Intersections of Racial Discrimination and LGB Victimization for Mental Health: A Prospective Study of Sexual Minority Youth of Color

Abstract

Sexual minority youth (i.e., lesbian, gay, and bisexual youth; LGB) of color have multiple minoritized identities, and few studies examine the implications of intersectional minority stressors for their prospective mental health. The current study tested three intersectional hypotheses: the additive hypothesis—racial discrimination and LGB victimization are independently associated with mental health; the multiplicative hypothesis—racial discrimination and LGB victimization interact in to exacerbate their negative association with mental health, and the inuring hypothesis—only racial discrimination or LGB victimization is associated with mental health. Data come from a sample of lesbian, gay, and bisexual youth of color (36% Black, 30% Latino, 26% Multi-racial, 4% Native American, and 3% Asian, Hawaiian, and Pacific Islander) from two U.S. cities, one in the Northeast (77%) and one in the Southwest, who were between ages 15–24 (M = 19) and surveyed four times over three years spaced nine months apart (N = 476; 38% bisexual; 67% free and reduced lunch; and 49% assigned female at birth). The multiplicative hypothesis was supported for depression symptoms, and the additive hypothesis was supported for suicidal ideation. Intersectional minority stressors undermine the mental health of sexual minority youth of color and warrant further investigation.

Introduction

Sexual minority youth (SMY¹) experience greater levels of compromised mental health compared to heterosexual youth (Ross et al., 2018). SMY are also subjected to greater levels of victimization, often tied to being lesbian, gay, bisexual, or other non-heterosexual identities (LGB victimization), that partly explains disparities in depression and suicide between SMY and heterosexual youth (Burton et al., 2013). Indeed, researchers have suggested that the typical developmental challenges that accompany adolescence (e.g., increased peer social regulation), are exacerbated for SMY who experience many milestones of sexual identity development (e.g., disclosing a sexual identity) during adolescence that make them vulnerable to LGB victimization (Russell & Fish, 2016, 2019). The compounding effects of LGB victimization with greater social regulation by peers during adolescence may explain the persistence of mental health disparities for SMY over time (Russell & Fish, 2016, 2019). However, research on LGB victimization and this “developmental collision” have neglected to attend to the challenges faced by SMY with multiple minoritized identities, such as their race and ethnicity, who may experience discrimination or victimization tied to both identities, further compromising their mental health (Thoma & Huebner, 2013). In response to these omissions, the present study investigates how experiences of racial discrimination and LGB victimization overlap in their association with the prospective mental health of SMY of color during adolescence and young adulthood.

Minority Stress

Minority stress theory (Meyer, 2003b) proposes that societal stigma linked to being a sexual minority exposes sexual minority people to additional and unique stressors related to their sexual orientation, such as victimization due to one’s sexual orientation, that explain the elevated levels of compromised mental health among sexual minority people. Cross-sectional research supports minority stress theory, generally indicating that greater LGB victimization is associated with greater depression symptoms and suicide risk (Burton et al., 2013). In longitudinal studies focused on or including SMY, Depression symptoms typically decline as youth age into young adulthood (Burton et al., 2013). Suicidality also declines with age among sexual minorities (Fish et al., 2019; Marshal et al., 2013). Unfortunately, greater LGB victimization also prospectively correlates with higher levels of psychological distress, which can maintain elevated rates of compromised mental health for SMY (Birkett et al., 2015). Thus, higher levels of compromised mental health earlier in life may partly explain the persistence of mental health disparities for sexual minority adults in later life (Fish et al., 2019). Although minority stress theory (Meyer, 2003b) acknowledges that these stress processes may differ across distinct identities, it does not explicitly consider that minority stressors themselves may not be tied to sexual identity alone.

Racism and Racial Discrimination

Racial discrimination is still a prevalent issue for the health and well-being of racial minorities (Pew Research Center, 2019). Racial discrimination is re-enforced by systems of racial inequality that privilege the experiences and needs of White people over those of people of color (Reskin, 2012). Given that SMY of color are also subjected to racial discrimination, it is important to understand how racial discrimination contributes to SMY’s mental health. Prospective studies on racial discrimination among racial minority youth suggest that, although there is important variability, racial discrimination tends to be highest in early and middle adolescence, and declines beginning in late adolescence (Hughes et al., 2016).

Few studies examine racial discrimination in the lives of SMY of color, and none are prospective. In one of the only studies including SMY of color and assessing multiple forms of discrimination, greater racial discrimination was associated with worse depression symptoms independent of LGB victimization (Thoma & Huebner, 2013). In the studies on or including SMY and assessing the association between racial discrimination and suicidal ideation, the results were mixed. In one cross-sectional study of Black SMY, there was a positive correlation between racial discrimination and suicidal ideation (Thoma & Huebner, 2013). Yet in another study of Black homeless youth, that included Black SMY, there was no association between racial discrimination and suicidal ideation (Gattis & Larson, 2016). Prior findings may be mixed because studies are cross-sectional, have diverse ages of participants, and focus on different populations.

Racial Discrimination and LGB Victimization among SMY of Color

Clear evidence exists that LGB victimization and racial discrimination independently compromise the mental health of SMY of color (Thoma & Huebner, 2013). However, a broader framing is needed to integrate these two areas of research in order to understand how LGB victimization and racial discrimination overlap to affect the long-term mental health of SMY of color. Intersectionality theory (Crenshaw, 1991) expands minority stress theory (Meyer, 2003b) and conceptualizations of racial discrimination by proposing that each individual stress process can be unique or overlapping based on the multiple identities held by SMY of color.

Generally three hypotheses related to how multiple experiences of discrimination intersect to affect mental health have been proposed—the additive hypothesis (Thoma & Huebner, 2013); the multiplicative hypothesis (Thoma & Huebner, 2013); and the inuring hypothesis (Raver & Nishii, 2010). These three hypotheses have only been tested in cross-sectional studies. However, applying these hypotheses to long-term mental health is warranted given the prospective associations between LGB victimization mental health among SMY and prospective associations between racial discrimination and mental health among racial minority youth (Birkett et al., 2015; Lee et al., 2020).

The additive hypothesis suggests that experiencing discrimination tied to multiple identities has an incrementally worse effect on mental health. Empirically, the additive model is tested by including both forms of discrimination in a statistical model. If the additive hypothesis is supported, SMY of color who experience more than one form of discrimination should have worse mental health for each additional form of discrimination experienced. The additive perspective has received only limited support in samples of SMY. In one cross-sectional study of Black SMY, Thoma & Huebner (2013) found that experiencing racial discrimination and sexual identity discrimination were both associated with more depression symptoms. Thus, there is cross-sectional evidence that supports the additive hypothesis for depression symptoms, but not for suicidal ideation, and this hypothesis has yet to be tested with longitudinal data. Because mental health typically improves over time, it would be expected that the additive effects of multiple forms of discrimination would mitigate improvements in mental health over time.

The multiplicative hypothesis suggests that the overlap in multiple forms of discrimination exacerbates mental health such that two forms of discrimination are associated with mental health that is worse than the association between any single form of discrimination and mental health. Empirically, researchers test the multiplicative hypothesis by assessing the interaction between two or more forms of discrimination. If the multiplicative hypothesis is supported, SMY of color who, for example, experience high levels of racial discrimination and LGB victimization would levels of depression symptoms twice those of SMY of color who only experience high levels of racial discrimination. The multiplicative hypothesis has only been tested among SMY of color in one cross-sectional study of Black SMY, and there was not a significant interaction between racial discrimination and LGB victimization when predicting depression symptoms or suicidal ideation (Thoma & Huebner, 2013). Longitudinally, the multiplicative hypothesis would be supported if there were an interaction between multiple forms of discrimination that slowed the improvement of mental health over time.

