an increase in depressive symptoms from 2014–2018. Those with female, Hispanic, AAPI, Other race/ethnicity, or LGBTQ+ identities experienced more depressive symptoms. After controlling for the main effects of the sociodemographic variables, 0.08% of variance in depressive symptoms remained attributed to the effects of intersectional identities.

Conclusion:

Evaluating disparities in depressive symptoms through an intersectional lens offers a more complete description of the epidemiology of depressive symptoms. Communities and institutions that serve marginalized people should consider the elevated burden of depressive symptoms that marginalized people may carry, and integrate culturally competent psychoeducation, assessments, and therapies where possible.

Introduction

Addressing the critical need for increased awareness of Major Depressive Disorder has become an international priority [1]. However, depression remains a threat to public health, consistently ranking among the leading causes of global disease burden [2]. A 2018 United States (U.S.) nationally representative survey reported that 13.8% of young adults aged 18–25 years and 11.4% aged 26–29 years experienced a depressive episode in the past year, the highest prevalence among all adult age groups [3]. The burden of depression is especially heavy during young adulthood, a period marked by instability, exploration of personal and professional identities, and pressure to succeed [4]. Depression is characterized by anhedonia, dysregulated emotion, and attenuated motivation [5], states which are seemingly antithetical with milestones of this critical developmental period. As a result, young adults who experience depression achieve lower occupational attainment [6], live more physically unhealthy lifestyles [7], and are less likely to marry [8].

Studying depression in large populations comes with measurement challenges. The gold standard for establishing a diagnosis of depression is the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, 5^th Edition (SCID-5) [5]. However, the SCID-5 is lengthy to conduct and requires specialized training by the administrator. Instead, many researchers use self-report survey measures that assess symptoms of depression. These reliable and valid self-report measures of depression are highly correlated with diagnoses of depression [9–11]. Therefore, the focus of the present study is on depressive symptoms, an important indicator for depression.

Substantial disparities exist across the affected population of those experiencing depressive symptoms. A recent meta-analysis indicated that women are nearly twice as likely to experience depressive symptoms compared to men [12]. Racial and ethnic disparities in depressive symptoms are also evident, but more complicated to disentangle. Whereas the white population experiences more frequent acute episodes of depressive symptoms compared to Hispanic, Black, and Asian populations, racial/ethnic minorities are more likely to suffer from undiagnosed, chronic, and severe depression, leading to a higher total disease burden [13]. When considering people who identify as multiple races or ethnicities, the prevalence is higher than the white population [3], and also higher in severity and chronicity [13]. Emerging evidence suggests that perhaps the greatest disparities in depressive symptoms are present in the LGBTQ+ community that includes lesbian, gay, bisexual, transgender, queer, and other sexual and gender minority identities. Compared to cisgender heterosexual individuals, LGBTQ+ individuals are estimated to be over two and a half times more likely to experience depressive symptoms [14]. The evidence is clear, sex, racial/ethnic, and LGBTQ+ identities are associated with disparate patterns in the epidemiology of depressive symptoms.

Research to date has identified several pathways through which disparities in depressive symptoms exist across sex, racial/ethnic, and LGBTQ+ identities. For example, a leading theory explaining sex disparities in depressive symptoms is the ABC model, which is conceptualized as affective, biological, and cognitive features uniquely interacting with negative life events to result in differences in rates of depression across sex [15]. Disproportionate micro-level negative life events such as interpersonal sexual harassment and sexual assault are potentially traumatic events that steer developmental trajectories towards depression [16]. Nationally representative surveys in the U.S. reveal that 56% of adolescent girls compared to 40% of boys experience sexual harassment behavior annually, and 18% of female adults, compared to 1.4% of males, report having been raped [17, 18]. The disproportionate effect of negative life events is also evident when considering macro-level oppressive factors. Structural gender inequalities include women receiving reduced pay for identical work, lower female representation in leadership, lower educational opportunity and achievement for developing girls, and higher burden of interpersonal relationship dynamics such as contraceptive responsibility and expected parenting roles [19]. While the underlying mechanisms linking structural gender inequality to depression are complex and remain under investigation, at a minimum, these inequalities lead to increases in stress, which is an important contributor to depression [20]. The ABC model posits that these disproportionate stresses interact with affect, biology, and cognition to exacerbate sex disparities in depression [15].

