Discrimination, relationship quality, and the mental health of sexual minority men in relationships: A test of the psychological mediation hypothesis

Abstract

Mental health disparities among sexual minority men (SMM) are commonly studied in the context of minority stress theory. The psychological mediation hypothesis posits that stigma is adversely associated with mental health in part through disruptions in interpersonal relationships. This study tested models hypothesizing internalized, enacted, and relationship-focused stigma would be associated with mental health through diminished relationship quality in a sample of SMM in relationships. A sample of 192 cisgender SMM (aged 18 to 35 and in a relationship with an adult cisgender man) responded to advertisements on social and sexual networking applications and completed survey assessments as part of the baseline for a larger randomized controlled trial. Results of path model analysis indicated that internalized, enacted, and relationship stigma were all in some way associated with one or more indicators of relationship quality. In turn, one indicator of relationship quality – barriers to leaving the relationship – was associated with depression and anxiety. Both internalized heterosexism and relationship marginalization were positively associated with depression and anxiety through indirect pathways involving the perception of barriers to leaving the relationship. Results generally aligned with minority stress theory and the psychological mediation hypothesis specifically. Existing interventions that address the impact of coping with stigma for sexual and gender minority people are presently focused on individuals. The development of relevant interventions for those in relationships would help to address extant mental health disparities in this population.

1. Introduction

Compared to heterosexual cisgender individuals, sexual minority people experience significantly higher rates of anxiety and depression (Cai et al., 2024). They simultaneously face barriers to behavioral and mental health treatment access (Williams et al., 2022). These disparities extend to sexual minority men (SMM) specifically, where elevated rates of depression (Lee et al., 2017) and anxiety (Lassiter et al., 2023) relative to heterosexual men are evident. Studying these mental health disparities among SMM at elevated risk of HIV infection is of particular public health significance. Mental health (most consistently depression) features prominently in research on psychosocial syndemics and the HIV epidemic (Liu et al., 2023; Stall et al., 2008; Starks et al., 2014).

Mental health disparities among cisgender sexual minority people are well-documented. The literature documenting disparities in depression and anxiety is particularly noteworthy. Mental health disparities among sexual minority people generally, and SMM specifically, are often explained using the minority stress framework (Brooks, 1981; Meyer, 1995, 2003). This framework suggests that sexual minority people experience unique stressors due to their minoritized identities, including distal (i.e., public policies or stigma enacted during interpersonal exchanges) as well as proximal (i.e., the internalization of negative societal attitudes about sexual minority people, known formerly as internalized homophobia and more commonly now as internalized heterosexism) processes (Szymanski & Mikorski, 2016). Although major events are also considered, the framework emphasizes the cumulative impact of chronic, everyday experiences of enacted stigma (i.e., discrimination, prejudice, and social disadvantage) on the mental and physical well-being of sexual minority people.

Hatzenbuehler (2009) subsequently put forward the psychological mediation hypothesis. It suggests that experiences of stigma are associated with adverse mental health outcomes because they lead to disruptions in interpersonal relationships as well as intra-individual emotional and cognitive processes. Although scholarship on the interpersonal consequences of stigma within the psychological mediation hypothesis initially said little about primary partner relationships, there are at least two ways in which these become relevant. Hatzenbuehler emphasized that stigma directed at an individual’s identity may lead to social isolation and diminished social support. Relatively recent research on relationship marginalization – the rejection of one’s relationship or primary partner by peers, family, and society generally – suggests that it is also significantly associated with mental health problems (Robles et al., 2022; Rosenthal & Starks, 2015). Notably, relationship marginalization typically refers to a general process of social rejection (Lehmiller, 2012). In contrast, measures of couple-level minority stress (Neilands et al., 2020) have been developed to assess group-specific stigma selectively directed at sexual and gender minority people in relationships.

The social isolation and pressure to conform resulting from experiences of stigma (including internalized and enacted stigma as well as relationship marginalization) may also diminish relationship quality. Multiple (quantitative and qualitative) studies on relationship marginalization demonstrate that such stigma diminishes relationship quality among same-gender couples (e.g., Frost et al., 2017; Robles et al., 2022; Rosenthal & Starks, 2015). These stressors can manifest in various ways, such as lower relationship investment and expending emotional resources that would otherwise be used to improve the quality of the relationship (Oren, 2024). Unfortunately, none of these studies have explicitly tested indirect paths in a manner aligned with the psychological mediation hypothesis, wherein associations between relationship marginalization and mental health might be explained in part through indirect associations involving relationship quality.

The relatively few studies of relationship quality and mental health among SMM have relied on global or overall assessments of relationship quality. This overlooks the fact that modern theories, including interdependence theory (Kelley & Thibaut, 1978) and the investment model (Rusbult et al., 1998), which are commonly used to explain health-related social control (Lewis et al., 2004; Lewis et al., 2006) and undergird couples’ interventions developed for SMM, suggest that relationship quality is multifaceted. In their theoretical review of the literature on relationship commitment (i.e., the intention to persist in a relationship), Stanley, Rhoades, and Whitton (2010) bifurcate components of relationship quality into two domains or determinants of commitment. The dedication domain includes aspects of relationship quality that capture desire for a relationship (positive features such as satisfaction). The constraint domain comprises those aspects of relationship quality that might compel an individual to feel as though they have to sustain a relationship (investment, the paucity of alternatives, or consequences associated with relationship termination).

The purpose of the current study was to test a hypothesized model derived from minority stress theory and the psychological mediation hypothesis to predict adverse mental health outcomes among SMM in relationships who report a recent history of sexual behaviors that may elevate their risk of sexual HIV or other STI transmission. The hypothesized model (see Figure 1) posited that stigma (internalized heterosexism, enacted discrimination, and relationship marginalization) would be positively associated with adverse outcomes (depression and anxiety) directly and through indirect pathways involving relationship quality. Each of these possible indirect pathways comprises two components. The first component is a direct, negative association between stigma and indicators of good relationship quality (or conversely, a positive association with indicators of poor relationship quality). The second component is a direct and negative association between indicators of good relationship quality and adverse mental health outcomes (or conversely, a positive association with indicators of poor relationship quality).

Figure 1.

Hypothesized model

NOTE: Directional hypotheses are indicated by “−” and “+” signs.