The inuring hypothesis suggests that mental health symptoms do not get worse beyond one form of discrimination. The inuring hypothesis is based on psychological adaptation theory (Helson, 1964), which suggests that when people are repeatedly exposed to the same stimulus, they tend to have a weaker response (habituate) to that same stimulus over time. The inuring hypothesis is tested the same way as the additive hypothesis; however, to confirm this hypothesis, each form of discrimination must have an independent association with mental health before testing both in the same model. The main difference between the inuring and additive hypothesis is that when including multiple forms of discrimination in a model, only one form remains significant (Thoma & Huebner, 2013). One study examined multiple forms of discrimination and their association with depression symptoms and suicidality among SMY and did not supported the inuring hypothesis (Thoma & Huebner, 2013). Longitudinally, the inuring hypothesis would be supported if improvement in mental health slowed when experiencing one, but not two forms of discrimination.

Current Study

SMY of color have multiple minoritized identities that make expose them to racial discrimination and LGB victimization, and both can compromise their mental health (Thoma & Huebner, 2013). Yet, existing research on SMY had given limited attention to the role of multiple sources of discrimination and victimization for their mental health, and it has been limited to cross-sectional studies. The current study addresses the limitations of existing research by investigating how intersectional minority stress compromises the prospective mental health of SMY of color. Using longitudinal data of SMY of color who were surveyed four times over three years, it is first examined if, consistent with prior research, racial discrimination, depression symptoms, and suicidal ideation decline over the course of the study (Research Question 1). Next the additive (Hypothesis 1), multiplicative (Hypothesis 2), and inuring (Hypothesis 3) hypotheses are tested to understand how racial discrimination and LGB victimization relate to depression symptoms and suicidal ideation, respectively.

Methods

Sample

SMY were recruited from LBGT community centers, college groups for LGBTQ youth, and referrals from other participants from two diverse cities—one in the Northeast and one in the Southwest. A waiver of parental consent was obtained, and assent procedures, including the presence of an independent youth advocate, were implemented for youth under the age of 18 to ensure that those who were not “out” to their parents would be well informed to give their assent. Participants provided contact information to complete follow-up surveys. SMY were between ages 15–24 at baseline (n = 585); the analytical sample was limited to 476 SMY of color. The data were collected at four time points spaced nine months apart. The mean age of the sample at baseline was about nineteen years old (M = 19, SD = 1.79). The sample break down by age was as follows 15 (n = 10), 16 (n = 27), 17 (n = 56), 18 (n = 74), 19 (n = 87), 20 (n = 84), 21 (n = 86), 22 (n = 50), 23 (n = 1), 24 (n = 1). The majority of the youth were from the Northeast data collection site (77%). Youth identified as lesbian or gay (58%), bisexual (38%), and queer or questioning (4%). The sample was racially and ethnically diverse with SMY who Black (36%) were, Latino (30%), Multi-racial (26%), Native American (4%), and Asian, Hawaiian, and Pacific Islander (3%). Half of the SMY were cisgender females (49%). The majority (67%) of SMY received free and reduced lunch. Response rates were 65% at W2 (n = 310), 53% at W3 (n = 254), and 49% at W4 (n = 235).

Measures

Depression symptoms.

The Beck Depression Inventory for Youth (BDI-Y) consists of twenty items (e.g. “I think that my life is bad”) measured on a four-point scale from (0) never to (3) always (Beck, 1996). A mean composite score was used, and higher scores represented greater levels of depression symptoms. The BDI-Y had excellent reliability across waves (See Table 1).

Table 1.

Correlations, Means, Standard Deviations and Cronbach Alpha for Longitudinal Variables

		1	2	3	4	5	6	7	8	9	10	11	12	13
1.	W1 Racial Discrimination	1.00
2.	W2 Racial Discrimination	0.51***	1.00
3.	W4 Racial Discrimination	0.34***	0.57***	1.00
4.	W1 Depression	0.31***	0.22***	0.11	1.00
5.	W2 Depression	0.27***	0.42***	0.25***	0.62***	1.00
6.	W3 Depression	0.17**	0.20**	0.15*	0.58***	0.58***	1.00
7.	W4 Depression	0.14*	0.21**	0.27***	0.39***	0.61***	0.65***	1.00
8.	W1 Suicidal Ideation	0.21***	0.13*	0.01	0.60***	0.33***	0.32***	0.28***	1.00
9.	W2 Suicidal Ideation	0.15**	0.19***	0.13	0.25***	0.46***	0.20**	0.33***	0.36***	1.00
10.	W3 Suicidal Ideation	0.09	0.08	0.08	0.28***	0.32***	0.53***	0.48***	0.43***	0.34***	1.00
11.	W4 Suicidal Ideation	0.09	0.09	0.05	0.27***	0.32***	0.30***	0.59***	0.42***	0.42***	0.51***	1.00
12.	W1 LGBT Victimization	0.30***	0.16**	0.11	0.21***	0.11*	−0.00	0.10	0.21***	0.23***	−0.00	0.23***	1.00
13	Age	−0.03	0.05	0.06	−0.08	−0.09	−0.04	−0.00	0.02	−0.04	0.02	0.10	0.10*	1.00
	Mean	0.95	0.76	0.70	0.73	0.64	0.62	0.66	0.48	0.25	0.29	0.34	0.67	19.00
	SD	0.91	0.92	0.96	0.62	0.54	0.53	0.63	0.78	0.52	0.58	0.63	0.76	1.79
	Range	0–4	0–4	0–4	0–2.8	0–2.75	0–2.9	0–2.85	0–4	0–3	0–3.63	0–3.5	0–3	15–24
	Alpha	0.68	0.78	0.85	0.95	0.94	0.95	0.96	0.93	0.92	0.94	0.94	0.85	-
	N	476	308	239	475	310	254	234	473	306	254	234	476	476

Note.

^*p < 0.05

^**p < 0.01

^***p < 0.001.

Suicide ideation.

The negative ideation subscale from the Positive and Negative Suicide Inventory was used to measure suicidal ideation in the past two weeks (Muehlenkamp et al., 2005). The eight-item subscale measured the frequency of suicidal thoughts (e.g., “How often have you seriously considered killing yourself because you could not live up to the expectations of other people?”) on a five-point scale from (0) none of the time to (4) most of the time. A mean composite score was used, and higher scores represented greater levels of suicidal ideation. The negative suicidal ideation subscale had good internal reliability across all waves (See Table 1).

Racial discrimination.

Two items from the Brief Racism and Life Experience and Stress Scale (RaLS-B) were used to assess racial discrimination in the past year: “how much have you personally experienced racism, racial discrimination or racial prejudice?” and “In general, how much stress has racism caused you?” (Harrell, 1995). These two items were selected because they were the only two items of the scale available longitudinally and referenced the same time frame, in the past year, as the items measured in later waves. Each item was rated on a five-point scale from not at all (0) to extremely (4). A mean score of the two items was computed, and higher scores represented greater racial discrimination. The RaLS-B had acceptable internal reliability at each wave (See Table 1).

LGB victimization.

Frequency of lifetime sexual orientation victimization at wave one was assessed with six items (e.g., verbal, physical, or sexual victimization due to being LGB) on a four-point scale (0) never to (3) three or more times (D’augelli et al., 2008). The mean score was used, and higher scores represented more frequent LGB victimization. The scale demonstrated good reliability (See Table 1). At waves two through four, the same questions were used but with reference to the past nine months. The frequency of victimization showed that the majority of the sample had a mean score of zero across the items after wave one: wave one (29.7%), wave two (52.6%), wave three (62%), and wave four (67.1%). The lower reports of victimization at later waves was likely due to the shorter referenced period (i.e., nine months) for the measure at waves two through four compared to the lifetime baseline measure.

Covariates.

Covariates included baseline age, site of data collection, race, baseline sexual identity, free/reduced lunch, and gender.

Attrition.