Racism has been among the most prevalent oppressive forces in American history [21, 22]. Discrimination based on race or ethnicity comes in many forms. Within the current sociopolitical context, overt examples include disproportionate policing of Black communities (which Bowleg, et al. has independently identified as a pathway to depression for Black men [23]), voter suppression of people of color [24], and increasingly restrictive immigration policy and anti-immigrant cultural norms [25, 26]. While the subjective experience of racism varies across the diverse spectrum of racial and ethnic identities, a recent meta-analysis inclusive of Black, Hispanic, and Asian races and ethnicities found that across groups, racial discrimination had a moderate to large negative effect on mental health, especially increasing depressive symptoms [27]. Carter et al.’s meta-analysis suggests that this effect is driven by increased biological and psychological stress reactions, reduced collective self-esteem, and negative cultural identity [27].

Recent legal, political, and cultural advances have improved public acceptance and understanding of LGBTQ+ people and issues; however, discrimination, oppression and stigmatization remain prevalent. Despite federal legalization, same-sex marriage is still met with many barriers in some states, and these restrictions can hinder access to next-of-kin rights [28]. Additionally, violence and stigma are disproportionately targeted towards individuals who identify as LGBTQ+ [29, 30]. Stigma against individuals who identify as LGBTQ+ can stem from the dominant cultural viewpoint that “others” non-heterosexual sexualities and non-traditional expressions of gender. These experiences of being different from, and looked down upon by, the dominant cultural viewpoint, can manifest in a host of stressors, including hypervigilance against harassment or discrimination, expectations of rejection, fear of disclosure, violence, and more [31]. Indeed, a systematic review confirms that individuals who identify as LGBTQ+ are subject to many disparate adversities, including unequal legal rights, hate crimes, bullying and social ostracization, difficulty coming out to friends and family members, and stigma [32]. Each of these challenges can independently influence the pathogenesis of depression, making depressive symptoms more likely and ultimately leading to a disproportionate mental health burden [32].

Examining disparities in depressive symptoms reveals that individuals with sociodemographic identities associated with social, legal, and economic inequality are at greater risk than their counterparts. An overarching theory that helps explain this phenomenon -- the Minority Stress Model -- suggests that marginalized populations face unique and hostile stressors (e.g., sexism, racism, homophobia) related to their identities and that these stressors have negative effects on their health [33, 34]. But identities are not monolithic. They exist within the context of, and alongside each other. In other words, they are intersectional. Intersectionality theory posits that multiple systems of privilege and oppression interlock to create unique struggles where multiple marginalized identities meet [35].

Intersectional theorists have stressed the importance of treating each of these unique intersections as their own social determinant when conducting population health research, rather than taking an “additive approach” [36, 37]. The additive approach to intersectionality treats each additional marginalized sociodemographic identity that a person holds as risk factors that combine to create greater cumulative risk. For example, an LGBTQ+ Black woman would be said to have three sociodemographic risk factors for negative outcomes in the additive conceptualization compared to a white heterosexual man. Instead, researchers are encouraged to conceptualize intersectional identities as cohesive social positions rather than separate dimensions that happen to coexist. Intersectional theorists assert that these intersectional identities are meaningful whole identities in and of themselves, not merely the sum of each individual dimension [38].

Research investigating disparities in depressive symptoms using an intersectional lens is limited. Patil et al. conducted the first systematic review investigating racial/ethnic, and gender intersectionality in the context of depressive symptoms during adolescence and young adulthood [39]. Drawing from the 25 studies meeting their criteria for inclusion, the authors concluded that some young people of color in the U.S. were at greater risk for depression and that girls were consistently at greater risk than boys across most racial/ethnic identities. The study noted that much of the reviewed literature focused on single dimensions of identity separately. Moreover, when multiple dimensions were considered, often it was done in the additive manner, using fixed-effect interactions, which is diametrically opposed to intersectionality scholars’ suggestions. The review concluded that people with marginalized identities are consistently subjected to more discrimination in the U.S., especially during critical periods of identity formation and socialization like adolescence and young adulthood. This increased discrimination likely predisposes some specific diverse groups to depression more than others. The authors called for more research that explicitly uses intersectional methods grounded in intersectional theory, treating each combination of intersectional identities as unique rather than interactions of several dimensions.