2. Methods

2.1. Participants

Data were gathered as part of the baseline assessment for project PARTNER (Starks, Robles, Pawson et al., 2019), a randomized controlled trial whose primary purpose was to evaluate the efficacy of a 4-session motivational interviewing intervention to reduce drug use and associated HIV transmission risk behaviors among SMM in relationships. Eligible participants met the following criteria: 1) cisgender man; 2) 18 to 29 years of age (with a partner aged 18 or older), or 30 to 34 (with a partner aged 18 to 29); 3) have a main partner (for at least one month) who is a cisgender man and 18 years or older; 4) HIV negative (confirmed by HIV rapid testing); 5) reported drug use in the past 30 days; 6) endorsed condomless anal sex with a casual partner or a non-monogamous primary partner (regardless of HIV status) or a primary partner who was living with HIV (regardless of sexual agreement) in the past 30 days; 7) able to read and speak English. Until March 2020, participation was restricted to those living in the New York City metropolitan area. In January 2021, the study design was changed (see Procedures) to facilitate remote participation and study enrollment was expanded to the US nationwide in September 2022.

2.2. Procedures

Participants were recruited through advertisements on social media sites (e.g., Facebook, Instagram) and smartphone-based dating applications to complete an online screener that assessed preliminary eligibility for various federally funded research studies and gathered contact information. In addition, study staff visited bars, nightclubs, and other social events in person to conduct SMM in the metropolitan area and screened participants on an iPad using the same preliminary screener used in online recruitment.

From January 2018 through June 2019, study staff contacted those who screened preliminarily eligible via telephone. On this call, they administered a study-specific screener. In response to participant feedback about communication preferences, this was converted to an online survey in July 2019. Study staff sent participants an email or text message containing a link to the study-specific screener. Those who were confirmed eligible were scheduled for a baseline appointment that occurred at a research center located in New York City. Study recruitment was suspended on March 15, 2020, due to the COVID-19 pandemic. Recruitment resumed on geospatial networking applications on January 2021. The transition to remote assessment removed geographical restrictions on participation. In September 2022, eligibility criteria were changed to include participants living in any US state. The purpose of this change was to maximize enrollment prior to closure.

Data used in this study were obtained from baseline survey responses. This was one data-collection point in a larger set of study procedures. From January 2018 to March 2020, baseline appointments were conducted in person at a research center located in New York City. Prior to the scheduled appointment, participants viewed written consent information and completed a survey administered through Qualtrics. When they arrived for the appointment, the research assistant verbally reviewed consent information and confirmed age eligibility. The participant then completed an in-office Qualtrics survey and a staff-administered timeline follow-back interview detailing recent sexual behavior and substance use. The participant also self-collected samples for gonorrhea and chlamydia (rectal and urethral) testing, for PrEP adherence testing using a dried blood spot (DBS) card, and for HIV testing using an OraQuick rapid test. Additionally, participants provided nail (finger or toenail) samples for drug use testing.

The study design changed substantially in response to the COVID-19 pandemic. When the study re-opened, baseline appointments were conducted remotely. Prior to the scheduled appointment, participants viewed written consent information and completed a survey administered through Qualtrics. Participants also received at-home HIV and STI testing materials shipped by Molecular Testing Labs as well as a nail collection kit mailed separately by study staff. To complete the appointment, participants used a link to join a video call hosted by a HIPAA-compliant Zoom account. During the call, the research assistant verbally reviewed consent information, confirmed age eligibility, and administered a timeline follow-back interview. The research assistant also provided instructions for self-collecting samples for gonorrhea and chlamydia (rectal and urethral) testing as well as for PrEP adherence using a DBS card. Additionally, participants completed a rapid HIV test (OraQuick) with guidance from the research assistant who provided pre- and post-test counseling. At the end of the appointment, the research assistant instructed participants to return biological specimens (rectal swab, urine sample, and DBS card) to the lab for testing using pre-paid shipping materials and instructed participants to collect and return nail samples independently to the research office using a pre-paid envelope.

Regardless of participation date, all participants were compensated a total of $50 for baseline activities, including $20 for survey completion, $20 for the timeline follow-back interview, and $10 for biological specimen collection. Prior to March 15, 2020, compensation was delivered in cash. Thereafter, compensation was delivered as an Amazon gift card via email. All study procedures (and related changes) were reviewed and approved by The City University of New York (CUNY) Integrated Institutional Review Board.

2.3. Measures

2.3.1. Demographics

Participants reported their age, PrEP uptake, sexual identity (heterosexual or straight, gay, bisexual, queer, or another identity; responses were dichotomized into gay versus bisexual or another identity), income (dichotomized as below $40,000 and $40,000 or above annually), and education (less than a four-year degree or above). Participants also indicated their ethnicity (Hispanic versus not Hispanic) and racial identity (White, Black or African American, American Indian or Alaska Native, Asian, Native Hawaiian or Pacific Islander, another identity). Responses were aggregated to distinguish among those who identified as ethnically Hispanic or Latino as well as among non-Hispanic respondents who identified as White, Black or African American, Asian/Hawaiian/Pacific Islander, multiracial or another identity.

Participants also reported their partners’ age, HIV status, and PrEP uptake. Partner race and ethnicity was assessed and aggregated analogous to personal race and ethnicity. Relationship length was assessed on an ordinal scale (<1 month; 1 or 2 months; 3 to 6 months; more than 6 but less than 12 months; 1 to 2 years; more than 2 years). Responses were dichotomized to indicate relationship length of more than 2 years versus 2 years or less. This distinction aligns with a hypothesized transition point in relationship development, moving from a merging of individual identities to a re-emergence of individual identities within a relationship (McWhirter & Mattison, 1984).

Sexual arrangement was assessed using a single item, “Regardless of your sexual agreement, how do you and your partner handle sex outside your relationship?” commonly employed in studies of SMM in relationships (Cain & Starks, 2024). Response options included 1) “neither of us have sex with others,” 2) “only I have sex with others,” 3) “only he has sex with others,” 4) “we only have sex with others together,” 5) “we both have sex with others separately and together,” 6) “we both have sex with others separately, 7) I have sex with others, I don’t know about my partner,” and 8) “I don’t have sex with others, I don’t know about my partner.” A two-category variable was created to describe those who had a monogamous arrangement (responses 1 and 8) and those who had a non-monogamous arrangement (responses 2 through 7).

2.3.2. Enacted discrimination

The Everyday Discrimination Scale (EDS) (Williams et al., 1997) begins with 9 items that assess the frequency of discrimination experiences (e.g., “You are treated with less respect than other people” and “You are called names or insulted”). For each item, participants reported the frequency of their everyday discrimination on a six-point Likert scale from 1 (never) to 6 (almost every day). Responses to these items were summed. Higher values indicated more frequent experiences of discrimination. The total score demonstrated good internal consistency (Cronbach’s α = 0.866).

