Substance Use and Discrimination in a Sample of U.S.-Based Latinx Sexual Minority Men and Their Main Partners

Abstract

Objective:

Substance use, including drug and alcohol misuse, is associated with myriad health conditions, including a higher risk for HIV infection. Although preliminary evidence suggests that higher levels of relationship functioning can buffer against the deleterious health consequences of discrimination on mental health broadly, such protective associations have been understudied with respect to alcohol and drug use. The topic is particularly understudied among Latino sexual minority men even though they are at greater risk for problematic substance use behaviors and are likely to experience multiple forms of discrimination (e.g., racism, homophobia).

Method:

To address this gap in the literature, we sampled 95 predominantly Latino sexual minority male couples to assess their drinking and drug use behaviors, relationship functioning, and experiences of discrimination. We used Actor–Partner Interdependence models to test our hypotheses.

Results:

We found that having a partner who experienced discrimination and higher partner reports of relationship functioning buffered against the negative relationship between own experiences of discrimination and drug use, but not problematic drinking.

Conclusions:

Our results suggest that higher relationship functioning serves as a buffer between the negative ramifications of discrimination on drug use, but not problematic drinking. We explicate implications for policy and practice to facilitate well-being among coupled Latino sexual minority men.

Substance misuse, encompassing drug and alcohol misuse, is one of the leading causes of preventable death in the United States (McGinnis & Foege, 1993; Mokdad et al., 2004; U.S. Burden of Disease Collaborators et al., 2018) and is associated with myriad health issues (Keaney et al., 2011), including increased risk for HIV infection (Norman et al., 2009; Wang & Maher, 2019). Although substance misuse is not unique to any single population in the United States, reports suggest that some populations are at elevated risk. For example, an emerging body of evidence suggests that Latino sexual minority men (SMM) are at increased risk for drug and alcohol misuse compared with their heterosexual Latino/x and non-Latino/x White sexual minority peers (Lamb et al., 2019; Martinez et al., 2017). The increased risk for substance misuse is particularly salient among Latino SMM, given this population's higher risk for HIV infection (Martinez et al., 2017).

Latino/x individuals in the United States report higher alcohol consumption, more binge drinking, and greater prevalence of alcohol use disorders in comparison to their non-Latino/x White counterparts (Lui & Zamboanga, 2019). Indeed, 25% of Latino/x individuals are diagnosed with or at risk for an alcohol use disorder in their lifetime (Substance Abuse and Mental Health Services Administration, 2017). These disparities can be largely explained by the negative ramifications of ethnoracial discrimination. One qualitative study suggests that Latino/x individuals who misuse alcohol do so to “distance themselves” from negative daily events, such as experiences of ethnoracial discrimination (Lee et al., 2021). Findings from a large nationally representative study exploring the association between discrimination and alcohol use disorder found that, among persons of color, experiences of ethnoracial discrimination increased the risk of meeting criteria for alcohol use disorder by 50% (Glass et al., 2020). In addition, findings from the National Latino and Asian American Survey suggest that among Latino/x individuals, heavy drinking and the frequency of drinking are linked to discrimination (Savage & Mezuk, 2014). Despite some evidence that links alcohol use and Latino/x identity, there is a dearth of such studies among Latino SMM, specifically. The limited available evidence suggests that Latino SMM may be at higher risk of alcohol misuse in comparison to their White SMM peers (Zollweg et al., 2023).

Similar to those regarding alcohol, associations between drug use and discrimination among Latino SMM are also comparatively understudied. One systematic review among Latino people in general indicated that perceived discrimination was associated with substance use, including illicit drug use (Andrade et al., 2021). Among Latino SMM, specifically, research has implicated drug use with perceived discrimination (Paul et al., 2014) and condomless anal sex, which places these men at greater risk for HIV infection (Wilson et al., 2009).