Evans and Erickson answered the call for quantitative intersectional methodology informed by theory using a novel method -- multilevel analysis of individual heterogeneity and discriminatory accuracy (MAIHDA) -- to examine intersections of gender, race/ethnicity, immigration status, and family income in a large, nationally representative sample of adolescents and young adults [40]. Their MAIHDA analysis operationalized each unique intersectional identity as one of 28 “social strata,” constituting each of the possible combinations of gender, race/ethnicity, immigration status, and income level. In doing this, each individual was assigned an intersectional “social stratum” based on the four sociodemographic variables included. MAIHDA models have traditionally been used to analyze geographical inequalities by partitioning variance into components explained by person-level variables and components explained by geographic-level variables, a statistic henceforth referred to as the Variance Partition Coefficient (VPC) [41]. By using this approach, Evans and Erickson’s models allowed for clustering of data around the assumed shared experience of each unique social stratum (e.g., female Hispanic immigrant of low-income level), thus applying an analysis technique to explore social strata as more than the sum of their parts [40].

Considerable disparities were found to exist between intersectional social strata [40]. Women, racial/ethnic minorities, immigrants, and individuals of low income reported more depressive symptoms. However, after controlling for the primary effects of gender, race/ethnicity, immigration status, and income level, only up to 0.28% of variance in depressive symptoms was attributed to intersectional identities. According to Merlo et al.’s proposed rules of thumb, this VPC is indicative of a very small effect size [42]. These findings suggested that intersections of gender, race/ethnicity, immigration status, and family income pose some additional risk for depression above and beyond singular sociodemographic determinants of depression during the period these data spanned (1995–2008); however, the effect was small [40]. In a complementary study using the same population, Evans reproduced this analysis using additional intersectional sociodemographic variables, including the participant’s sexual identity and their parent’s highest level of education, and arrived at a similar conclusion [43]. However, additional research with more contemporary samples is needed.

In the present study, we build on, and extend, the state of this literature by examining disparities in depressive symptoms through an intersectional lens in three ways. First, by using a more recent sample of young adults, we hope to offer more context to the mental health challenges that may present to certain social strata within the current sociopolitical environment. The current American sociopolitical period is much different than preceding eras, with the very forces that foster oppression -- sexism, racism, and LGBTQ+ discrimination and stigmatization -- being increasingly felt [44–50]. The national proliferation of sexism, racism, and LGBTQ+ discrimination and stigmatization over time can be inferred by tracking the number of hate groups (e.g., anti-immigrant, anti-LQBTQ+, anti-Muslim, Ku Klux Klan, Neo-Nazi, White Nationalist, etc.) around the U.S. each year, which are actively monitored by the Southern Poverty Law Center. In 2000, after a decade of relative stability in the number of hate groups, there remained 599 hate groups across the U.S. This number nearly doubled to 1,018 by 2011 [51]. The proliferation of hate is also reflected in reports from the National Center for Education Statistics that indicate a nearly 50% increase in hate crimes committed on college campuses from 2011–2016 [52]. Modernized analyses using an intersectional framework are needed in many health contexts to illuminate how this proliferation of overt public hate may have contributed to health disparities, such as disparities in depressive symptoms. Second, we build on and extend the existing literature by using a more inclusive assessment of LGBTQ+ identities. While no conceptualization of combinations of sociodemographic variables constitutes the “perfect” model of intersectionality, considering the known magnitude of depressive symptom disparities in the LGBTQ+ community [14], thoughtful assessments of LGBTQ+ identities should surely be considered. Third, we build on and extend the existing literature by following the general MAIHDA tradition when constructing our models while also integrating growth curve analysis methodology using longitudinal data with seven waves. Importantly, longitudinal data from our sample was collected from 2014–2018, a period that roughly parallels the resurgence of overt sexism, racism, and LGBTQ+ discrimination and stigmatization in public discourse [44–52].

The present study applies the MAIHDA framework to investigate longitudinal trajectories of depressive symptoms across a period of four years in young adulthood among intersections of sex, race/ethnicity, and LGBTQ+ identities. We hypothesize (1) linear growth in depressive symptoms across the four-year period, (2) that individuals with marginalized identities will demonstrate elevated depressive symptoms, and (3) that intersectional identities will explain variance in depressive symptoms after controlling for the primary effects of the person-level sociodemographic variables that construct intersectional identities.