2.3.3. Internalized heterosexism

Internalized stigma was assessed using the Internalized Heterosexism Scale (Meyer & Dean, 1998). Participants indicated the extent to which they agreed with each of 9 items (e.g., “I often feel it is best to avoid personal or social involvement with other gay/bisexual men” and “I have tried to stop being attracted to men in general”) using a Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Items were summed to create a total score, with higher scores indicating greater internalized heterosexism. This total score demonstrated good reliability (Cronbach’s α = 0.865)

2.3.4. Relationship stigma

Stigma directed at a main or primary relationship was assessed using an adapted form of the Relationship Marginalization Scale (RMS) (Lehmiller & Agnew, 2006) used in previous studies (Robles et al., 2022; Rosenthal & Starks, 2015). The scale comprised 6 items-assessing the extent to which individuals anticipated peers and friends (e.g., “My friends are generally not accepting of my romantic partner”), family members (e.g., “My family is generally not accepting of my romantic partner”), and society in general (e.g., “I believe that most other people (who I do not know) would generally disapprove of my romantic relationship”) disapproved of their relationship. The original scale included 4 items with the assessment of friends’ and family members’ attitudes combined. Participants reported their responses on a Likert-type scale from 1 (strongly disagree) to 5 (strongly agree). Item responses were summed, with higher scores indicating more relationship stigma. This total score demonstrated acceptable reliability (Cronbach’s α = 0.782).

2.3.5. Depressive symptoms

The Center for Epidemiologic Studies - Depression (CESD-10) scale (Andersen et al., 1994) comprises 10 items assessing depressive symptoms experienced in the past week (e.g., “I was bothered by things that usually don’t bother me” and “I felt depressed”). Response options ranged from 0 (Rarely or none of the time – less than 1 day last week) to 3 (Most of the time – 5–7 days). Item responses were summed to create a total score with higher scores indicating more symptoms of depression. This total score demonstrated good internal consistency (Cronbach’s α = 0.841).

2.3.6. Anxiety symptoms

The Generalized Anxiety Disorder-7 (GAD-7) asks participants to report how often they have experienced anxiety symptoms (e.g., “feeling nervous, anxious, or on the edge”) over the past 2-weeks (Spitzer et al., 2006) using a Likert-type scale ranging from 0 (not at all) to 3 (nearly every day). Responses were summed with higher scores indicating more anxiety symptoms. This total score had good internal consistency (Cronbach’s α = 0.897).

2.3.7. Relationship quality

Relationship functioning was assessed using Kurdek’s (1995, 1998) Multiple Determinants of Relationship Commitment Inventory (MDRCI). The scale comprises 6 subscales, each containing 4 items. Of these, 4 subscales assess dimensions of relationship quality typically considered to be strengths: rewards (e.g., “One advantage to my relationship is having someone to count on”); comparison to an ideal standard (e.g., “My current relationship comes close to matching what I would consider to be my ideal relationship”); investments (e.g., “I’ve put a lot of energy and effort into my relationship”) and barriers to leaving the relationship (e.g., “I would find it difficult to leave my partner because I would feel obligated”). The remaining subscales assess subdomains typically associated with poor relationship quality: costs (e.g., “I give up a lot to be in my relationship”); attractiveness of alternatives (e.g., “As an alternative to my relationship, I would like to date someone else”). Response options ranged from 1 (strongly disagree) to 5 (strongly agree).

Internal consistency for 5 MDRCI subscales ranged from adequate to good (Cronbach’s α = .756, .824, .849, .758, and .723 for rewards, costs, comparison to ideal, attractiveness of alternatives, and barriers to leaving respectively). In contrast, the investments subscale of the MDRCI had poor reliability (Cronbach’s α = .634) due to one specific item “A part of me is tied up in my relationship.” When the item “A part of me is tied up in my relationship” was removed, subscale reliability improved appreciably (Cronbach’s α = .797). This same phenomenon has been observed in a previous study of SMM in relationships (Starks, Robles, Bosco et al., 2019). Consistent with that, we utilized this 3-item version of the investments subscale based upon superior internal consistency.

2.3.8. Problematic substance use

The 10-item Drug Abuse Screening Test (DAST-10) (Bohn et al., 1991) assessed symptoms of substance use. Participants who indicated the use of drugs for other than medical purposes were asked to indicate whether they experienced each of 10 problems or symptoms as a result of drug use. Responses were summed to produce a total score. Those participants who did not indicate the use of any assessed substances were assigned a value of 0. This total score had adequate internal consistency for research purposes (Cronbach’s α = .622). Subsequently, the total score was dichotomized at the cutoff for moderate impairment (3 or more symptoms versus 2 symptoms or fewer).

3.0. Analytic plan

Data were inspected to ensure conformity with analytic assumptions. Descriptive statistics were calculated in SPSS (v. 29). Pearson’s product-moment (r) and point-biserial correlations quantified bivariate associations among constructs of primary interest (mental health outcomes, experiences of stigma, and indicators of relationship quality) and demographic covariates.

Following preliminary descriptive analyses, a path model was calculated in Mplus version 8.1. Depression and anxiety were both regressed on indicators of relationship quality (MDRCI subscale scores: rewards, costs, approximation to an ideal, investment, valuation of alternatives, and barriers to leaving the relationship) as well as stigma scores (internalized heterosexism, enacted discrimination, and relationship marginalization). Relationship quality subscale scores were in turn regressed on stigma scores. The prediction of all endogenous variables controlled for relationship length; participants’ own problematic drug use scores (above versus below clinical DAST-10 clinical cutoff for moderate impairment), age, race and ethnicity, and sexual identity; as well as the age, race and ethnicity, and sexual identity of participants’ relationship partners. The residual correlation between depression and anxiety was freely estimated, recognizing the comorbid nature of these two outcomes.

Model calculation employed full information maximum likelihood estimation. Model significance was evaluated by comparison of the specified model to a null model (wherein all regression coefficients were constrained to be zero). Where the pattern of direct associations suggested a plausible indirect pathway, these were tested using the Model Constraint command in Mplus (Muthén & Muthén, 2018). Contingent upon model significance, associations between specific predictions and endogenous variables were evaluated by inspection of regression coefficients using two-tailed p-values and α = .05.