Sexual and gender minorities of color are at increased risk for negative mental health outcomes as a result of experiencing persistent discrimination based on their sexual/gender identity as well as their ethnoracial identity (Molina et al., 2014; Whitehead et al., 2016). Members of the LGBTQ+ community generally are at higher risk for negative mental health outcomes, including substance misuse, because of discrimination (Garcia-Perez, 2020; Green et al., 2020; Mereish et al., 2017; Pulice-Farrow et al., 2023). Evidence from multiple previous studies supports the hypothesis that sexual orientation–related discrimination and minority stress are associated with increased drinking and alcohol use disorder among SMM specifically (Bogart et al., 2021; Plöderl & Tremblay, 2015; Wray et al., 2016). Within the larger LGBTQ+ population, evidence indicates that Latino/x individuals reported higher levels of anticipated and internalized stigma compared with other racial groups (Shangani et al., 2020). These results highlight the importance of considering the multiple forms of discrimination individuals may be facing, in terms of stigma and discrimination from two systems of oppression, such as Latino SMM.

The minority stress model contends that interpersonal and structural homophobia explains health disparities among sexual minorities (Brooks, 1981; Meyer, 2003). The concept of intersectionality (Bowleg, 2012; Crenshaw, 1989; Hill Collins, 1990) purports that the negative ramifications of minority stress may be exacerbated among those with multiple marginalized identities, such as Latino SMM (Hsieh & Ruther, 2016). Latino SMM often live in complicated systems of oppression that are interlinked and interconnected with their various identities such as racism, homophobia, and linguistic hierarchy (Bogart et al., 2021; Garcia-Perez, 2020; Velez et al., 2019). Not surprisingly, experiencing multiple forms of discrimination, such as racism and homophobia, is correlated with poorer mental health outcomes, including substance use (Hayes et al., 2011; Puckett et al., 2015). Despite the ascendance of intersectionality in research, studies that focus on the intersection of racism and homonegativity when examining health disparities among Latino SMM are limited and generally only incorporate information from one type of discrimination (Layland et al., 2022).

The psychological mediation framework (Hatzenbuehler, 2009) has extended our understanding of minority stress and emphasized the potential for discrimination to disrupt interpersonal relationships, which in turn can have an impact on mental health. LGBTQ+ individuals who are in romantic or sexual relationships live in a social context that stigmatizes both their individual identity and their relationship status (Frost & Meyer, 2009; Peplau & Fingerhut, 2007). Indeed, in recent research, Starks and colleagues (2023) assert that supportive state-level LGBTQ policies are predictive of queer men being partnered, in comparison to states that have antiLGBTQ policies, and that partnered men are more likely to have better mental health. In another study, Robles and colleagues (2024) found that social support and its influence on mental health were conditional on the state policy environment. In terms of intercultural queer couples, stressors and stigma are more complicated and amplified as these couples must navigate intersecting systems of oppression that target their gender/sexual identities, racial/ethnic identities, and their relationship(s). This premise highlights the unique vulnerabilities that might be experienced by Latino SMM in relationships and reinforces that social context can influence health.

The experience and impact of discrimination may be different for SMM in relationships compared with those who are single. Stress experienced by one partner will likely have a negative impact on the other partner (Otis et al., 2006). Although limited, existing research among SMM has demonstrated that alcohol consumption is associated with poorer relationship quality (Starks et al., 2019). Subsequent research has shown that dyadic adjustment (i.e., high relationship quality) can buffer the negative association between relationship discrimination and depression (Robles et al., 2022; Starks et al., 2017).

Previous studies examining the potential for relationship quality to buffer against the deleterious effects of discrimination have been limited to mental health outcomes (e.g., Robles et al., 2022) and in some cases focused on heterosexual samples (e.g., McNeil Smith et al., 2020). To the best of our knowledge, no studies have examined the possibility that relationship quality may moderate associations between discrimination and substance use in ways that attend to the multiple forms of discrimination experienced by Latino SMM. The purpose of the current study was to test hypothesized models predicting alcohol consumption and illicit drug use from ethnoracial and sexual identity–based discrimination and relationship quality. Specifically, we hypothesize that relationship quality will moderate associations between ethnoracial and sexual identity–based discrimination and alcohol as well as illicit drug use. We anticipated that the effects of discrimination on these outcomes would be significant and positive at low levels of relationship quality and attenuate as relationship quality increases.