Methods

Participants and Procedure

Participants were 3,575 students aged 18–29 from 24 metropolitan colleges in Texas who were assessed every six months for two and a half years (starting fall 2014) and then a one year follow up (spring 2018), for a total of seven waves. After providing informed consent, they self-reported data via a web-based survey at each assessment. This sample was a subset from a larger population health surveillance study of Texas college students designed to assess tobacco use and related behaviors: the Marketing and Promotions Across Colleges in Texas Project (“Project M-PACT”) [53–55]. Project M-PACT was approved by the Institutional Review Board at the University of Texas at Austin. All Project M-PACT participants who provided data on date of birth, sex, race/ethnicity, LGBTQ+ identity, and at least two waves of depressive symptoms data were included.

Measures

Sociodemographic Variables

Participants self-reported data via a web-based survey on date of birth, sex, race/ethnicity, and LGBTQ+ identity. Sex, age, and race/ethnicity were recorded at wave 1. Sex was assessed with one item (“What is your sex?”). The response options included “female” and “male” and was used as a binary variable with male as the reference category. Age was calculated from date of birth. Race/ethnicity was assessed by asking if the participant was Hispanic, and then asking them to indicate one or more of the following racial identities that they identified with: white, Black/African American, Asian, American Indian/Native Alaskan, Hawaiian/Pacific Islander, and “Other”. For the present analysis, a nominal variable was created denoting membership to one of the following racial categories: white, Black, Hispanic, Asian, or Other, with white as the reference category. “Other” indicated all possible combination of remaining racial/ethnic identities including American Indian/Native Alaskan, Hawaiian/Pacific Islander, self-identified “other,” or being of multiple racial or ethnic identities. The decision to collapse several racial/ethnic identities into “other” was made to preserve sample size.

LGBTQ+ identity was assessed at waves three through seven, which allowed for developmental fluidity in this construct. Participants were asked to check all of the following sexual and gender identity categories that they identified with: “Heterosexual or straight,” “Gay or Lesbian,” “Bisexual,” “Queer,” “Transgender,” “Other.” For the present analysis, LGBTQ+ identity was coded as a binary variable with 0 representing cisgender heterosexual individuals, and 1 indicating membership within the LGBTQ+ community at any assessment during the study. Any self-report of any identity other than cisgender and heterosexual was coded as LGBTQ+. For example, if a participant reported being heterosexual and bisexual in the same wave, then they were coded as LGBTQ+.

Intersectionality

Intersectional social strata were constructed for all possible combinations of sex, race/ethnicity, and LGBTQ+ identity, totaling 20. Frequencies for these 20 social strata are reported in Table 1.

Table 1.

Sample Characteristics and Mean Estimates of Depressive Symptoms with 95% C.I. across all Waves

Variable	N	Estimate	95% C.I.
Total	3575	7.87	7.80 – 7.95
Sex
Male	1,244	7.40	7.28 – 7.52
Female	2,331	8.13	8.04 – 8.22
Race/Ethnicity
White	1,260	7.61	7.49 – 7.74
Hispanic	1,088	7.89	7.76 – 8.02
Black	270	8.06	7.80 – 8.32
Asian	687	8.13	7.96 – 8.29
Other	270	8.23	7.98 – 8.49
Sexual and Gender Identity
Cisgender Heterosexual	2,930	7.37	7.30 – 7.45
LGBTQ+	645	10.18	9.99 – 10.38
Social Strata	20
Cis-Heterosexual white Male	366	6.42	6.21 – 6.62
Cis-Heterosexual Black Male	42	6.39	5.85 – 6.93
Cis-Heterosexual Hispanic Male	304	6.97	6.74 – 7.19
Cis-Heterosexual AAPI Male	237	7.62	7.36 – 7.89
Cis-Heterosexual Other Male	79	7.91	7.46 – 8.36
LGBTQ+ white Male	91	9.31	8.80 – 9.82
LGBTQ+ Black Male	11	10.47	8.90 – 12.03
LQBTQ+ Hispanic Male	64	9.88	9.26 – 10.50
LGBTQ+ AAPI Male	35	9.35	8.59 – 10.11
LGBTQ+ Other Male	15	8.39	7.07 – 9.71
Cis-Heterosexual white Female	638	7.23	7.06 – 7.39
Cis-Heterosexual Black Female	183	7.64	7.36 – 7.93
Cis-Heterosexual Hispanic Female	594	7.67	7.50 – 7.85
Cis-Heterosexual AAPI Female	351	7.86	7.64 – 8.09
Cis-Heterosexual Other Female	136	8.11	7.74 – 8.49
LGBTQ+ white Female	165	10.82	10.43 – 11.21
LGBQ+ Black Female	34	11.58	10.63 – 12.52
LQBTQ+ Hispanic Female	126	10.16	9.71 – 10.61
LGBTQ+ AAPI Female	64	10.76	10.18 – 11.34
LGBTQ+ Other Female	40	9.24	8.57 – 9.91