To better understand who completed the follow up waves of the study, participants who did and did not complete waves two through four were compared on baseline demographic characteristics and baseline measures of racial discrimination, LGB victimization, depression symptoms, and suicidal ideation. The results from the attrition analyses at each wave are shown in Appendices A – C. Mental health partly explained differences between participants who did and did not complete the follow up surveys. Participants with fewer depression symptoms and less suicidal ideation were more likely to have completed wave two; participants who competed wave three also reported less suicidal ideation. While there were no demographic differences between SMY who completed wave two, at waves three and four, gay and lesbian youth and youth with a higher SES were more likely to complete the surveys. Notably there were no differences by race, gender, age, LGB victimization or site of data collection at any of the follow up waves. Thus, the differences between those who completed and did not complete each wave were not consistent, but across waves youth who identified as gay or lesbian, had higher SES, and had less suicidal ideation tended to complete more waves of data. Given that missing data were partly explained by covariates, the missing cases were considered missing at random (Little, 2013). Variables for which there were differences between SMY who did and did not complete each wave were included in analyses, which in tandem with Full Information Maximum Likelihood Estimation (FIML) were used to estimate unbiased parameters and standard errors (Little, 2013).

Analysis Plan

Data were managed in Stata 15 (StataCorp, 2017), analyzed in Mplus 8 (Muthén & Muthén, 1998–2017), and visualized with the R package ggplot2 (Wickham, 2016). Model fit of the growth curve models was assessed with the Chi-Square Test, Root Mean Square Error of Approximation (RMSEA), Comparative Fit Index (CFI); and Standardized Root Mean Residual (SRMR). Model fit will be considered acceptable with a chi-square p-value greater than 0.05, a RMSEA below 0.08 and 90% confidence interval between 0 and 0.10, a CFI > .90, and a SRMR < 0.08 (Kline, 2016). Missing data were handled with FIML.

For racial discrimination, a linear model was fit, because there were only three waves of data for this variable. For depression symptoms and suicidal ideation, both linear and quadratic models were fit and compared using the chi-square test of difference with the Satorra-Benter Correction, and the better fitting model was selected (Satorra, 2000). Next, three sets of parallel process models (i.e., the general name for when the intercepts and slopes of two latent growth curve models are correlated or regressed on each other) of racial discrimination and mental health trajectories were estimated. In the first set of models, the intercept of racial discrimination predicted the intercept and slope of the mental health outcome (i.e., depression symptoms and suicidal ideation) and the slope of racial discrimination predicted the slope of the mental health outcome. In second third set of models, LGB victimization was added to predict the intercept and slope of each mental health outcome, which tested the additive and inuring hypotheses. In the final set of models, the interaction between the intercept of racial discrimination and LGB victimization was added to the parallel process model and tested the multiplicative hypothesis.

The latent interaction between baseline racial discrimination and baseline LGB victimization was estimated with the XWITH command in Mplus. Traditional model fit indices were not provided because the estimation of the latent interaction utilizes a random effects estimator, so the model with the interaction was compared to the model without the interaction with the likelihood ratio test (LRT) using the Satorra-Benter scaling correction (Satorra, 2000). The model without an interaction is a less complex model and uses fewer degrees of freedom, which should provide a worse fit to the data than the more complex model with an interaction. If the model fit was acceptable for the model without an interaction, and the LRT test was significant, this indicated that the interaction model should also provide acceptable model fit (Maslowsky et al., 2015). When interactions were significant, we utilized the Johnson-Neyman Technique (J-N) to identify regions of significance, which indicated at which values of the moderator the interaction was significant (Johnson & Neyman, 1936). Results from the J-N technique informed which values were used to plot the simple slope of significant interactions.

Results

Individual Growth Curve – Racial Discrimination

The linear model for racial discrimination with constrained residuals had poor fit (X²(3) = 13.77, scaling correction = 1.30, p < 0.01; RMSEA = 0.09, 90% CI [0.04, 0.14]; SRMR = 0.06; CFI = 0.92) due to a significant p-value and RMSEA confidence intervals that exceeded an upper limit of 0.10. Upon inspection of the model, there was a notable drop in mean scores between wave one and wave two that the linear model did not fully capture. The factor loading for the slope at wave two was freed so that it was estimated and provided a better fit to the data (Little, 2013). This approach changes the magnitude of the correlation between the intercept and slope as well as the magnitude of the slope coefficient, but the different specifications resulted in a similar substantive interpretation of the growth curve. The main difference was that freeing the wave two loading reduced the variability around the slope, which was likely artificially high when constraining the wave two factor loading to be one, because it overestimated the mean at wave two. Although a chi-square difference test did not indicate that the model without the loading free was a worse fitting model (p = 0.29), the model with the wave two loading freed had an acceptable RMESA and a non-significant p-value (X² (2) = 1.77, scaling correction = 1.38, p = 0.41; RMSEA = .00, 90% CI [0.00, 0.09]; SRMR = 0.03; CFI = 1.00). Thus, the model with the wave two loading freed was selected for further analysis.

Racial discrimination declined significantly over time (M_Intercept = 0.95, p < 0.001; M_Slope = −0.10, p < 0.001). There was significant variability around the intercept (SD = 0.71, p < 0.001) and slope (SD = 0.22, p < 0.001). There was also a significant negative correlation between the intercept and slope (r = −0.36, p < 0.001; See Table 2), indicating that SMY with higher baseline racial discrimination had greater declines in racial discrimination. For covariates, SMY from the Northeast site (b = −0.23, p = 0.04) had lower baseline levels of discrimination compared to the Southwest site. Further, sexual minority women had a less steep decline in racial discrimination (b = 0.08, p = 0.047) compared to sexual minority men (See Table 3).

Table 2.

Estimates for Unconditional Latent Growth Curves

		Mean	SE	SD	SE	rintercept and slope
Discrimination	Intercept	0.95***	0.04	0.71***	0.06	−0.36***
	Slope	−0.10***	0.02	0.22***	0.01
Depression	Intercept	0.71***	0.03	0.52***	0.04	−0.49***
	Slope	−0.03*	0.01	0.15**	0.01
Suicidal Ideation	Intercept	0.48***	0.04	0.64***	0.07	−0.86***
	Slope	−0.05**	0.02	0.12	0.01

Note.

Correlation is between intercept and slope within each trajectory.

^*p < 0.05

^**p < 0.01

^***p < 0.001.

Table 3.

Individual Growth Curve Models with Covariates

	Discrimination				Depression				Suicidal Ideation
	Intercept		Slope		Intercept		Slope		Intercept		Slope
Variables	b	SE	b	SE	b	SE	b	SE	b	SE	b	SE
Northeast site (ref = Southeast)	−0.23*	0.11	0.05	0.05	−0.19**	0.07	0.03	0.03	−0.21	0.11	0.05	0.04
Native American (ref = Black)	0.38	0.22	−0.14	0.10	0.31	0.17	0.06	0.08	0.20	0.20	0.02	0.06
Multiracial (ref = Black)	0.08	0.11	0.04	0.06	0.11	0.06	0.00	0.03	−0.08	0.09	0.03	0.03
Latino (ref = Black)	0.16	0.11	−0.01	0.05	0.16*	0.07	−0.03	0.04	−0.01	0.09	0.00	0.03
AHPI (ref = Black)	0.25	0.28	−0.03	0.09	0.11	0.19	0.04	0.06	0.08	0.24	0.01	0.06
Women (ref = Men)	−0.08	0.09	0.08*	0.04	0.05	0.06	0.04	0.03	0.03	0.07	0.00	0.02
Reduced Lunch (ref = No)	0.01	0.10	−0.03	0.05	0.01	0.06	−0.01	0.03	0.03	0.09	0.00	0.03
Age	0.00	0.03	0.02	0.01	−0.01	0.02	0.00	0.01	0.02	0.02	−0.01	0.01
Bisexual (ref = Gay)	0.07	0.09	0.01	0.04	0.20**	0.06	−0.06*	0.03	0.13	0.08	−0.03	0.02
Queer (ref = Gay)	0.35	0.27	0.13	0.16	0.20	0.13	0.08	0.06	−0.10	0.15	0.03	0.04

Note.