4.0. Results

The screening survey was completed 129,183 times. Of these, 128,286 (99.3%) were determined to be ineligible. Most were ineligible due to location before September 2022 (115,570, 90.1%) or age (8,984, 7%). Separately, 276 (0.2%) ineligible participants did not identify as cisgender man or were living with HIV; 1,232 (1.0%) had been in their relationship for less than 1 month; and 230 (0.2%) reported the occurrence of physical or sexual intimate partner violence and indicated they did not feel safe in their current relationship. Another 1,567 (1.2%) were ineligible because they did not meet substance use or sexual behavior criteria. Finally, 156 (0.1%) otherwise eligible respondents declined to participate. Of the 897 participants screened eligible, 225 (25.1%) completed the baseline appointment. Of these, 27 (12%) withdrew after baseline; 2 (0.9%) participants were identified as comprising a dyad and withdrawn from analyses; and 4 (1.8%) were missing survey data. This resulted in the final analytic sample of 192 participants.

Table 1 displays descriptive data for the analytic sample. On average, the sample was 26.86 years old (SD = 2.78). More than a quarter identified as Hispanic. Approximately half identified as Non-Hispanic and White. The majority identified as gay, completed a 4-year college degree, and earned more than $40,000 annually. More than half had a current PrEP prescription.

Table 1.

Sample characteristics (n = 192)

	Personal Characteristics	Partner Characteristics
	n (%)	n (%)
Race and ethnicity
White	99 (51.0)	105 (54.1)
Black or African American	20 (5.2)	20 (10.3)
Latino or Hispanic	53 (27.3)	46 (23.7)
Asian, Pacific Islander	15 (7.7)	13 (6.7)
Multiracial or another identity	15 (7.7)	8 (4.1)
Sexual Identity
Gay	157 (81.8)	169 (88.0)
Bisexual or another identity	35 (18.2)	23 (12.0)
Region of residence
Northeast	178 (91.8)
Elsewhere in the US	14 (7.2)
Education
Less than a 4 year degree	43 (22.2)	36 (18.8)
4 year degree or higher	149 (76.8)	156 (81.3)
Annual Income
Less than $40,000	94 (48.5)
$40,000 or more	98 (50.5)
HIV status and PrEP uptake
HIV negative PrEP Rx	104 (54.2)	73 (38.0)
HIV negative no PrEP Rx	88 (45.8)	109 (56.8)
Living with HIV	0 (0.0)	10 (5.2)
DAST-10 scores
3 or greater	46 (24.1)
2 or less	145 (75.5)
Relationship Length
2 years or less	86 (44.8)
more than 2 years	106 (55.2)
Sexual Arrangement
Monogamous	13 (6.8)
Non-monogamous	179 (93.2)
	M(SD)	M(SD)
Age	26.86 (2.78)	28.96 (6.78)

Table 1 also presents data on participants’ relationships and their relationship partners. On average, main partners were 28.96 years old (SD = 6.78). About a quarter indicated their main partner identified as Hispanic. Approximately half indicated their main partner identified as Non-Hispanic White. Most participants indicated their relationship partners identified as gay and had earned a 4-year degree or higher. The majority indicated their main partner was not living with HIV, and more than half further specified that their main partner was HIV-negative and did not have a current PrEP prescription. More than half indicated they had been in their relationship for 2 years or more.

4.1. Bivariate associations

4.1.1. . Correlations among adverse mental health outcomes and stigma

Table 2 displays bivariate correlations involving constructs of primary interest to the hypothesized model. Depression and anxiety were highly correlated. Enacted discrimination was positively associated with depression and anxiety as was internalized heterosexism and relationship marginalization.

Table 2.

Correlations among constructs of primary interest

	1	2	3	4	5	6	7	8	9	10	M(SD)
Mental Health outcomes
1. Depression	–	–	–	–	–	–	–	–	–	–	8.96 (5.57)
2. Anxiety	.715**	–	–	–	–	–	–	–	–	–	5.76 (4.81)
Stigma
3. Internalized Heterosexism	.304**	.287**	–	–	–	–	–	–	–	–	13.38 (5.56)
4. Enacted Discrimination	.258**	.271**	−.014	–	–	–	–	–	–	–	10.01 (6.58)
5. Relationship Marginalization	.218**	.150*	.386**	.238**	–	–	–	–	–	–	12.42 (4.11)
Relationship Quality
6. Rewards	−0.126	.017	−.227**	.008	−.371**	–	–	–	–	–	17.67 (2.13)
7. Costs	.357**	.266**	.327**	.048	.205**	−.212**	–	–	–	–	10.05 (3.52)
8. Approximation to an Ideal	−.250**	−.112	−.233**	−.089	−.334**	.682**	−.303**	–	–	–	15.25 (3.38)
9. Value of Alternatives	.247**	.092	.280**	.021	.294**	−.451**	.461**	−.466**	–	–	9.66 (3.27)
10. Investment	.151*	.135	0.005	.003	−.170*	.450**	.244**	.258**	−.294**	–	12.61 (2.01)
11. Barriers to Leaving	.377**	.327**	0.132	.200**	0.02	.230**	.340**	0.078	−0.019	.611**	14.80 (3.13)

NOTE:

^*p < .05;

^**p < .01.

4.1.2. Adverse mental health outcomes and relationship quality

Depression had the most pervasive associations with relationship quality. It was positively associated with two dimensions of poor relationship quality – perceived costs and valuation of alternatives – and negatively associated with the perception that the relationship approximated an ideal – an indicator of good relationship quality. Contrary to expectations, depression was positively associated with relationship investment and perceived barriers to leaving the relationship, typically indicators of good relationship quality. In contrast, anxiety was positively associated only with relationship costs and barriers to leaving the relationship.

4.1.3. Stigma and relationship quality.

Internalized heterosexism was negatively associated with two aspects of good relationship quality – perceive rewards and approximation to an ideal – and positively associated with two aspects of poor relationship quality – perceived costs and valuation of alternatives. The same pattern of associations was observed for relationship marginalization. In addition, relationship marginalization was negatively associated with investment. In contrast, enacted discrimination was only associated (positively) with barriers to leaving the relationship.

4.2. Path model results

The hypothesized model provided significantly better fit to the data compared to a null model (χ² (138) =311.84, p < .001) and it accounted for 33.6% of the variance in depression and 29.6% of the variance in anxiety. Associations among outcomes of primary interested are depicted in Figure 2. As anticipated, the residuals for depression and anxiety were significantly correlated with one another. Table 3 displays comprehensive path-model results for the regression of depression and anxiety. Tables 4a, 4b, and 4c display comprehensive regression coefficients for the regression of indicators of relationship quality.