Outcome Variable

Participants self-reported depressive symptoms using the Center for Epidemiologic Studies–Depression Scale-10 (CES-D-10) [9], a 10-item self-report measure of depression commonly used for population research. It consists of a series of questions related to depression (e.g., “I felt depressed.”), negative affect (e.g., “I felt fearful.”), and functional impairment (e.g., “I was bothered by things that don’t usually bother me.”), with possible responses in the Likert-form ranging from 0 = “none of the time,” to 3 = “all of the time.” A higher CES-D-10 total score indicates more depressive symptoms. The CES-D-10 has been validated for use to assess depressive symptoms in the general population, demonstrating strong convergent validity with other assessments of depression and negative affect [56]. In the present study the CES-D-10 Cronbach’s α = 0.88.

Analysis

Hypothesis 1 was that there would be linear growth in depressive symptoms across the four-year period. To address hypothesis 1, we fit linear and quadratic unconditional growth models of depressive symptoms. Akaike Information Criterion (AIC) was used to determine the best model of change, where lower values indicate better fit.

Hypothesis 2 was that individuals with marginalized identities would demonstrate elevated depressive symptoms. To address hypothesis 2, we examined the fixed effects of sex, race/ethnicity, and LGBTQ+ identities on depressive symptoms. Then, we examined the fixed effects of the interactions of sex, race/ethnicity, and LGBTQ+ identities with time and plotted trajectories of depressive symptoms separated by these sociodemographic identities.

Hypothesis 3 was that intersectional identities would explain variance in depressive symptoms after controlling for the primary effects of the person-level sociodemographic variables that construct intersectional identities. To address hypothesis 3, we examined the random effects of social strata. Then, we plotted mean estimates of depressive symptoms with 95% confidence intervals (C.I.) for each social strata across all waves.

Analyses were performed using multilevel regression modeling and following the general procedures of the MAIHDA framework [40–43, 57]. The dependent variable was depressive symptoms. The multilevel model accounted for the nesting of study waves within individuals, and individuals within intersectional social strata. The fixed effects in the model included wave (which was coded 1 to 7: representing intervals of time between Fall 2014 and Spring 2018), age at wave 1, sex, race/ethnicity, LGBTQ+ identity, sex × wave, race/ethnicity × wave, and LGBTQ+ identity × wave. The random effects in the model included an intercept (i.e., a unique participant ID), a slope (i.e., wave), and intersectionality (i.e., a unique identifier that corresponded with one of the 20 social strata). Results are presented in two multilevel models. First, the “main effects model” examined the independent effects of sex, race/ethnicity, and LQBTQ+ identities. Then, the “interaction effects model” examined the interactions of sex, race/ethnicity, and LQBTQ+ identities with time. We calculated the VPCs for the pooled fixed effects, for the random intercept, for the random slope, and for social strata. VPC’s were calculated as a proportion with the variance in depressive symptoms attributed to each variable as a numerator and the total variance in depressive symptoms as a denominator, multiplied by 100. Multilevel analyses were conducted using the mixed procedure of Stata 15.1 [58].

Results

The AIC for the unconditional linear growth model was 144741.8, and for the unconditional quadradic growth model was 144768.2, indicating that the linear model of growth was the better fit.

Parameter estimates and VPCs for the multilevel regression models are reported in Table 1 (the main effects model) and Table 2 (the interaction effects model). In the main effects model, wave was significantly associated with depressive symptoms (β = 0.09, 95% C.I. = 0.07, 0.11), indicating that depressive symptoms increased linearly over time, on average. Female sex (β =0.79, 95% C.I. = 0.51, 1.07), Hispanic race/ethnicity (β =0.34, 95% C.I. = 0.01, 0.68), AAPI race/ethnicity (β =0.77, 95% C.I. = 0.38, 1.16), Other race/ethnicity (β =0.71, 95% C.I. = 0.15, 1.27), and LGBTQ+ identity (β =2.80, 95% C.I. = 2.43, 3.17) were significantly associated with depressive symptoms, indicating that these marginalized identities experienced higher depressive symptoms. In the interaction effects model, female x wave was significantly inversely associated with depressive symptoms (β = −0.09, 95% C.I. = −0.14, −0.03), indicating that the growth in depressive symptoms for females occurred at a slower rate than for males during the study period. Linear trajectories of depressive symptoms separated by sex, race/ethnicity, and LGBTQ+ identities are presented in Figure 1, Figure 2, and Figure 3 respectively.