^*p < 0.05

^**p < 0.01

^***p < 0.001.

Individual Growth Curve – Depression symptoms

The linear model for depression symptoms (X²(5) = 10.02, scaling correction = 1.51, p = 0.08; RMSEA = 0.05, 90% CI [0.00, 0.09]; SRMR = 0.03; CFI = 0.98) did not have worse fit (TRd = 0.72, ∆df = 4, p = 0.95) compared to the quadratic model (X²(1) = 0.10, scaling correction = 1.04, p = 0.75; RMSEA = 0.00, 90% CI [0.00, 0.08]; SRMR = 0.00; CFI = 1.00). However, there were negative variances for the linear and quadratic slope of the quadratic model, so the linear model was selected for further analysis.

Depression declined significantly over time (M_Intercept = 0.71, p < 0.001; M_Slope = −0.03, p = 0.03), and there was significant variability around the intercept (SD = 0.52, p < 0.001) and slope (SD = 0.15, p < 0.01). There was as significant negative correlation between the intercept and slope (r = −.49, p <.001; See Table 2), which suggested that those who had higher baseline levels of depression symptoms had a steeper negative slope. For covariates, SMY from the Northeast site (b = −0.19, p = 0.01) had lower baseline depression symptoms compared to SMY from the Southwest site. Latinx SMY (b = 0.16, p = 0.03) and bisexual youth (b = 0.20, p < 0.01) had higher baseline depression symptoms compared Black SMY and gay and lesbian youth, respectively. Bisexual youth (b = −0.06, p = 0.02) also had a steeper decline in depression symptoms compared to gay and lesbian youth (See Table 3).

Finally, as part of testing the inuring hypothesis, LGB victimization was included as a predictor in the individual growth curve of depression symptoms with covariates to assess if it was independently associated with the intercept and slope of depression symptoms. The model had good fit (X²(30) = 41.51, scaling correction = 1.10, p = 0.08; RMSEA = 0.03, 90% CI [0.00, 0.05]; SRMR = 0.02; CFI = 0.97), and LGB victimization was positively associated with baseline depression symptoms (b = 0.20, p = < 0.001) and negatively associated with the slope of depression symptoms (b = −0.05, p = 0.03). That is, higher baseline levels of LGB victimization were associated with higher baseline levels and steeper declines in depression symptoms.

Individual Growth Curve –Suicidal Ideation

For the growth curve model of suicidal ideation, both linear (X²(5) = 25.09, scaling correction = 1.72, p <0.001; RMSEA = 0.09, 90% CI [0.06, 0.13]; SRMR = 0.09; CFI = 0.74) and quadratic (X²(1) = 3.57, scaling correction = 1.25, p = 0.06; RMSEA = 0.07, 90% CI [0.00, 0.16]; SRMR = 0.03; CFI = 0.97) models were initially tested. However, the linear model had poor fit and the quadratic model resulted in negative variances for the linear and quadratic slopes. As with the model for racial discrimination, there was a notable drop in mean values between wave one and wave two, so models with the factor loadings freed were tested. The final model selected was a linear model where wave two and wave three loadings were allowed to be estimated, and the residuals were constrained to be the same over time (X²(6) = 6.54, scaling correction = 2.27, p = 0.37; RMSEA = 0.01, 90% CI [0.00, 0.06]; SRMR = 0.06; CFI = 0.99).

Suicidal ideation declined significantly over time (M_Intercept = 0.48, p < 0.001; M_Slope = −0.05, p < 0.01), and there was significant variability around the intercept (SD = 0.64 p < 0.001), but not the slope (SD = 0.12, p = 0.25). There was as significant negative correlation between the intercept and slope (r = −0.86, p < 0.001; See Table 2), which indicated that SMY who had higher baseline levels of suicidal ideation had steeper declines in suicidal ideation over time. There were no significant covariates of the intercept or slope of suicidal ideation (See Table 3).

Finally, as part of testing for the inuring hypothesis, LGB victimization was included as a predictor of the growth curve of suicidal ideation with covariates to assess if it was independently associated with the intercept and slope of suicidal ideation. The model fit was marginal due to a CFI below 0.90 and significant p-value (X²(28) = 57.69, scaling correction = 1.20, p = 0.001; RMSEA = 0.05, 90% CI [0.03, 0.06]; SRMR = 0.04; CFI = 0.86). Higher baseline LGB victimization was associated with higher suicidal ideation (b = 0.25, p = < 0.001), but was not associated with the slope of suicidal ideation (b = −0.03, p = 0.12).

Parallel Process Model – Racial Discrimination and Depression symptoms

The parallel process model with baseline racial discrimination predicting baseline depression symptoms and the slope of depression symptoms, and the slope of racial discrimination predicting the slope of depression symptoms had good model fit (X²(49) = 91.47, scaling correction = 1.08, p < 0.001; RMSEA = 0.04, 90% CI [0.03, 0.06]; SRMR = 0.03; CFI = 0.94). Baseline racial discrimination was positively associated with baseline depression symptoms (b = 0.32, p < 0.001), which indicated that higher baseline racial discrimination was associated with higher baseline depression symptoms. The slope of racial discrimination was positively associated with the slope of depression symptoms (b = 0.30, p < 0.001), which suggested that steeper declines in racial discrimination were associated with less steep declines in depression symptoms over time. There was not a significant association between the baseline racial discrimination and the slope of depression symptoms (b = −0.04, p = 0.17). The associations between the covariates and the intercept and slope of depression symptoms were similar to those in the individual growth curve model for depression symptoms (See Table 4).

Table 4.

Parallel Process Model Prediction with Covariates

	Depression				Suicidal Ideation
	Intercept		Slope		Intercept		Slope
Variable	b	SE	b	SE	b	SE	b	SE
Discrimination Int	0.32***	0.06	−0.03	0.03	0.26***	0.07	−0.03	0.02
Discrimination Slope	--	--	0.30**	0.10	--	--	0.09	0.06
Northeast site (ref = Southeast)	−0.11	0.07	0.01	0.03	−0.15	0.11	0.04	0.04
Native American (ref = Black)	0.19	0.16	0.11	0.07	0.09	0.21	0.04	0.06
Multiracial (ref = Black)	0.08	0.06	0.00	0.03	−0.10	0.09	0.03	0.03
Latino (ref = Black)	0.11	0.07	−0.02	0.04	−0.05	0.09	0.00	0.03
AHPI (ref = Black)	0.03	0.14	0.06	0.06	0.01	0.20	0.02	0.05
Women (ref = Men)	0.08	0.05	0.01	0.03	0.05	0.07	−0.01	0.02
Reduced Lunch (ref = No)	0.01	0.06	−0.01	0.03	0.03	0.09	0.00	0.02
Age	−0.01	0.02	0.00	0.01	0.02	0.02	−0.01	0.01
Bisexual (ref = Gay)	0.17**	0.05	−0.07**	0.03	0.11	0.08	−0.02	0.02
Queer (ref = Gay)	0.09	0.16	0.06	0.08	−0.19	0.15	0.03	0.05

Note.

^*p < 0.05

^**p < 0.01

^***p < 0.001.

The parallel process model extended to include a main effect of LGB victimization predicting the intercept and slope of depression symptoms had good model fit (X²(52) = 98.77, scaling correction = 1.08, p < 0.001; RMSEA = 0.04, 90% CI [0.03, 0.06]; SRMR = 0.03; CFI = 0.94). LGB victimization was associated with baseline depression symptoms (b = 0.10, p = 0.02) but not the slope of depression symptoms (b = −0.02, p = 0.40). Notably, baseline racial discrimination remained associated with baseline depression symptoms (b = 0.28, p < 0.001) and remained non-significantly associated with slope of depression symptoms (b = −0.03, p = 0.33). Further, the slope of racial discrimination remained significantly associated with the slope of depression symptoms (b = 0.30, p < 0.001 See Table 5). Thus, these results would not indicate support for the inuring hypotheses and align with the additive hypothesis pending tests for an interaction between racial discrimination and LGB victimization.