Figure 2:

Path model results

NOTE: **p < .01; *p < .05; Pathways not depicted were non-significant at α < .05. All coefficients associated with direct paths are unstandardized regression coefficients (B).

Table 3.

Path Model Results for Adverse Mental Health Outcomes

	Depression					Anxiety
	R2 = .336, p < .001					R2 = .296, p < .001
Stigma	B	95% CI	β	p		B	95% CI	β	p
Internalized Heterosexism	0.137	(−0.022, 0.296)	.138	.092		0.153	(0.018, 0.287)	.178	.026
Enacted Discrimination	0.124	(0.000, 0.248)	.149	.049		0.118	(−0.000, 0.236)	.163	.051
Relationship Marginalization	0.072	(−0.109, 0.252)	.053	.437		0.045	(−0.126, 0.215)	.039	.608
Relationship Quality
Rewards	0.070	(−0.472, 0.613)	.027	.800		0.374	(−0.116, 0.863)	.168	.134
Costs	0.178	(−0.114, 0.469)	.113	.232		0.226	(−0.025, 0.476)	.167	.077
Approximation to an Ideal	−0.219	(−0.527, 0.090)	−.133	.164		−0.116	(−0.361, 0.130)	−.082	.356
Valuation of Alternatives	0.156	(−0.107, 0.418)	.092	.245		−0.049	(−0.277, 0.179)	−.034	.674
Investment	0.049	(−0.414, 0.512)	.018	.835		−0.184	(−0.593, 0.225)	−.078	.379
Barriers to Leaving	0.538	(0.267, 0.810)	.309	<.001		0.412	(0.168, 0.656)	.274	.001
Personal covariates
Problematic drug use	1.108	(−0.545, 2.762)	.086	.189		1.696	(0.166, 3.227)	.153	.030
Age	0.058	(−0.167, 0.283)	.029	.615		0.162	(−0.060, 0.385)	.095	.153
Race and ethnicity
Black or African American	−1.691	(−5.296, 1.914)	−.068	.358		−1.317	(−4.725, 2.091)	−.062	.449
Latino or Hispanic	0.279	(−1.142, 1.700)	.023	.701		−0.152	(−1.517, 1.214)	−.014	.828
Multiracial or another identity	−0.271	(−2.476, 1.934)	−.018	.810		−0.408	(−2.338, 1.523)	−.031	.679
Sexual Identity (referent = bisexual or queer)	−1.720	(−3.578, 0.138)	−.121	.070		−1.198	(−2.940, 0.544)	−.097	.178
Date of assessment (referent = pre-COVID)	−0.236	(−1.668, 1.197)	−.021	.747		−0.446	(−1.632, 0.739)	−.046	.461
Partner and relationship covariates
Age	−0.084	(−0.198, 0.030)	−.103	.146		−0.128	(−0.219, -0.037)	−.182	.006
Race and ethnicity
Black or African American	0.197	(−2.225, 2.620)	.011	.844		−0.140	(−2.277, 1.996)	−.009	.898
Latino or Hispanic	0.438	(−1.301, 2.177)	.034	.538		0.744	(−0.831, 2.320)	.067	.355
Multiracial or another identity	−1.764	(−3.738, 0.209)	−.100	.108		−1.231	(−3.062, 0.600)	−.081	.187
Sexual Identity (referent = bisexual or queer)	0.371	(−1.656, 2.398)	.022	.851		1.536	(0.076, 2.996)	.105	.039
Relationship Length (referent: ≤ 2 years)	0.355	(−1.163, 1.873)	.032	.719		−0.475	(−1.843, 0.739)	−.050	.496

Table 4a.

Coefficients for the regression of indicators of relationship quality

	Rewards					Costs
	R2 = .272, p < .001					R2 = .198, p < .001
	B	95% CI	β	p		B	95% CI	β	p
Stigma
Internalized Heterosexism	0.017	(−0.039, 0.074)	.044	.552		0.174	(0.056, 0.291)	.274	.004
Enacted Discrimination	0.047	(0.002, 0.092)	.144	.041		0.016	(−0.062, 0.095)	.030	.683
Relationship Marginalization	−0.178	(−0.258, -0.097)	−.341	<.001		0.065	(−0.058, 0.187)	.075	.301
Personal covariates
Problematic drug use	−0.512	(−1.146, 0.123)	−.102	.114		0.414	(−0.780, 1.608)	.050	.497
Age	−0.081	(−0.184, 0.023)	−.105	.126		0.017	(−0.143, 0.178)	.014	.832
Race and ethnicity
Black or African American	−0.065	(−1.413, 1.283)	−.007	.924		1.355	(−0.773, 3.482)	.086	.212
Latino or Hispanic	0.364	(−0.266, 1.283)	.076	.257		0.209	(−0.851, 1.270)	.027	.699
Asian, Pacific Islander	−0.708	(−1.631, 0.214)	−.120	.132		1.234	(−0.348, 2.816)	.127	.126
Sexual Identity (referent = bisexual or queer)	0.698	(−0.091, 1.487)	.126	−.083		−0.841	(−2.112, 0.429)	−.092	.194
Date of assessment (referent = pre-COVID)	−0.126	(−0.662, 0.411)	−.029	.646		−0.291	(−1.218, -.636)	−.041	.539
Partner and relationship covariates
Age	0.004	(−0.039, 0.046)	.012	.858		0.056	(−0.013, 0.125)	.108	.110
Race and ethnicity
Black or African American	−0.716	(−1.493, 0.062)	−.102	.071		0.191	(−1.357, 1.739)	.017	.809
Latino or Hispanic	−0.670	(−1.382, -0.042)	−.134	.065		0.229	(−0.815, 1.272)	.028	.667
Multiracial or another identity	0.610	(−0.334, 1.554)	.089	.205		0.951	(−0.703, 2.605)	.084	.260
Sexual Identity (referent = bisexual or queer)	0.078	(−0.749, 0.906)	.012	.853		−0.768	(−2.317, 0.781)	−.071	.331
Relationship Length (referent: ≤ 2 years )	0.899	(0.334, 1.465)	.209	.002		1.169	(0.076, 2.262)	.165	.036

Table 4b.

Coefficients for the regression of indicators of relationship quality (cont.).