Table 2.

Multilevel Model Parameter Estimates for Depressive Symptoms – Main Effects Model (n=3575)

Variable	Estimates	95% C.I.
Fixed Effects
Intercept	6.49	5.20, 7.77
Wave	0.09	0.07, 0.12
Age	−0.02	−0.08, 0.04
Sex
Male (reference)	-	-
Female	0.79	0.51, 1.07
Race/Ethnicity
White (reference)	-	-
Hispanic	0.34	0.01, 0.68
Black	0.39	−0.16, 0.94
AAPI	0.77	0.38, 1.16
Other	0.71	0.15, 1.27
LGBTQ+ Identity
No (reference)	-	-
Yes	2.80	2.43, 3.17
Random Effects
Intercept	10.93	9.78, 12.22
Wave	0.16	0.14, 0.18
Social Strata	0.02	0.01, 0.03
Residual	15.24	14.93, 15.56
VPC For Fixed Effects	57.84%
VPC For Random Intercept (%)	41.48%
VPC For Random Slope (%)	0.60%
VPC For Social Strata (%)	0.08%

Bold indicates statistical significance (p<0.05)

VPC = Variance Partition Coefficient

Figure 1.

Trajectories of Depressive Symptoms by Sex (n=3575)

Figure 2.

Trajectories of Depressive Symptoms by Race/Ethnicity (n=3575)

Figure 3.

Trajectories of Depressive Symptoms by sexual and gender identity (n=3575)

Variance estimates for social strata are identical in both the main effects and interaction effects models (${\mathrm{\sigma }}_v^2$ = 0.02, 95% C.I. = 0.01–0.03), indicating that a very small amount of variance is attributed to intersectionality after controlling for the primary effects of the person-level sociodemographic variables that construct intersectional identities. VPCs were calculated as 57.28–57.84% for the pooled fixed effects, 41.48–41.50% for the random intercept, 0.60% for the random slope, and 0.08% for social strata. These VPCs indicated that after controlling for age and the primary effects of the person-level sociodemographic variables that construct intersectional identities, as well as random variation assumed within and between individuals, 0.08% of the variance in depressive symptoms remained attributed to the effects of intersectional identities. Mean estimates of depressive symptoms with 95% C.I. plots for each social strata across all waves are presented in Figure 4.

Figure 4.

Mean Estimates of Depressive Symptoms with 95% C.I. Plots for each Social Strata across all Waves (n=3575)

Discussion

Our substantive interest focused on investigating disparities in depressive symptoms within a young adult population in Texas through an intersectional lens, using a contemporary data set. The finding that linear change fit the data well and the significant fixed effect for wave in the main effects model supports our first hypothesis that depressive symptoms would increase over time, specifically that from 2014–2018 young adults in Texas would experience an increase in depressive symptoms. This is consistent with the overall epidemiological trend in depressive symptoms, which show that since at least 2005, depression has been on the rise, especially among youth and young adults [78].

The significant fixed effects in the main effects model and the plotted longitudinal trajectories of depressive symptoms indicated that young adults in Texas with any of the following identities, female, Hispanic, AAPI, Other race/ethnicity, and LGBTQ+, demonstrated elevated depressive symptoms from 2014–2018. This finding is consistent with previous research examining disparities in depressive symptoms and supports our second hypothesis that people who are marginalized would have greater burden of depressive symptoms [3, 12–14]. This finding is also consistent with recent evidence from a study of Texas young adults that reported mental health disparities among sexual minority youth [80]. While it is impossible to determine exogenous factors that contributed to disparities in depressive symptoms using our data, we point to the resurgence of discourse related to overt sexism, racism, and LGBTQ+ discrimination and stigmatization that precedes and parallels the time frame of our study as an area for future research [44–52].