Method

Participants and procedure

Latino SMM were recruited via a sequential index-partner design using a geo-targeted sexual/dating networking mobile application and social media outlets for participation in the online survey (for more general information on sequential index-partner designs, see Robles et al., 2019). Index partners consented to taking the survey and were screened for eligibility. Index partners had to be 18 years or older; be in a relationship with another male-identifying person; reside in the 50 U.S. states, the District of Columbia, or Puerto Rico; identify as cismale; identify as Latino/x; and be able to read in Spanish or English.

After completing the survey, participants were compensated $10 via an Amazon gift card and were provided with a recruitment email template with a unique survey link and an invitation for their main partner to complete the study. The inclusion criteria for the recruited partner included residing in the 50 states, the District of Columbia, or Puerto Rico; identifying as male; being 18 years of age or older; and being able to read in Spanish or English. All index partners were required to be Latino/x, but recruited partners could be of any ethnoracial identity. Both index and recruited partners were compensated for their time. Study procedures were approved by the University Integrated Institutional Review Board at the City University of New York.

Measures

Demographics. In the current study, we used demographic information that included participants'ethnicity, sexual orientation, level of education, HIV serostatus, pre-exposure prophylaxis (PrEP) usage, level of outness, relationship length, and age. Participants were asked if they identified as Latino/x. Participants were also asked if they identified as heterosexual, gay, bisexual, or queer and could specify another sexual orientation. In the current study, we created a dichotomous variable to indicate if participants identified as gay or any other sexual orientation. Participants were asked to indicate their highest level of education. We then created a dichotomous variable to indicate if participants completed a 4-year college degree or higher or had lower educational attainment (i.e., high school, currently enrolled in college, etc.).

Participants were asked about their self-reported HIV serostatus. Participants were also asked about PrEP usage. We created a binary variable to indicate no PrEP use or current PrEP use. Participants were asked how “out” they were regarding their sexual orientation. Participants responded on a 4-point Likert-type scale that ranged from not at all to completely. Higher scores indicate a higher degree of outness. Participants were asked to self-report the length of their relationship. In the current study, we averaged both partners' self-reported relationship length to produce the only couple-level variable in the analyses. Finally, we calculated participants' age based on their self-reported birthday.

Experiences of discrimination. To measure discrimination, we used two versions of the Everyday Discrimination Scale (EDS; Forman et al., 1997). The EDS is a 9-item measure of experiences of everyday discrimination and was modified to include a lead-in to assess the attributable cause of the discrimination. In the current study, the lead to the scale was modified to assess if participants experienced discrimination based on their sexual orientation and a repeat of the same questions but with a focus on race. Examples of items include, “You are treated with less courtesy than other people” and “You are treated with less respect than other people.” The response format included a 6-point scale with responses ranging from never to almost every day. Items were summed together, and higher scores indicate a higher frequency of discrimination. The internal consistency of the sexual orientation and race EDS scores was strong, with Cronbach's alpha levels of .94 and .96, respectively. Because of skewness of the data and sample size constraints, we created a binary indicator of discrimination after averaging the two discrimination scales together. Zero was coded as values below the median (e.g., low scores) and one was coded as above the median (e.g., high scores).

Relationship functioning. Relationship functioning was measured using the Revised Dyadic Adjustment Scale (DAS; Busby et al., 1995). The DAS is a 14-item measure of a couple's adjustment to relationship conflict and stressors that asks how often members of the couple disagree on everyday stressors. Examples items include, “How often do you discuss, or have you considered, divorce, separation, or terminating your relationship?” or “How often do you and your partner disagree on religious matters?” Respondents answer each question on a 5-point scale from always disagree/occurs all the time to always agree/never occurs. Items were summed, and higher scores signify greater dyadic adjustment. Internal consistency of the scale was considered good (α = .84). We mean-centered the final scale to facilitate the interpretation of interaction effects.