	Approximation to Ideal					Valuation of alternatives
	R2 = .226, p < .001					R2 = .196, p < .001
	B	95% CI	β	p		B	95% CI	β	p
Stigma
Internalized Heterosexism	−0.044	(−0.152, 0.064)	−.073	.424		0.053	(−0.034, 0.140)	.090	.235
Enacted Discrimination	0.010	(−0.064, 0.084)	.020	.787		−0.025	(−0.096, 0.046)	−.050	.493
Relationship Marginalization	−0.219	(−0.355, -0.083)	−.268	.002		0.154	(0.030, 0.279)	.194	.015
Personal covariates
Problematic drug use	−0.936	(−2.008, 0.136)	−.119	.087		0.169	(−0.847, 1.185)	.022	.744
Age	0.007	(−0.162, 0.177)	.006	.933		0.004	(−0.176, 0.183)	.003	.967
Race and ethnicity
Black or African American	0.463	(−1.533, 2.459)	.031	.649		1.180	(−0.744, 3.104)	.080	.229
Latino or Hispanic	0.966	(−0.035, 1.967)	.129	.059		0.777	(−0.236, 1.790)	.106	.133
Multiracial or another identity	0.096	(−1.173, 1.365)	.010	.882		1.063	(−0.152, 2.279)	.118	.086
Sexual Identity (referent = bisexual or queer)	0.028	(−1.136, 1.191)	.003	.963		−0.750	(−1.897, 0.397)	−.089	.200
Date of assessment (referent = pre-COVID)	0.555	(−0.294, 1.405)	.081	.200		−0.246	(−1.125, 0.632)	−.037	.582
Partner and relationship covariates
Age	−0.015	(−0.079, 0.048)	−.031	.633		0.025	(−0.045, 0.094)	.051	.485
Race and ethnicity
Black or African American	−1.010	(−2.432, 0.411)	−.092	.164		1.427	(0.198, 2.657)	.134	.023
Latino or Hispanic	−1.591	(−2.734, -0.448)	−.202	.006		0.401	(−0.674, 1.475)	.052	.465
Multiracial or another identity	0.763	(−0.806, 2.332)	.071	.340		0.402	(−1.019, 1.823)	.038	.579
Sexual Identity (referent = bisexual or queer)	0.942	(−0.323, 2.207)	.091	.144		−1.024	(−2.322, 0.274)	−.102	.122
Relationship Length (referent: ≤ 2 years )	1.054	(0.075, 2.032)	.156	.035		−0.640	(−1.623, 0.344)	−.097	.202

Table 4c.

Coefficients for the regression of indicators of relationship quality (cont.).

	Investment					Barriers to leaving
	R2 = .248, p < .001					R2 = .280, p < .001
	B	95% CI	β	p		B	95% CI	β	p
Stigma
Internalized Heterosexism	0.081	(0.026, 0.137)	.223	.004		0.158	(0.072, 0.244)	.277	<.001
Enacted Discrimination	0.024	(−0.018, 0.066)	.077	.267		0.112	(0.043, 0.180)	.232	.001
Relationship Marginalization	−0.079	(−0.158, 0.001)	−.159	.052		−0.020	(−0.148, 0.109)	−.026	.763
Personal demographic covariates
Problematic drug use	−0.489	(−1.136, 0.158)	−.103	.139		−0.307	(−1.248, 0.631)	−.042	.521
Age	−0.153	(−0.249, −0.056)	−.209	.002		−0.252	(−0.410, −0.094)	−.222	.002
Race and ethnicity
Black or African American	−0.250	(−1.531, 1.031)	−.027	.702		−0.157	(−2.418, 2.105)	−.011	.892
Latino or Hispanic	−0.228	(−0.896, 0.441)	−.050	.504		−0.914	(−1.928, 0.099)	−.129	.077
Multiracial or another identity	−0.363	(−1.173, 0.447)	−.065	.379		−0.028	(−1.347, 1.290)	−.003	.966
Sexual Identity (referent = bisexual or queer)	0.486	(−0.288, 1.260)	.093	.218		0.801	(−0.282, 1.884)	.098	.147
Date of assessment (referent = pre-COVID)	0.480	(−0.073, 1.033)	.116	.089		0.558	(−0.251, 1.368)	.087	.176
Partner and relationship covariates
Age	−0.001	(−0.041, 0.039)	−.003	.965		−0.009	(−0.073, 0.055)	−.019	.781
Race and ethnicity
Black or African American	−0.757	(−1.699, 0.186)	−.114	.116		−1.630	(−2.812, −0.447)	−.158	.007
Latino or Hispanic	−0.414	(−1.021, 0.193)	−.087	.181		−0.829	(−1.846, 0.188)	−.112	.110
Multiracial or another identity	0.541	(−0.271, 1.353)	.083	.192		0.662	(−0.436, 1.761)	.065	.230
Sexual Identity (referent = bisexual or queer)	0.232	(−0.631, 1.095)	.037	.598		−0.442	(−1.716, 0.832)	−.045	.497
Relationship Length (referent: ≤ 2 years )	1.585	(1.023, 2.146)	.389	<.001		2.251	(1.319, 3.184)	.354	<.001

Internalized Heterosexism.

Consistent with hypotheses, internalized heterosexism had a statistically significant and positive direct association with anxiety; however, its association with depression was non-significant. Consistent with expectations, internalized heterosexism was positively associated with perceptions of relationship costs (an indicator of poor relationship quality); however, it was also positively associated with relationship investment and perceptions of barriers to leaving the relationship (indicators of good relationship quality) and unrelated to perceptions of rewards, the extent to which the relationship approximates an ideal, and valuation of relationship alternatives.

Enacted Discrimination.

Consistent with hypotheses, enacted discrimination had a statistically significant and positive direct association with depression; however, its association with anxiety was non-significant. Contrary to expectations, enacted discrimination was associated (positively) with perceptions of rewards and barriers to leaving the relationship, and associations with other relationship quality subscales were uniformly non-significant.

Relationship Marginalization.

Contrary to expectations, relationship marginalization was not directly associated with depression or anxiety scores. Consistent with expectations, relationship marginalization was negatively associated with two indicators of good relationship quality, including: perceptions of rewards and the extent to which the relationship approximated an ideal. Likewise, consistent with expectations, it was positively associated with valuation of alternatives an indicator of poor relationship quality). In contrast, it was unrelated to perceived costs and barriers to leaving the relationship.

Relationship quality.

One domain of relationship quality – the perception of barriers to leaving the relationship – was significantly associated with anxiety and depression. Although this subscale assesses a construct that has typically been construed to indicate a relationship strength (the more barriers to leaving a relationship, the greater a respondent’s sense of commitment), the association between barriers to leaving the relationship and depression was positive in direction. Also, contrary to expectations, no other domains of relationship quality were significantly associated with depression.