Research suggests that individuals with higher education levels and those who are members of Millennial generation or Generation-Z cohorts have a strong affinity for equity and social justice [59–61]. Our sample was young adult college students in 2014–2018, which suggests our participants were members of these generations who had obtained at least some higher education. Thus, overt hate and discrimination in public discourse may be particularly offensive to the young adults in our sample. This may cause cognitive dissonance with one’s internalized cultural schema, potentially leading to rumination and reappraisal of one’s roles in society and influencing growth of depressive symptoms. Some evidence supports this interpretation within the context of immigrant and intergenerational cultural dissonance [62, 63], however future research should study this phenomenon within the proposed context.

Sociopolitical changes that create or reenforce discriminatory structures and practices also may have influenced depressive symptoms within our population, particularly for those with marginalized identities. For example, Texas state laws continue to disproportionally affect already marginalized populations. In the 87th Texas Legislature session (most recent as of this writing), lawmakers prioritized banning the teaching of critical race theory (TX HB3979 [64]), banning abortions occurring 6-weeks after conception (TX SB8 [65]), restricting transgender children’s access to gender-transition-related medical care (TX SB1311 [66]), and restricting voting access in areas with larger communities of color (TX SB7 [67]). These are just the most recent examples of a decades-long Texas lawmaking campaign that gained favor in 1995 and accelerated in 2003 when statewide leadership transitioned firmly towards a more conservative political ideology [68].

The legal structures in Texas are particularly concerning when viewed through the Minority Stress Model [33, 34]. For example, banning the teaching of critical race theory perpetuates colonized education practices (i.e., powerful white men establishing self-aggrandizing curriculums) that stifle academic achievement and decrease well-being for students of color [69]. Restricting abortion is a longstanding patriarchal strategy which shifts the locus of control for a woman’s body from self to legal doctrine, which has been shown to profoundly affect the mental health of women [70]. Limiting transgender children’s accesses to medical care is a structural barrier to LGBTQ+ equity that compromises the physician-patient relationship, making shared decision making extremely difficult or impossible. Shared decision making has been shown to benefit both mental health and physical health [71], especially among transgender youth [72]. Thus, disproportionately restricting shared decision making for LGBTQ+ individuals may increase their depressive symptoms. Restricting voting access for communities of color is another longstanding strategy of oppression that dilutes the democratic voice of entire neighborhoods and demographics. The downstream effects of such voter suppression are profound, and include increased exposure to environmental risk, disempowerment, social exclusion, psychological distress, and physical violence [73]. While it is indeed possible that the Texas legal structures influenced depressive symptoms in the study population, the present study nevertheless was not able to not test for this. Future research to test this hypothesis is necessary.

Our results also showed a significant interaction between sex and wave, indicating that the growth in depressive symptoms over time was slower for females than for males. While females started with more depressive symptoms and remained higher than males throughout the study, this finding suggests that young adult males in Texas have an accelerated rate of worsening depression symptoms. This finding calls for future research to investigate the underlying mechanisms of the potential worsening mental health of young men in Texas. Furthermore, the recent passing of TX SB8 [65] which effectively bans abortions occurring 6-weeks after conception threatens the mental health of women in Texas both receiving and not receiving an abortion [79]. Thus, researchers should also closely monitor the epidemiology of depressive symptoms among females in Texas.

The multilevel models allowed for time (level 1) nested within individuals (level 2), and individuals nested within social strata (level 3) and indicated that intersectional identities explained a very small amount of variance in depressive symptoms after controlling for the primary effects of the person-level sociodemographic variables that construct intersectional identities. Findings in the multilevel models suggested that a little over half of the variance in depressive symptoms is attributed to the pooled fixed effects of wave, age, sex, race/ethnicity, and LQBTQ+ identity; about 40% of the variance in depressive symptoms is attributed to unmeasured or random variation between individuals; only 0.60% of variance in depressive symptoms is attributed to random variation within individuals; and just 0.08% of variance in depressive symptoms is attributed to social strata. This finding supports our third hypothesis that intersectional identities would explain variance in depressive symptoms over and above the person-level variables, although the effect is extremely small [42]. A small effect for social strata is consistent with previous literature which calculated VPC’s for social strata equal to 0.06–0.28% [40]. These consistent findings may suggest that sociodemographic disparities in depressive symptoms are driven by the main effects of each single marginalized identity across social strata. However, future research should continue to incorporate an intersectional lens into their work to further understand this relationship.