Table 5.

Parallel Process Model Prediction Main Effect of LGB Victimization and Interaction with Covariates

	Depression								Suicidal Ideation
	Intercept		Slope		Intercept		Slope		Intercept		Slope		Intercept		Slope
Variable	b	SE	b	SE	b	SE	b	SE	b	SE	b	SE	b	SE	b	SE
Discrimination Int	0.28***	0.07	−0.03	0.03	0.28***	0.07	−0.03	0.03	0.19*	0.08	−0.02	0.02	0.16	0.11	−0.01	0.06
Discrimination Slope	--	--	0.30**	0.10	--	--	0.33**	0.11	--	--	0.09	0.06	--	--	0.05	0.12
LGB Victimization	0.10*	0.04	−0.02	0.03	−0.08	0.08	0.07	0.06	0.18**	0.06	−0.02	0.02	−0.07	0.21	−0.04	0.16
Discrimination Int X Victim	--	--	--	--	0.15*	0.07	-0.08*	0.04	--	--	--	--	0.21	0.17	0.01	0.15
Northeast site (ref = Southeast)	−0.13	0.07	0.02	0.03	−0.14	0.07	0.02	0.03	−0.18	0.11	0.04	0.04	−0.17	0.16	−0.04	0.16
Native American (ref = Black)	0.17	0.16	0.12	0.07	0.17	0.16	0.13	0.07	0.07	0.20	0.04	0.06	0.08	0.25	0.04	0.07
Multiracial (ref = Black)	0.07	0.06	0.00	0.03	0.08	0.06	−0.01	0.03	−0.13	0.09	0.03	0.03	−0.11	0.09	0.02	0.11
Latino (ref = Black)	0.10	0.07	−0.02	0.04	0.10	0.07	−0.02	0.04	−0.06	0.09	0.00	0.03	−0.06	0.09	0.03	0.04
AHPI (ref = Black)	0.04	0.14	0.06	0.06	0.05	0.13	0.05	0.06	0.01	0.20	0.02	0.05	0.07	0.21	0.00	0.02
Women (ref = Men)	0.10	0.05	0.01	0.03	0.10	0.05	0.01	0.03	0.09	0.07	−0.02	0.02	0.09	0.07	0.02	0.06
Reduced Lunch (ref = No)	0.00	0.06	0.00	0.03	0.00	0.06	0.00	0.03	0.01	0.09	0.00	0.02	−0.01	0.10	−0.02	0.02
Age	−0.01	0.02	0.00	0.01	−0.01	0.02	0.00	0.01	0.02	0.02	−0.01	0.01	0.02	0.02	0.01	0.02
Bisexual (ref = Gay)	0.20***	0.05	−0.08**	0.03	0.20***	0.05	−0.08**	0.03	0.16*	0.08	−0.03	0.02	0.17*	0.08	−0.01	0.01
Queer (ref = Gay)	0.13	0.15	0.05	0.08	0.15	0.14	0.03	0.08	−0.11	0.14	0.02	0.05	−0.06	0.13	−0.03	0.02

Note.

^*p < 0.05

^**p < 0.01

^***p < 0.001.

The parallel process model with the main effects of the intercept of racial discrimination and LGB victimization (−2LL = −4878.99, −2LL correction factor = 1.22, parameters = 137) had a worse fit compared to the model with the interactions (−2LL = −4873.64, −2LL correction factor = 1.21, parameters = 139; TRd = 9.12, ∆df = 2, p = 0.01). The interaction between baseline racial discrimination and LGB victimization was associated with baseline depression symptoms (b = 0.15, p = 0.03) and the slope of depression symptoms (b = −0.08, p = 0.04; See Table 5). The J-N technique was utilized to identify regions of significance across all values of the moderator (Johnson & Neyman, 1936; i.e., LGB victimization). For the association between the interaction and the intercept of depression symptoms, the interaction was significant at all values of LGB victimization—therefore the values were plotted at the lowest possible value of LGB victimization (−0.63 SD) below the mean and one unit higher (+0.67 SD) above the mean of LGB victimization (Figure 1, Panel A). As shown in Figure 1, the simple slopes analyses indicated that while both slopes were significant: at high values of LGB victimization (b = 0.38, p < 0.001) the association between baseline racial discrimination and baseline depression symptoms was twice the strength than at low values of LGB victimization (b = 0.19, p = 0.01). For the slope of depression symptoms, the interaction was only significant at values above one for LGB victimization—thus, the low value of the simple slopes was set at the lowest possible value of LGB victimization (−0.67 SD), and at the value where the slope becomes significant (+1 SD) of LGB victimization (Figure 1, Panel B). The simple slope analysis indicated that at high levels of LGB victimization, the association between baseline racial discrimination and the slope of depression symptoms becomes stronger leading to a steeper decline in depression symptoms. Taken together, the results for depression symptoms aligned with the multiplicative hypothesis.

Figure 1.

Interaction between LGB Victimization and the Intercept of Racial Discrimination Predicting the Intercept of Depressive Symptoms (A) and Interaction between LGB Victimization and the Intercept of Racial Discrimination Predicting the Slope of Depressive Symptoms (B).

Parallel Process Model – Racial Discrimination and Suicidal Ideation

The parallel process model with the intercept and slope of racial discrimination predicting the intercept and slope of suicidal ideation had good model fit (X²(47) = 72.06, scaling correction = 1.12, p = 0.01; RMSEA = 0.03, 90% CI [0.02, 0.05]; SRMR = 0.04; CFI = 0.93). Baseline racial discrimination was positively associated with baseline suicidal ideation (b = 0.26, p < 0.001), which suggested that higher baseline racial discrimination was associated with higher baseline suicidal ideation. The intercept (b = −0.03, p = 0.12) and the slope (b = 0.09, p = 0.12) of racial discrimination were not associated with the slope of suicidal ideation. Estimates for covariates were consistent with the individual trajectory of suicidal ideation (Table 4).

The parallel process model extended to include a main effect of LGB victimization predicting the intercept and slope of suicidal ideation had good model fit (X²(50) = 94.20, scaling correction = 1.11, p <0.001; RMSEA = 0.04, 90% CI [0.03, 0.06]; SRMR = 0.04; CFI = 0.90). Higher baseline levels of LGB victimization were associated with higher baseline suicidal ideation (b = 0.18, p <0.001), but LGB victimization was not associated with the slope of suicidal ideation (b = −0.02, p = 0.39). Baseline racial discrimination remained associated with baseline suicidal ideation (b = 0.19, p <0.001). Estimates for covariates in this model were similar to those in previous models (See Table 5). Thus, these results did not indicate support for the inuring hypotheses, and the results with baseline suicidal ideation as the outcome, indicated support for the additive hypothesis pending tests for the multiplicative hypothesis.

The parallel process model with the main effects of the baseline racial discrimination and LGB victimization (−2LL = −5137.45, −2LL correction factor = 1.30, parameters = 139) compared to the model with the interaction (−2LL = −5129.36, −2LL correction factor = 1.46, parameters = 141) was not a worse fit (TRd = 1.24, ∆df = 2, p = 0.54). There was not a significant interaction between the intercept of racial discrimination and LGB victimization for baseline suicidal ideation (b = 0.21, p = 0.12) or the slope of suicidal ideation (b = 0.01, p = 0.94; See Table 5), which indicated that neither the intercept nor slope of suicidal ideation were associated with the overlap in racial discrimination and LGB victimization. Given that no significant interaction was found between racial discrimination and LGB victimization the results for suicidal ideation aligned with the additive hypothesis.