Indirect pathways linking stigma and depression.

Based on the pattern of direct associations, two of the originally hypothesized indirect pathways from stigma to depression were plausible (see Figure 2). The first pathway comprised the direct association between internalized heterosexism and barriers to leaving the relationship as well as the direct association between barriers to leaving the relationship and depression. This indirect pathway was statistically significant. Internalized heterosexism was positively associated with depression scores in part through variance shared with barriers to leaving the relationship. The second pathway comprised the direct association between enacted discrimination and barriers to leaving the relationship as well as the direct association between barriers to leaving the relationship and depression. This indirect pathway was also statistically significant. Enacted discrimination was positively associated with depression scores in part through variance shared with barriers to leaving the relationship.

Indirect pathways terminating in anxiety were similar to those involving depression. Two of the originally hypothesized indirect pathways were plausible based upon the pattern of direct associations. The first comprised the direct association between internalized heterosexism and barriers to leaving the relationship as well as the direct association between barriers to leaving the relationship and anxiety. This indirect pathway was statistically significant. Internalized heterosexism was positively associated with anxiety scores in part through the variance shared with barriers to leaving the relationship. The second pathway comprised the direct association between enacted discrimination and barriers to leaving the relationship as well as the direct association between barriers to leaving the relationship and anxiety. This indirect pathway was also statistically significant. Enacted discrimination was positively associated with anxiety scores in part through the variance shared with barriers to leaving the relationship.

4.3. Post hoc analysis: Sequential versus simultaneous entry

In the hypothesized model, relationship quality subscales were entered simultaneously as competing predictors of mental health outcomes. This use of MDRCI subscales aligns with previous studies of SMM relationships involving this scale. The approach was further supported by inspection of bivariate correlations (in Table 2) for concerns regarding collinearity. These were uniformly below .70 and most (all but two) fell between .20 and .47. With simultaneous entry, regression coefficients represent the uniquely shared variance between a given predictor and the outcome. In this context, findings indicate that barriers to leaving the relationship emerged as having the only statistically significant unique associations with mental health outcomes after controlling for other components of relationship quality and model covariates.

Although simultaneous entry has the benefit of quantifying the unique association of a given predictor with an outcome, parsing out variance shared among multiple predictors may diminish associations that would be significant if these were examined in isolation. For example, although barriers to leaving the relationship was the only component of relationship quality that was significantly associated with depression and anxiety in the model using simultaneous entry, 5 components (costs, approximation to an ideal, valuation of alternatives, investment, and barriers to leaving the relationship) had significant bivariate associations with depression and 2 (costs and barriers to leaving the relationship) had significant bivariate associations with anxiety. An alternative analytic approach might involve a sequence of models in which each component of relationship quality was examined individually or orthogonally from the others. This would quantify the association of each relationship quality component with depression and anxiety, irrespective of the variance shared with other components. Results of these alternative models and a brief summary comparing and contrasting findings with those produced by simultaneous entry are provided in Supplemental Materials.

5.0. Discussion

This paper is among the first to test a priori hypothesized models in a sample of partnered SMM linking internalized, enacted, and relationship stigma to mental health outcomes through indirect pathways involving relationship quality. Where significant, results generally aligned with the psychological mediation hypothesis and its specific premise that disruption of interpersonal relationships is one mechanism linking stigma to mental health (Hatzenbuehler, 2009). Findings suggested that internalized, enacted, and relationship stigma were all in some way associated with one or more indicators of relationship quality, and some indicators of relationship quality were in turn associated with depression and anxiety (specifically the perception of barriers to leaving the relationship). Both internalized heterosexism and relationship marginalization were positively associated with depression and anxiety through indirect pathways involving the perception of barriers to leaving the relationship.

5.1. Internalized and enacted stigma: Direct and indirect associations with mental health

This study went beyond previous investigations by testing three different forms of stigma (internalized stigma, enacted stigma, and relationship marginalization) as additive or competing predictors in multivariable models. These findings broadly align with the larger literature on internalized and enacted stigma and mental health outcomes (Curtis et al., 2022; Lattanner et al., 2022). Internalized and enacted stigma were directly associated with anxiety and depression respectively in the observed path model (as hypothesized). These associations were detected above and beyond relationship marginalization in this sample of SMM in relationships. As described in supplemental materials, models in which relationship quality components were tested sequentially rather than simultaneously generally resulted in both internalized and enacted stigma having statistically significant and positive associations with depression and anxiety, consistent with previous findings (Curtis et al., 2022; Lattanner et al., 2022).

In addition to these direct associations, internalized and enacted stigma were positively associated with depression and anxiety indirectly through the perception of barriers to leaving the relationship. In some respects, the organization and direction of these associations broadly aligned with the psychological mediation hypothesis (Hatzenbuehler, 2009). Their valence was positive (more internalized or enacted stigma was indirectly associated with greater depression as well as anxiety). The dimension of relationship quality involved – perceptions of barriers to leaving the relationship – was not anticipated.

The perception of barriers to leaving the relationship has generally been understood in the context of interdependence theory as an aspect of relationship maintenance, a cognitive strategy used to increase commitment to the long-term durability of a relationship (Rusbult, 1980, 1983). Contrasting with this characterization as a relationship strength, the perception of barriers to leaving the relationship was positively associated with internalized and enacted stigma as well as depression and anxiety – a pattern more aligned with a psychological risk factor. This mirrors the findings of at least one previous study of partnered SMM (Starks, Robles, Bosco, et al., 2019) in which barriers to leaving the relationship was positively associated with the frequency of illicit drug use (but not cannabis use).

Stanley, Rhoades, and Whitton (2010) suggested that constraint components of commitment (such as barriers to leaving a relationship) are typically not aversive until or unless the relationship ceases to be sufficiently satisfying to the individual. Another possible explanation is that the barriers to leaving the relationship subscale is assessing a construct that is meaningfully different from a relationship maintenance strategy in these samples of partnered SMM, many of whom have non-monogamous sexual agreements. The scale comprises items that refer to leaving the relationship as a source of emotional pain or suggest the respondent would feel an ongoing attachment to their former partner or a sense of obligation to maintain the relationship (Kurdek, 1995, 1998). For some partnered SMM, particularly those who have experienced stigma and discrimination, these items may activate negative internal working models of the self (Ainsworth, 1985; Bowlby, 1977). Viewed as such, the items may be sensitive to a respondent’s belief that they are indebted to their partner or would be unable to cope successfully with the emotions involved in relationship termination. It is also plausible that for some, their primary relationship partner provides a source of safety from the threat of stigma. This might increase their sense of loss if the relationship were to end. Future research may advance understanding of relationship quality for male couples by re-examining the construct of barriers to leaving a relationship and assessing the relationship maintenance strategies deployed by male couples.