As suggested by intersectionality theory and minority stress theory, the substantive findings of the present paper findings are not unexpected [33–38]. We propose an interpretation: emergent sociological evidence suggests that after a long period of social progress in the U.S -- in terms of reducing discrimination and prejudice directed at marginalized sociodemographic identities -- current young adults are facing an increase in these societal stressors [44–52, 74, 75]. Stress and discrimination are supported with robust evidence as being predisposing or causal of depressive symptoms [15, 20, 23, 27, 32, 76, 77]. It is reasonable to suggest that a social environment with increased stress and discrimination may lead to the development of more depressive symptoms among people subject to unique pathways of stress and discrimination. These unique pathways of stress and discrimination may actively work against public health initiates aiming for health equity. Future research would be keen to directly assess mediating factors in the relationship between sociodemographic identities and depressive symptoms.

This study should be considered in light of its limitations. First, the sample is not representative of the entire U.S. and is limited to college students in Texas. However, this large sample was diverse and included students from private and public universities, as well as community colleges and other 2-year institutions. Second, depressive symptoms are self-reported and may be subject to biases. However, self-report is the most feasible way to assess depressive symptoms in large populations, and the assessment we used is widely accepted. Third, this paper does not empirically examine the underlying mechanisms that may explain disparities in depressive symptoms; rather it is a descriptive epidemiologic investigation of how depressive symptoms may vary by intersectional identities. Future research is needed to examine causal mechanisms, such as discrimination or exposure to hate. Fourth, while our assessment of LGBTQ+ identities improved upon those used in previous research investigating depression through an intersectional lens, we were limited in our ability to fully assess sexual and gender identities. Best practices [81] require assessing sexual orientation and gender identity (SOGI) in separate questions, yet accurate and inclusive measures for SOGI are limited and still evolving [82]. Lastly, intersectionality is a dynamic theory, and the present study only integrated three intersections of sociodemographic identity. It is possible that with different or additional measures of identity, findings may change. Future research is needed that examines additional sociodemographic identities such as socioeconomic status, education level, immigration status, among others.

Conclusion

Young adults in Texas experienced an increase in depressive symptoms from 2014–2018. Those with female, Hispanic, AAPI, Other race/ethnicity, or LGBTQ+ identities experienced more depressive symptoms, but growth in depressive symptoms among males was faster than among females. Evaluating these trends through an intersectional lens offers a more nuanced and complete description of depressive symptoms in people with multiple marginalized identities. This study offers unique contributions to the literature because of its attention to thoughtful assessments of identity, a diverse sample, a data collection period that roughly parallels the resurgence of prejudice and hate-speak in public discourse, and an integration of intersectional theory with a robust analytic approach. Findings underscore the need to better understand the interplay of depressive symptoms, intersectionality, and sociopolitical trends. Communities and institutions that serve marginalized people should consider the elevated burden of depressive symptoms that marginalized people may carry, and integrate culturally competent psychoeducation, assessments, and therapies where possible.

Table 3.

Multilevel Model Parameter Estimates for Depressive Symptoms – Interaction Effects Model (n=3575)

Variable	Estimates	95% C.I.
Fixed Effects
Intercept	6.33	5.03, 7.62
Wave	0.14	0.09, 0.20
Age	−0.02	−0.08, 0.04
Sex
Male (reference)	-	-
Female	1.06	0.73, 1.39
Race/Ethnicity
White (reference)	-	-
Hispanic	0.29	−0.10, 0.68
Black	0.37	−0.27, 1.01
AAPI	0.88	0.43, 1.33
Other	0.73	0.08, 1.37
LGBTQ+ Identity
No (reference)	-	-
Yes	2.68	2.26, 3.11
Sociodemographic*Wave Interactions	-	-
Female*Wave	−0.09	−0.14, −0.03
Hispanic*Wave	0.02	−0.05, 0.08
Black*Wave	0.01	−0.10, 0.11
AAPI*Wave	−0.04	−0.11, 0.04
Other*Wave	−0.01	−0.11, 0.10
LGBTQ+ Identity*Wave	0.04	−0.03, 0.11
Random Effects
Intercept	10.94	9.79, 12.24
Wave	0.16	0.14, 0.18
Social Strata	0.02	0.01, 0.03
Residual	15.24	14.93, 15.55
VPC For Fixed Effects	57.82%
VPC For Random Intercept (%)	41.50%
VPC For Random Slope (%)	0.60%
VPC For Social Strata (%)	0.08%

Bold indicates statistical significance (p<0.05)

VPC = Variance Partition Coefficient