Sensitivity Analyses

Some unexpected findings were that greater levels of baseline racial discrimination and LGB victimization were not associated with changes in suicidal ideation and that, in the context of high levels of LGB victimization, higher levels of racial discrimination were associated with a steeper decline in depression symptoms. Yet prior research from cross-sectional and longitudinal studies suggesting positive associations between racial discrimination and LGB victimization with worse mental health (Birkett et al., 2015; Thoma & Huebner, 2013). Thus, a respecified longitudinal model that incorporated baseline racial discrimination and LGB victimization as lagged covariate of depression symptoms and suicidal ideation at each wave was estimated to assess if the association between baseline racial discrimination, baseline LGB victimization, and their interaction were associated with mental health at specific time points in the growth curve model. This alternative model specification helps to elucidate if the strength of the associations between racial discrimination and LGB victimization with mental health differed across time points, which may explain why there was no association between racial discrimination and LGB victimization with suicidal ideation, and why there were declines in depression symptoms at high levels of LGB and racial discrimination. Model fits for these models can be found in Appendix D and results in Appendix E.

For depression, baseline racial discrimination and LGB victimization were both associated with depression symptoms at waves one and two. At wave three only baseline racial discrimination remained associated with wave three depression symptoms. Neither racial discrimination nor LGB victimization were associated with wave four depression. Notably, the associations between baseline racial discrimination and LGB victimization were generally weaker at later waves, suggesting that the early negative associations of racial discrimination and LGB victimization grow weaker over time. There were no interactions between baseline racial discrimination and baseline LGB victimization and depression at waves one, two, three or four. The lack of significant interactions suggests that the interaction between baseline racial discrimination and LGB victimization and baseline depression symptoms should be interpreted with caution, and that the association between the interaction was related to changes in depression symptoms over time rather than depression symptoms at a specific wave.

For the suicidal ideation, racial discrimination and LGB victimization were both associated with suicidal ideation at wave one. Racial discrimination was not associated with depression symptoms at wave two, three or four. LGB victimization was associated with suicidal ideation at waves two and four. The associations between racial discrimination and depression generally grew weaker over time, while the association between LGB victimization and suicidal ideation grew stronger over time except for at wave three. The lack of an association between racial discrimination and suicidal ideation at later waves and inconsistent associations between LGB victimization and suicidal ideation may explain why there were no associations between baseline racial discrimination and LGB victimization with changes in suicidal ideation. The model examining the interaction between baseline racial discrimination and LGB victimization on each wave of suicidal ideation could not be interpreted due to poor model fit as indicated by a p-value less than .05 and a CFI lower than .90, which suggests that this specification of the model with a lagged interaction was not a good fit to the data.

Discussion

SMY are at greater risk for compromised mental health and experiencing victimization due to their sexual identity (Burton et al., 2013). SMY of color may be vulnerable to both racial discrimination and LGB victimization, which can exacerbate their risk for compromised mental health (Thoma & Huebner, 2013). The current study addressed the limitations of previous cross-sectional research that has neglected to assess multiple forms of discrimination among SMY by exploring three hypotheses, the additive, multiplicative hypothesis, and an inuring hypothesis, to understand how racial discrimination and LGB victimization might overlap in their association with the mental health of SMY of color over three years. First, the results of the study align with research in other samples and show that in a racially diverse sample of SMY of color, racial discrimination, depression symptoms, and suicidal ideation declined over time. Second the results of the study supported the multiplicative hypothesis for depression symptoms such that at high levels of LGB victimization, the association between racial discrimination and the intercept and the slope of depression symptoms was stronger, indicating higher baseline and a steeper decline in depression symptoms. Lastly, for suicidal ideation, the results supported the additive hypotheses and indicated that greater racial discrimination and LGB victimization were both associated with greater levels of suicidal ideation when tested simultaneously. These results also extend previous findings in two distinct ways.

First, although this study found patterns of racial discrimination depression symptoms, and suicidal ideation that are consistent with other studies, these unconditional trajectories do not account for all the relevant factors that have been shown to affect them (Birkett et al., 2015; Hughes et al., 2016). Yet the findings do point to their being a developmental process that typically leads to improved mental health as youth move into early adulthood. The findings highlight the need for concern for the well-being of SMY of color when they experience stigma, but are hopeful in pointing to recovery or resilience across time (Kuper et al., 2014).

For racial discrimination, prior studies among racial minority youth have considered demographic differences and differential exposure to racial discrimination to explain changes in reports of racial discrimination (e.g., Hughes et al., 2016). However, there were few demographic differences in the intercepts and slopes of the racial discrimination. Baseline racial discrimination differed for SMY from the Northeast site compared to the Southwest site—potentially due to differences in local racial diversity and access to LGBT communities of color (McConnell et al., 2018). Additionally, sexual minority women had a less steep decline in racial discrimination compared sexual minority men, and thus had higher overall levels of racial discrimination across time. This finding aligns with the notion that racial discrimination is also gendered (Lewis & Neville, 2015).

Depression symptoms also declined over time, and there were several important demographic differences in this trajectory. Compared to gay youth, bisexual youth had higher levels of depression symptoms, but also a steeper decline in depression symptoms. The higher baseline level of depressive symptomology among bisexual youth compared to gay youth is consistent with previous research, and adds to the body of research the finding that there is variability in how depression symptoms change over time between bisexual and gay or lesbian youth (Ross et al., 2018). Because this pattern was consistent across models that included racial discrimination and LGB victimization, future studies should further explore the overlap in racial discrimination and sexual orientation-based discrimination or victimization—particularly bi-stigma—which may explain the higher risk for compromised mental health among bisexual youth (Ross et al., 2018).

The decline in suicidal ideation was not associated with any covariates, which suggested that changes in suicidality did not differ by demographics in the current study. This finding does not align with cross-sectional research that typically finds racial and ethnic differences in mental health among SMY (Bostwick et al., 2014). Similarly, studies on age differences in suicidality also indicate racial ethnic difference among sexual minorities persist into adulthood (Layland et al., 2020). However, across these studies, racial and ethnic sexual minorities, particular Black and Latinx SMY tend to have lower rates of suicidal ideation compared to White SMY, which may help explain the lack of demographic differences in suicidal ideation in a sample of SMY of color (Bostwick et al., 2014).

The second contribution of this study was the examination of the overlap of racial discrimination and LGB victimization and their longitudinal associations with mental health utilizing three intersectional hypotheses of how racial discrimination and mental health may overlap in their associations with mental health. The results from the parallel process models contributed to the understanding of how racial discrimination and LGB victimization overlap to affect SMY of color’s long-term mental health and contributed to the literature on intersectionality and health more broadly. Importantly, support for the additive and multiplicative hypotheses varied across mental health outcomes. This suggests that a monolithic hypothesis about how intersections of discrimination affect mental health likely obscures important complexity in how intersectional stigma affects distinct aspects of mental health.

For depression symptoms, the multiplicative hypothesis was supported, but with an important caveat. At high levels of LGB victimization and racial discrimination, SMY had higher initial levels of depression symptoms, but also steeper declines in depression symptoms. Thus, SMY of color who experience both racial discrimination and LGB victimization were at a substantially higher risk for depression symptoms earlier in their life course. SMY who experienced an overlap in racial discrimination and LGB victimization had a steeper decline in depression symptoms over time; this decline may be attributed to a higher baseline score. The steeper declines in depression symptoms in the context of racial discrimination and LGB victimization seems counter intuitive; however, given that SMY who experienced both racial discrimination and LGB victimization started out higher at baseline, they also had a greater distance to cover in order to “catch up” with more typical levels of depression symptoms over time. Depression symptoms declined over time even when they were elevated earlier in time—a pattern of shock and recovery. Thus, although SMY experiencing high levels of LGB victimization and racial discrimination eventually “catch up” in terms of mental health with peers experiencing less stigma, they may stay at higher levels of depressive symptomology for a longer period of time, which could increase risk for correlates of depression symptoms such as substance abuse and suicidality (Burns et al., 2015). The exacerbating risk for depression symptoms among SMY of color is particularly troubling given that SMY of color can experience LGB victimization outside of the LGB community and racial discrimination with the LGB community (Ghabrial, 2017; Kuper et al., 2014). However, it is also important to note that even in the face of these overlapping sources of stigma, the decline in depression symptoms also points to the resilience of SMY of color in the face of intersectional stigma. For youth of color, family racial socialization and a positive racial-ethnic identity buffer the negative effects of racial discrimination and promote mental health (Phinney & Chavira, 1995)—the same may be true for SMY of color (Ghabrial, 2017; Kuper et al., 2014).