5.2. Relationship marginalization: A predictor of diminished relationship quality

Relationship marginalization was not directly associated with mental health outcomes in multivariable models and indirect associations were non-significant as well. As described in supplemental materials, this remained true even when relationship quality subscales were examined sequentially rather than simultaneously. Nevertheless, these findings underscore the extent to which relationship marginalization is associated with diminished relationship quality. In bivariate analyses, it was negatively correlated with three indicators of good relationship quality (rewards, approximation to an ideal, and investment) and positively associated with two indicators of poor relationship quality (costs and valuation of alternatives). For two aspects of good relationship quality (rewards and approximation to an ideal) as well as one aspect of poor relationship quality (valuation of alternatives), these associations persisted in multivariable models.

These findings align with previous studies linking relationship marginalization to relationship quality in sexual minority women (Whitton et al., 2024) and transgender women (Gamarel et al., 2014). Previous studies involving SMM in relationships have largely examined whether relationship quality moderated associations between relationship marginalization and mental health (Robles et al., 2022; Rosenthal & Starks, 2015), rather than testing the association between them in ways aligned with the psychological mediation hypothesis. This study also goes beyond previous research by testing relationship marginalization as an additive predictor or covariate in models including internalized and enacted stigma.

5.1. Clinical implications

Existing interventions that address the impact of coping with stigma for sexual and gender minority people are presently focused on individuals (Pachankis et al., 2022; Parsons et al., 2017). Similarly, the mostly widely circulated guidance for conducting LGBTQ affirmative cognitive behavioral therapy is focused on individual interventions (Pachankis et al., 2022). Comparable work on the development of affirmative psychotherapies for couples is limited to pilot investigations (Pentel et al., 2021). Although brief motivational interviewing has demonstrated some potential to reduce perceptions of stigma when delivered to individual sexual minority youth living with HIV (Budhwani et al., 2021), analogous effects of motivational interviewing with couples (Starks, 2022) have not yet been tested.

The effects of stigma on relationship quality may be addressed through a variety of interventions, including communication skills building, behavioral activation that increases pleasant and rewarding interactions between partners, or the deployment of coping skills that help partners respond to experiences of stigma and discrimination. For example, clinicians working with couples might focus on fostering effective communication skills, cognitive restructuring, or social support building to address stigma. For those in relationships, this might include guiding partners to seek each other’s support in ways that validate, rather than dismiss, each other’s experiences—something that can be especially important for interracial couples or those couples with differing gender presentations. Clinicians can teach active listening techniques and encourage partners to recognize and validate emotions, whether they stem from internalized heterosexism or enacted stigma. Existing intervention protocols developed for use with male couples already incorporate elements of relationship or communication skill building (Newcomb et al., 2023; Starks et al., 2022; Starks, Dellucci, et al., 2019; Wu et al., 2011). These might be leveraged and adapted to guide such practice.

5.2. Limitations

This study focused on a highly specific subgroup of participants – cisgender SMM in primary relationships with other (cisgender) men; who use drugs; and engage in condomless sex with casual partners or main partners who are living with HIV. The focus on this relatively circumscribed population arises from the epidemiology of HIV infection in the US. SMM account for the majority of new HIV infections in the US (Centers for Disease Control and Prevention, 2024); many, possible most, new HIV infections among SMM are transmitted between primary partners (Goodreau et al., 2012; Sullivan et al., 2009). Mental health problems (e.g., symptoms of depression or anxiety) and drug use represent synergistic and co-occurring psychosocial risk factors that additively compound HIV infection risk for SMM generally (Stall et al., 2008; Stall et al., 2003) and those in relationships specifically (Starks et al., 2016). Although understanding factors associated with mental health outcomes among partnered SMM at high risk of HIV infection is therefore of particular public health significance, these findings may not be representative of the general population of SMM or sexual and gender minority people broadly.

Whereas the internalized heterosexism measure was specifically relevant to sexual identity, both the enacted discrimination and relationship marginalization scales used in this study were sensitive to experiences of stigma irrespective of the identity or perceived motivation for that discrimination. The items of the enacted discrimination scale that assess how often participants experience various forms of discrimination do not ask them to differentiate between events they perceive as directed at their sexual identity versus other aspects of identity (e.g., race and ethnicity, culture, language, religious affiliation). Similarly, the relationship marginalization scale assesses perceived acceptance of one’s relationship by peers, family, and society at large, regardless of the cause. The sex and gender composition of a couple is only one reason why a relationship might be subject to marginalization. Although their use in this study is consistent with previous (Robles et al., 2022; Rosenthal & Starks, 2015), future studies should examine whether effects differ if sexual identity-specific measures of enacted stigma and relationship marginalization are employed.

Generalizability is further limited by eligibility criteria and other aspects of study design. Respondents were aged 18 to 35, most were in non-monogamous relationships, and all reported recent drug use. The majority of participants are from the Northeast and largely urban areas. Most participants were recruited through social media and geospatial networking applications. As such, findings may not be representative of the experiences of SMM who do not utilize these platforms. As with any cross-sectional examination of indirect effects, these findings should not be construed as connoting causal connections among constructs (Kline, 2015). Although the organization of hypothesized models was guided by theory – specifically the psychological mediation hypothesis (Hatzenbuehler, 2009) – alternative equivalent models are plausible and longitudinal data would be necessary to establish temporal order.

5.3. Conclusions

Despite these limitations, this study is among the first to directly test the premise that stigma may be associated with mental health in part through associations with relationship quality. Although results generally conformed to assumptions of the psychological mediation hypothesis, they also provided some indication that studies in this area may come to variable conclusions depending upon which outcomes, forms of stigma, and dimensions of relationship quality they examine. Findings suggest that the development of interventions that support couples’ coping with stigma may meaningfully enhance the mental health of SMM in relationships.

Public Significance Statement:

Experiences of stigma, including internalized and enacted stigma directed at an individual as well as directed at a primary sexual-romantic relationship, were negatively associated with relationship quality. On the other hand, those with better relationship quality reported fewer symptoms of anxiety and depression. These findings suggest that stigma is associated with mental health not only through individual experiences but also, at least in part, through its impact on relationship processes.