Problematic drinking. Problematic drinking was measured using the Alcohol Use Disorders Identification Test (AUDIT; Bohn et al., 1995). The AUDIT has three dimensions: quantity of consumption, problems related to drinking, and injury from drinking. Drinking frequency is assessed on a 5-point scale from never to 4 or more times per week. Alcohol consumption on a typical day is measured on a 5-point scale from 1–2 drinks to 10 or more drinks. Problems related to alcohol consumption are measured with 6 items on a 5-point scale from never to daily or almost daily. Physical injury to self or others as a result of drinking is measured on a 3-point scale from no to yes, during the last 6 months. Concern from others is measured on a 3-point scale from no to yes, during the last 6 months. All 10 items are summed together, with higher scores indicating greater problematic drinking. Internal consistency was strong (α = .86). Because of the skewness of data in this sample, we created a binary variable to indicate nonproblematic drinking (coded as zero, in which individuals score below 6 on the AUDIT scale) versus problematic drinking (coded as one, based on the cutoff AUDIT score of 6 or higher). This cutoff was based on recommendations of prior research (e.g., Babor et al., 2001).

Drug use. Participants were initially asked if they had used drugs (from a list of possible drugs) within the past 30 days. Examples of the types of drug use assessed included cannabis, opioids, stimulants, club drugs, and prescription sedatives. Because of the limited number of participants using drugs from all of the available options, we created a binary variable to indicate no drug use (coded as zero) versus any drug use (coded as one).

Statistical analyses. The similarity of partners' responses was evaluated using intraclass correlations for continuous variables and statistics for binary or categorical variables. We calculated Actor–Partner Interdependence Models (APIMs) assuming exchangeable dyads (Kenny et al., 2020) using generalized estimating equations (GEEs) within Stata Version 17 (StataCorp LP, College Station, TX). We estimated APIM GEEs for each outcome (i.e., AUDIT and drug use) independently. APIMs are multi-level (Hong & Kim, 2019), with individual participants nested within a couple. At Level 1, actor effects represent the association between a participant's own predictor score and the outcome; meanwhile, partner effects quantify the association between a partner's predictor score and the outcome. In addition to predictors of primary interest—relationship functioning and discrimination—the model included actor and partner effects of demographic covariates (e.g., ethnicity, sexual orientation, level of education, HIV serostatus, level of outness, and age) at Level 1. Level 2 of the APIMs contained couple-level variables in which both members of the couple have identical or shared responses (i.e., relationship length). For the current analyses, Level 2 was limited to relationship length.

We first ran a main effects model, followed by a model that included two-way interactions between actor and partner effects of relationship functioning and discrimination. Subsequently, nonsignificant interactions were excluded in a reverse-hierarchical procedure to produce final parsimonious models. Model fit was evaluated using quasi-likelihood under the independence model criterion (QIC; Cui, 2007; Pan, 2001), an extension of Akaike's information criterion for GEE models.

Results

A total of 625 index partners completed the online survey. Ninety-five index partners successfully recruited their main partners. Only complete dyads were included in the analysis of the current study. Our sample consisted of 190 individuals (95 dyads). All our observations had complete data on study variables; therefore, there are no considerations for missing data.

The majority of participants identified as Latino (80%) by design, as at least one of the partners within the couple was required to be of Hispanic or Latinx descent. Table 1 includes descriptive statistics for all study variables. The majority of participants had a college education or higher (55.79%), identified as gay (88.42%), and were HIV negative (85.79%). The average age of participants was 31.63 years (SD = 7.80). The average relationship length was about 4 years (48.8 months; SD = 53.38).

Table 1.