The additive hypothesis was supported for initial levels of suicidal ideation, but not the slope. The independent positive associations between both racial discrimination and LGB victimization with suicidal ideation highlights their additive associations, a finding more in line with the double jeopardy (Beal, 1969) framing of intersectionality. One way to think about the independent associations of racial discrimination and LGB victimization is that SMY of color have “twice” the risk of having heightened levels of suicidal ideation; SMY of color are exposed to both racial discrimination and LGB victimization, while White sexual minority youth face risk “only” for LGB victimization. This is an important framing of the additive hypothesis, which is often discounted as not intersectional (Bowleg, 2008), yet SMY of color have twice the “opportunity” of exposure to risk for higher levels of suicidal ideation, clearly an intersectional concern. The findings related to suicidal ideation also elucidate some of the mixed findings in the literature regarding racial differences in risk for suicidal ideation between white SMY and SMY of color—it may be a question of whether SMY are experiencing one or more forms of stigma.

Lastly, the non-significant associations between racial discrimination and LGB victimization with the trajectory of suicidal ideation may imply that the negative additive effects of racial discrimination and LGB victimizations for suicidal ideation are concurrent rather than prospective—this would make sense because that suicidal ideation was assessed in the past two weeks. It is likely that the negative association between racial discrimination and LGB victimization may be more acute and difficult to capture with too long of a lagged variable (Burton et al., 2013). However, given that this is one of the first studies to test these intersectional hypotheses longitudinally, this association should be further investigated.

As with all studies, there were several limitations. First, racial discrimination and suicidal ideation had substantial drops in scores between waves one and two. This required that the time points in the latent growth curve to be estimated in order to improve model fit. While this strategy is not uncommon in growth curve models, the time points used to estimate a latent growth curve can affect the estimation of the intercept and slope (Little, 2013). Future studies should seek to replicate these findings in another longitudinal study with SMY of color using different measures of racial discrimination and suicidal ideation—in particular, the use of the two items for the RaLES-B was a limitation of this study. The RaLES-B has not been used in samples of SMY of color, or longitudinally, so future studies should further evaluate the utility of this two-item measure and the full scale longitudinally. Additionally, collecting more waves of data with a shorter interval between each wave might also help to address the attrition, which was also a limitation of the current study. For suicidal ideation, the positive and negative suicide inventory is a well-validated measure that has been used longitudinally, so the drop in symptoms may be partly explained by lower levels of mental health symptomatology at baseline for those who completed the second wave of the study (Muehlenkamp et al., 2005). Future longitudinal studies of SMY of color should consider ways to retain SMY with higher baseline levels of compromised mental health.

Although declines in racial discrimination, depression symptoms, and suicidal ideation were consistent with previous research (Birkett et al., 2015; Hughes et al., 2016), it has also been suggested that when constructs that assess negative experiences such as discrimination are assessed over time and have a small range of scores, they tend to decline over time (Little, 2013). These types of measures are also more sensitive to declines over time because people are less likely to report stable negative states, or to report regularly increasing negative states, and may not fully recall all of their negative experiences when there is substantial time between waves of data collection (Little, 2013). In other words, experiences of racial discrimination, depression symptoms, and suicidal ideation are time sensitive experiences that may require shorter follow up in order to better capture change over time.

Ideally, a longitudinal test of additive, inuring, and multiplicative hypotheses would be done separately by racial/ethnic group or sexual identity to account for the unique experiences of sexual minority and/or racial minority youth. Unfortunately, it was not possible to test such models due to sample size limitations. In comparison to most previous studies, ours included multiple racial/ethnic, and sexual minority groups, while prior studies usually focused on one race/ethnic group, typically African American youth (Gattis, 2013; Grollman, 2012; Thoma & Huebner, 2013) or sexual minority group (Hightow-Weidman et al., 2011). Given this distinction, the findings from this study should be further explored in a larger and more diverse samples to allow for understanding these processes among specific groups.

Racial discrimination and LGB victimization capture distinct forms of stigma and exclusion. The measure of racial discrimination utilized in the current study assessed perceptions of racial discrimination and stress related to it, while the measure of victimization measured the frequency of explicit acts of verbal and physical violence tied to being LGB. An insidious component of measuring identity-based discrimination is that it is not always explicit, and suspicions of unfair treatment due to having a minoritized identity may arise in ambiguous situations (Meyer, 2003a). In contrast, it is not ambiguous when someone threatens you or physically assaults you because of your identity. The explicit nature of victimization may be difficult to measure because it is less likely to occur in shorter (e.g. past nine-months) rather than longer (e.g. ever) reference periods—a pattern consistent with findings in the current study, and has been noted in other longitudinal studies (e.g. Birkett et al., 2015). Racial discrimination and experiences of LGB victimization are not monolithic, and future studies should account for this variability and consider testing for multiple trajectories of racial discrimination and LGB victimization in order to understand how variability in these experiences may differently affect SMY’s mental health. When testing how intersections of stigma matter for mental health, the similarity in how stigma is measured may be important to consider, because how multiple forms of stigma overlap may vary based on which aspects of discrimination and victimization are measured.

Conclusion

Racial discrimination and LGB victimization have consistently and independently been shown to be associated with worse mental health for racial minority youth and SMY respectively (Birkett et al., 2015; Hughes et al., 2016). Yet their overlap and association with mental health for SMY of color has received relatively little attention or been examined with longitudinal data. Informed by intersectionality and minority stress theories, the present study utilized longitudinal data from a diverse sample of SMY of color to advance understandings about the link between intersectional discrimination and mental health. The findings of the present study indicated that racial discrimination, depression symptoms, and suicidal ideation declined over time, that high levels of racial discrimination and LGB victimization doubled the strength of the association between racial discrimination and LGB victimization with depression symptoms concurrently and prospectively, and that racial discrimination and LGB victimization were independently associated with suicidal ideation concurrently. Together the finding of the present study illustrates normative trends in racial discrimination, depression symptoms, and suicidal ideation among sexual minority adolescents and young adult of color, but also unique vulnerabilities and resiliency during critical periods of development. Moreover, the study findings highlight unique stressors for the mental health of SMY of color may inform prevention and intervention efforts that seek to promote the mental well being of SMY during adolescence and young adulthood.

Biography

Allen B. Mallory is a Presidential Postdoctoral Scholar at The Ohio State University. His major research interests include intersectionality and health among sexual and gender diverse people.

Stephen T. Russell is a Professor at the University of Texas at Austin. His major research interests include the health and well-being so sexual and gender diverse youth.

Appendix

Appendix A. Attrition Analysis for Wave 2

	Not Completed		Completed
	n	Mean (SD)	n	Mean (SD)	t(df)	p-value
W1 Depressive Symptoms (N = 475)	166	0.81 (0.69)	309	0.68 (0.58)	1.94 (290.05)	0.05
W1 Negative Ideation (N = 473)	166	0.63 (0.84)	307	0.39 (0.73)	3.10 (300.21) **	0.002
W1 Racial Discrimination (N = 476)	166	0.81 (0.81)	310	1.03 (0.94)	−2.61 (382.44) **	0.01
W1 LGB Victimization (N = 476)	166	0.75 (0.87)	310	0.63 (0.70)	1.50 (278.82)	0.14
W1Age (N = 476)	166	18.82 (1.87)	310	19.07 (1.74)	−1.42 (317.42)	0.15
	Not Completed		Completed
	n	%	n	%	X2(df)	p-value
Northeast Site (n = 368)	132	79.5	236	76.1	0.71 (1)	0.40
Black (n = 173)	61	36.8	110	35.5	0.08 (1)	0.78
Native American (n =19)	6	3.6	13	4.2	0.10 (1)	0.76
Multiracial (n = 125)	43	25.9	82	26.5	0.02 (1)	0.89
Latino (n = 137)	47	29.0	90	29.4	0.01 (1)	0.93
AHPI (n = 16)	5	3.0	11	3.5	0.10 (1)	0.76
Cisgender female (n = 231)	76	45.8	155	50.0	0.77 (1)	0.38
Gay/Lesbian (n = 276)	90	54.2	186	60.0	1.48 (1)	0.22
Bisexual (n = 181)	67	40.4	114	36.8	0.59 (1)	0.44
Queer (n = 19)	9	5.4	10	3.2	1.36 (1)	0.24
Free and reduced lunch (n = 316)	119	73.0	198	64.3	3.53 (1)	0.06

Note.