Sample description (N = 190)

Individual characteristics	n	%	Similaritya k
Ethnoracial identityb
Latino	152	80.00
Non-Latino	38	20.00
Education
<4-year degree	84	44.21	.36***
4-year degree or higher	106	55.79
Sexual orientation
Gay	168	88.42	.19
Not gay	22	11.58
HIV status
Positive	27	14.21	.61***
Negative	163	85.79
PrEP use
No PrEP use	158	83.16	.40***
Current PrEP use	32	16.84
Discrimination
Low discrimination	95	50.00	.31***
High discrimination	95	50.00
Problematic drinking (AUDIT)
Scores below 6	149	78.42	.29***
Scores 6 and above	41	21.58
Any drug use
Absence of drug use	96	50.53	.41***
Presence of any drug use	94	49.47
	M	SD	ICC
Age	31.63	7.8	.59***
Relationship length, in monthsc	48.8	53.38	–
Outness	3.38	0.89	.41***
Relationship functioning (DAS)	33.02	7.20	.22**

Notes: PrEP = pre-exposure prophylaxis; AUDIT = Alcohol Use Disorders Identification Test; ICC = intraclass correlation; DAS = Dyadic Adjustment Scale.

^a Similarity between Actor and Partner scores was assessed using intraclass correlations for continuous variables and statistics for binary or categorical variables;

^b all index participants were Latino by design; therefore, similarity cannot be calculated;

^c relationship length was calculated from the average of the Index Participant and Recruited Partner self-report responses.

^**p < .01;

^***p < .001.

Problematic drinking results

Our results indicated that the main effects model did not account for a significant amount of variance in problematic drinking, χ²(19) = 28.84, p = .069, which is contrary to our study hypothesis. Subsequently, including all interaction terms decreased model fit (delta QIC = -3.817), and no interaction terms were significant. Partner high relationship functioning was associated with lower odds of drinking (OR = 0.44, 95% CI [0.20, 1.00], p = .049) compared to those with low relationship functioning. Table 2 contains comprehensive parameters for the final main effects model.

Table 2.

Summary of multivariable results for problematic drinking and any drug use

Variable	Problematic drinking OR [95% CI]	Any drug use OR [95% CI]
Couple level
Relationship length	0.99 [0.98, 1.01]	1.00 [0.99, 1.01]
Individual level
Latino (ref.: non-Latino)
Actor	0.47 [0.17, 1.27]	0.87 [0.34, 2.20]
Partner	2.22 [0.67, 7.41]	0.43 [0.16, 1.15]
Sexual orientation (ref.: gay)
Actor	0.79 [0.25, 2.51]	0.93 [0.30, 2.88]
Partner	0.51 [0.17, 1.53]	0.33* [0.11, 0.95]
Education (ref.: less than college)
Actor	1.83 [0.91, 3.65]	1.20 [0.55, 2.62]
Partner	0.81 [0.39, 1.70]	0.64 [0.30, 1.35]
HIV status (ref.: negative)
Actor	2.03 [0.55, 7.46]	1.50 [0.32, 7.13]
Partner	0.42 [0.18, 1.53]	1.36 [0.32, 5.74]
PreP status (ref.: negative)
Actor	1.31 [0.40, 4.22]	1.01 [0.35, 2.92]
Partner	0.81 [0.27, 2.42]	4.36** [1.48, 12.83]
Age
Actor	0.99 [0.93, 1.06]	1.03 [0.98, 1.08]
Partner	0.99 [0.93, 1.06]	1.00 [0.95, 1.06]
Outness
Actor	1.02 [0.67, 1.55]	1.20 [0.81, 1.78]
Partner	1.82* [1.10, 3.01]	1.38 [0.91, 2.10]
Discrimination (D) (ref.: low)
Actor	2.20 [0.98, 4.93]	11.54*** [3.36, 39.70]
Partner	1.09 [0.51, 2.31]	1.84 [0.55, 6.15]
DAS
Actor	0.71 [0.30, 1.64]	0.56 [0.27, 1.19]
Partner	0.44* [0.20, 1.00]	0.84 [0.26, 2.74]
Interaction effects
Actor D × Partner DAS		0.20** [0.07, 0.59]
Partner D × Partner DAS		3.06 [0.94, 10.01]
Actor D × Partner D		0.16* [0.03, 0.76]

Notes: OR = odds ratio; CI = confidence interval; ref. = referent group; PrEP = pre-exposure prophylaxis; D = discrimination; DAS = Dyadic Adjustment Scale.

^*p < .05;

^**p < .01;

^***p < .001.

Drug use results

Our results indicated that the main effects model did not account for a significant amount of variance in drug use, χ²(19) = 27.75, p = .088. Subsequently, the inclusion of all interaction terms improved model fit (delta QIC = 5.745), and the joint test of significance of the interaction terms was statistically significant, χ²(6) = 23.92, p < .001. This model did explain a statistically significant amount of variation in drug use, χ²(25) = 50.66, p = .002. Based on the improvement in QIC, model testing advanced to the removal of nonsignificant interaction terms. In this model, one interaction term (Partner D × Partner DAS) was no longer significant. We removed this interaction, but this led to a decrease in model fit (delta QIC = 2.003). Because of this decrease in model fit, we reverted to the previous model, which yielded a final parsimonious model that included three interaction terms, two of which were statistically significant. Table 2 contains comprehensive model parameters for the final parsimonious model.

Consistent with hypotheses, the interaction between actor discrimination and partner DAS was significant (OR = 0.20, 95% CI [0.07, 0.59], p = .003). Likewise consistent with hypotheses, the interaction between Actor discrimination and Partner discrimination was statistically significant (OR = 0.16, 95% CI [0.03, 0.76], p = .021). Figures 1 and 2 depict these interactions. At lower levels of self-reported relationship quality (partner DAS scores below the median), the effect of actor discrimination was associated with higher odds of drug use (OR = 11.54, 95% CI [3.36, 39.70], p < .001). In contrast, at higher levels of self-reported relationship quality (partner DAS above the median), the effect of actor discrimination was not statistically significant (OR = 2.31, 95% CI [0.73, 7.32], p = .155; Figure 1). Similarly, among men whose partners reported low discrimination, the effect of actor discrimination was associated with higher odds of drug use (OR = 11.54, 95% CI [3.36, 39.70], p < .001); meanwhile, among those with partners who reported high levels of discrimination, the effect of actor discrimination was not statistically significant (OR = 1.83, 95% CI [0.53, 6.35], p = .343; Figure 2).

Figure 1.

Marginal odds ratio of drug use in the interaction between actor discrimination and partner relationship functioning. DAS = Dyadic Adjustment Scale.

Figure 2.

Marginal odds ratio of drug use in the interaction between actor discrimination and partner discrimination

Discussion

These findings comport with a growing body of literature suggesting that relationship quality contextualizes the associations between experiences of discrimination and drug use among partnered SMM. Generally, we find evidence of cross-over effects, such that the effect of actor discrimination on drug use is buffered by higher levels of partner relationship functioning and partner discrimination. In terms of problematic drinking, we do not find evidence of the buffering role of higher levels of relationship functioning. Our results for drug use, but not problematic drinking, are similar to those in other literature that indicates that couple-based discrimination and depression are buffered by relationship functioning (e.g., Robles et al., 2022). In our study, we find that higher levels of relationship functioning and having a partner who experiences discrimination are associated with lower levels of drug use.

We interpret these findings in the context of a potential psychological mediation framework, as results indicate that the deleterious effect of discrimination may be buffered by higher levels of relationship functioning. We speculate that higher relationship functioning may mitigate the need for predominantly Latino SMM to escape from their experiences of discrimination by using drugs, whereas previous literature indicates (e.g., De Santis et al., 2014) that participants with lower relationship functioning may opt instead for ineffective coping mechanisms, such as selfmedicating. We speculate that having a partner who also experiences discrimination may serve as a source of coping because they can more readily empathize with their partner's experiences. Although alcohol use is generally seen as ubiquitous among Latino SMM (Martinez et al., 2016), evidence of interactions was observed only for drug use and not problematic drinking, providing evidence that the profound ramifications of experiencing discrimination may require more powerful substances such as stimulants to escape, which may partially explain these differential effects in terms of substance used.