T-tests were calculated for unequal variance between groups. Reference group of Northeast site was the Southwest site.

^*p < 0.05

^**p < 0.01

^***p < 0.001.

Appendix B. Attrition Analysis for Wave 3

	Not Completed		Completed
	n	Mean (SD)	n	Mean (SD)	t (df)	p-value
W1 Depressive Symptoms (N = 475)	221	0.78 (0.68)	254	0.68 (0.57)	1.78 (433.29)	0.08
W1 Negative Ideation (N = 473)	220	0.57 (0.80)	253	0.39 (0.75)	2.50 (453.36) *	0.01
W1 Racial Discrimination (N = 476)	222	0.91 (0.86)	254	1.00 (0.95)	−1.10 (473.33)	0.27
W1 LGB Victimization (N = 476)	222	0.75 (0.83)	254	0.61 (0.69)	1.91 (430.32)	0.06
W1Age (N = 476)	222	18.90 (1.80)	254	19.06 (1.79)	−0.95 (464.89)	0.34
	Not Completed		Completed
	n	%	n	%	X2(df)	p -value
Northeast Site (n = 368)	179	80.6	189	74.4	2.61 (1)	0.11
Black (n = 173)	89	40.1	82	32.3	3.14 (1)	0.08
Native American (n =19)	11	4.0	8	3.2	1.01 (1)	0.32
Multiracial (n = 125)	53	23.9	72	28.4	1.22 (1)	0.27
Latino (n = 137)	59	27.2	78	31.1	0.85 (1)	0.36
AHPI (n = 16)	5	2.3	11	4.3	1.58 (1)	0.21
Cisgender female (n = 231)	103	46.4	128	50.4	0.76 (1)	0.38
Gay/Lesbian (n = 276)	110	49.6	166	65.4	12.15 (1) ***	<0.001
Bisexual (n = 181)	99	44.6	82	32.3	7.62 (1) **	0.01
Queer (n = 19)	13	5.7	6	2.4	3.77 (1)	0.05
Free and reduced lunch (n = 316)	161	73.9	155	61.5	8.09 (1) **	0.004

Note.

T-tests were calculated for unequal variance between groups. Reference group of Northeast site was the Southwest site.

^*p < 0.05

^**p < 0.01

^***p < 0.001.

Appendix C. Attrition Analysis for Wave 4

	Not Completed		Completed
	n	Mean (SD)	n	Mean (SD)	t(df)	p-value
W1 Depressive Symptoms (N = 475)	240	0.75 (0.64)	235	0.70 (0.61)	0.73 (472.75)	0.47
W1 Negative Ideation (N = 473)	239	0.52 (0.75)	234	0.43 (0.81)	1.10 (467.03)	0.27
W1 Racial Discrimination (N = 476)	241	0.95 (0.89)	235	0.96 (0.92)	−0.14 (472.01)	0.89
W1 LGB Victimization (N = 476)	241	0.71 (0.81)	235	0.63 (0.71)	1.13 (468.02)	0.26
W1Age (N = 476)	241	18.86 (1.84)	235	19.12 (1.74)	−1.60 (473.67)	0.11
	Not Completed		Completed
	n	%	n	%	X2(df)	p-value
Northeast Site (n = 368)	194	80.5	174	74.0	2.83 (1)	0.09
Black (n = 171)	90	37.3	81	34.5	0.43 (1)	0.51
Native American (n =19)	14	5.8	5	2.1	4.21 (1) *	0.04
Multiracial (n = 125)	56	23.2	69	29.4	2.31 (1)	0.13
Latino (n = 137)	69	29.2	68	29.3	0.003 (1)	0.99
AHPI (n = 16)	7	2.9	9	3.8	0.31 (1)	0.58
Cisgender female (n = 231)	112	46.5	119	50.6	0.83 (1)	0.36
Gay/Lesbian (n = 276)	127	52.7	149	63.4	5.60 (1) *	0.02
Bisexual (n = 181)	101	41.9	80	34.0	3.12 (1)	0.08
Queer (n = 19)	13	5.4	6	2.6	2.51 (1)	0.11
Free and reduced lunch (n = 316)	178	75.1	138	59.2	13.45 (1) ***	<0.001

Note.

T-tests were calculated for unequal variance between groups.

Reference group of Northeast site was the Southwest site.

^*p < 0.05

^**p < 0.01

^***p < 0.001.

Appendix D. Model Fit for Sensitivity Analyses

Title	X2	df	p-value	Scaling Correction	RMSEA	90%CI LL	90%CI UL	SRMR	CFI
Depression Symptoms w Lagged W1 Racial Discrimination and LGB Victimization	35.105	25	0.09	1.01	0.03	0.00	0.05	0.02	0.98
Depression Symptoms w Lagged W1 Racial Discrimination, LGB Victimization, and Interaction	34.258	25	0.10	1.00	0.03	0.00	0.05	0.02	0.98
Suicidal Ideation w Lagged W1 Racial Discrimination and LGB Victimization	42.351	26	0.02	1.18	0.04	0.01	0.06	0.03	0.93
Suicidal Ideation w Lagged W1 Racial Discrimination, LGB Victimization, and Interaction	80.743	28	0.00	1.40	0.06	0.05	0.08	0.05	0.77

Note.

RMSEA = Root Mean Square Error of Approximation.

CFI = Comparative Fit Index.

SRMR Standardized Root Mean Residual.

Appendix E. Sensitivity Analysis - Alternative Model

	Depression				Suicidal Ideation
	Main Effects		Interactions		Main Effects		Interactionsa
Variables	β	SE	β	SE	β	SE	b	SE
W1 Racial Discrimination → W1 Outcome	0.22***	0.05	0.22***	0.05	0.14**	0.05	--	--
W1 LGB Victimization → W1 Outcome	0.20***	0.05	0.18***	0.05	0.20***	0.05	--	--
W1 Racial Discrimination → W2 Outcome	0.19***	0.06	0.19***	0.05	0.09	0.05	--	--
W1 LGB Victimization → W2 Outcome	0.12*	0.06	0.11	0.06	0.22***	0.07	--	--
W1 Racial Discrimination → W3 Outcome	0.12*	0.05	0.11*	0.05	0.09	0.07	--	--
W1 LGB Victimization → W3 Outcome	0.01	0.06	0.03	0.06	−0.03	0.06	--	--
W1 Racial Discrimination → W4 Outcome	0.10	0.07	0.09	0.07	0.02	0.07	--	--
W1 LGB Victimization → W4 Outcome	0.11	0.09	0.11	0.09	0.28***	0.10	--	--
W1 Discrimination X Victimization → W1 Outcome	--	--	0.09	0.05	--	--	--	--
W1 Discrimination X Victimization → W2 Outcome	--	--	0.05	0.06	--	--	--	--
W1 Discrimination X Victimization → W3 Outcome	--	--	-0.05	0.05	--	--	--	--
W1 Discrimination X Victimization → W4 Outcome	--	--	0.00	0.08	--	--	--	--

Note.

^aNot shown due to poor model fit. Coefficients are standardized.

^*p < 0.05

^**p < 0.01

^***p < 0.001.