Relatedly, the absence of relationship functioning moderating the effects of discrimination and problematic drinking should not be misinterpreted to indicate that relationships do not affect drinking. We theorize that problematic drinking may be an activity that couples engage in jointly, mitigating the buffering role of relationship functioning in terms of problematic drinking. In comparison, drug use may be an activity that couples do not jointly engage in, with only one partner engaging in the activity. In this sense, the protective effects of relationship functioning may be observed because the problematic behavior is only performed by one of the partners. Hence, future research is needed to disentangle the substance use patterns of couples to examine the role of relationship functioning in substance use more thoroughly.

Implications for policy and practice

We interpret our findings to facilitate advancements in policy and practice, both in the short term and in the longer term. Historically, the focus of relationship functioning has been on general communication, sexual communication, and accessing social support from one's partner. However, less attention has been paid to how to communicate with one's partner and leverage support related to discrimination. This lack of attention means that couples may not be fully equipped to have these conversations in a relationship-functioning framework. As a result, there is a need to incorporate strategies and approaches that can help couples address discrimination. By doing so, couples can build stronger, more resilient relationships that can withstand the challenges, such as drug use, that arise from discrimination.

Although shorter-term goals may fill an immediate need, it assumes that SMM should do something extra to buffer against oppressive forces, when the residual effects of oppressive forces should actually be the primary target of intervention. More comprehensive and longer-term strategies should therefore include measures to dismantle structural oppressive forces as a matter of health equity. Policy changes that promote equity among SMM, including addressing structural racism, are needed to eliminate the burden on individuals, regardless of their relationship status.

Limitations

We interpret our findings within the context of several limitations, which may temper study results. By design, this study only included SMM currently in same-gender relationships. Therefore, our results are not immediately generalizable to SMM not in a relationship or who are not cisgender. Future studies should consider how being in a romantic relationship further influences risk and protection for substance and alcohol misuse. In addition, future studies should also leverage samples from SMM that include transmen.

In a similar vein, SMM couples who enrolled in this study could have been influenced by selection as a result of the use of a convenience sample that is not necessarily representative of the population at large. Relatedly, we did not oversample subpopulations within the Latino SMM community, which largely identified as White Latino. Future studies using nationally representative data, including oversampling of subpopulations (i.e. Black/Afro-Latinos, multiracial-Latinos, indigenous-Latinos, and mono/multiracial Asian-Latinos) would be beneficial to ensure that findings are generalizable to the population at large. The Latino/x community is a heterogeneous group, with distinct use behaviors and experiences of discrimination across a range of subgroups such as by national origin or immigration status. In addition, cultural factors, such as beliefs, values, and identity, may further shape substance use behaviors. Future research should therefore leverage more representative data to disentangle how a multitude of factors that influence Latino/x identity shape substance use behaviors and experiences of discrimination.

Our study was a cross-sectional study with a small sample size. Future studies should leverage longitudinal study designs to enhance the robustness of study conclusions by being able to draw causal conclusions. Moreover, the psychological mediation framework could not be robustly tested in the current analysis given the cross-sectional nature of our study; therefore, longitudinal studies are needed to test this framework more rigorously. The relatively small size of our sample may limit the generalizability and/or the robustness of the study findings. In the current study, we could not explore true intersectional discrimination (e.g., racism and homophobia) and were limited in the use of psychometrically sound substance use measures because of data constraints. Future research is therefore needed to replicate our study findings using larger samples to overcome this limitation.

Conclusions

Although this study has its limitations, it addresses an important knowledge gap about how experiences of discrimination and dyadic adjustment confer risk/protection to substance misuse, but not problematic drinking, in a primarily Latino sample of SMM. Our results support that SMM are particularly at risk for substance misuse when experiencing discrimination, especially at lower levels of relationship function or when their partner does not also experience discrimination. This point is particularly important within the context of this sample, where the majority of respondents carry multiple marginalized identities. We lay out future directions for practice and policy to meet the immediate need, but more important is the need to address discrimination at the macro level, which would eliminate, or at least temper, the need for populations to develop resiliency.