An Analysis of Chronic Stress, Substance Use, and Mental Health Among a Sample of Young Sexual Minority Men in New York City: The P18 Cohort Study

Abstract

Introduction: Sexual minority men (SMM) are at increased risk for psychosocial stressor exposure, substance use, and poor mental health relative to heterosexual men. While the burden of mental health is growing in the United States, among SMM these trends are increasing at a greater rate, driving health disparities. Methods: Framed within a minority stress framework, these analyses examine how stressors explain substance use and poorer mental health over time. Participants were asked questions on stressor exposure (stigma, discrimination, internalized homophobia, perceived stress), mental health (anxiety, depression, PTSD), and substance use (alcohol to intoxication, club drugs, poly club drugs) over 36 months among 528 SMM in NYC. Results: Perceived stress increased frequency of all substance use, whereas discrimination decreased days of club and poly club drug use. Depression severity predicted increased days of club drug and poly club drug use. PTSD severity predicted increased days of club drug and poly club drug use. Conclusion: We are able to expand on the literature with granular substance use data to highlight associations with stressors and mental health. These findings support an increased need for systematic policy solutions and public health interventions to address drivers of substance use disparities among young SMM.

1. Introduction

Depression is expected to be the leading cause of global disease by 2030 (World Health Organization, 2008). The mental health burden is growing in the United States, with approximately 1.5 million additional adults reporting poor mental health each year (Mental Health America, 2020). Emerging young adulthood (19–25 years old) is a highly vulnerable and unique time of identity exploration (Arnett, 2000). Young adult sexual and gender minorities are at increased risk for depression (Alvy et al., 2011; Marshal et al., 2011), anxiety (Batchelder et al., 2017; S. Sun et al., 2020), and post-traumatic stress disorder (PTSD) (Beidas et al., 2012). A recent cross-sectional analysis in the United States found 49% of sexual minorities self-reported depression, compared to 19.5% of heterosexual participants (Miller et al., 2024). Self-reported mental health is important as a formal diagnosis only captures the disorder in a binary manner, but an individual’s mental health state lies along a spectrum (Keyes, 2002).

Substance use trends in the United States are stable or decreasing (Miech et al., 2023), but not for sexual and gender minorities (SGMs) who are at increased risk for substance use (Caputi et al., 2018). Lesbian, gay, and bisexual individuals have triple the prevalence of opioid use (Substance Abuse and Mental Health Services Administration, 2018), higher prevalence of binge alcohol use (Substance Abuse and Mental Health Services Administration, 2018), and higher prevalence of club drugs compared to heterosexual individuals (Halkitis et al., 2005a, 2005b). Among sexual minority men (SMM), bisexual men have higher prevalence relative to gay men (Substance Abuse and Mental Health Services Administration, 2018). Although alcohol use may be similar between SGM and cisgender heterosexual men, SGM face worse alcohol-related consequences (Schipani-McLaughlin et al., 2022).

Results from the 2017 National Survey on Drug Use and Health (NSDUH) indicate that gay men had the highest prevalence of “chemsex” drugs (defined as using club drugs including ecstasy, ketamine, GHB, and amyl nitrite) with 45.3% reporting past-year use (Rosner et al., 2021). Results from the 2019 NSDUH indicate sexual minority men had higher prevalence of substance use disorder (15.5%) compared to heterosexual men (10%) (Hodges et al., 2023). There is ample evidence of the importance of the social environment on club drug use prevalence among young SMM, particularly in New York City (NYC) (Halkitis & Palamar, 2006; Johnston et al., 2008). This study will expand upon a previous cohort of young sexual minority men in NYC (Halkitis et al., 2014) in order to provide further follow-up and identify more recent data related to stress, mental health, and substance use.

Several conceptual models describe how stressors in a minority population lead to both substance use and mental health outcomes (Hatzenbuehler, 2009; Meyer, 2003). Minority Stress Theory (MST) indicates that minority populations are subject to both heightened chronic stressors and more frequent acute stressors related to their minoritized identity over the life course and are additionally subject to stressors specific to marginalized populations such as discrimination, prejudice, and stigma (Meyer, 2003).

For prior studies examining health disparities among SGM, MST is the most prominent underlying theoretical framework (Parent et al., 2018). Much of the literature using these models utilizes a cross-sectional design which may fail to capture the latent effect of these stressors on mental health. Several recent studies utilize longitudinal design and methodology, which is advantageous for studying cause and effect through mediation analysis and observing changes in individuals over the course of a study, but have several limitations in how the sample was categorized for analysis. Most studies utilize the term “Men who have sex with men” (MSM), but this term is inadequate because it categorizes individuals based solely on behavior rather than identity, failing to capture meaningful differences in risk and experiences among SMM (Galupo et al., 2015), for example, between gay and bisexual men (Chan et al., 2020; Green & Feinstein, 2012; Semlyen et al., 2016).

Many existing longitudinal studies lack sufficient detail describing substance use behavior or collapse substances into broad categories solely for sample size purposes or due to primary data source limitations, preventing the nuances of substance use behavior from being fully understood. Prior literature often specifically groups club drugs into broader categories with non-club drugs, such as “other drugs” (Parent et al., 2018), impacting interpretability. While poly substance use among sexual minorities has been studied, prior studies look at past-year use (Kecojevic et al., 2016) or lifetime prevalence (Turpin et al., 2020) or, due to the way data was collected, report poly substance use without being able to specify each specific substance (McCabe et al., 2022). Our study design allows us to expand upon the literature, as we have sufficient sample size of sexual minority men and enough granular substance use data to extend current knowledge.

1.1. Theoretical Framework

This study aims to understand how psychosocial stressors affect mental health and substance use outcomes in a diverse sample of SMM. According to Minority Stress Theory (Meyer, 2003), sexual minority individuals experience unique stressors which can be distal (indirect, chronic) or proximal (direct, immediate). Proximal stressors (i.e., internalized homophobia) originate within the individual, while distal stressors (i.e., discrimination) are external to the individual. Our study is grounded within this framework and extends previous work by analyzing how these stressors affect mental health and substance use outcomes. Further, by including a temporal component in our analyses and analyzing substance use, a contextually relevant level (i.e., club drugs and poly club drugs) for our population, our study will clarify the complex pathways in which minority stress impacts health over time.

1.2. Current Study

Given these crucial challenges there is a need to understand more nuanced substance use behavior among young SMM in order to contribute to the overall understanding of the complex interactions between chronic stressors, substance use, and mental health.

This study (a) provides unique evidence for the impact of minority stress on moderating substance use on a more granular substance level than exists in the literature by adequately reporting poly substance use. (b) We expand on prior studies by exploring substances of crucial importance to SGM, including club drugs and the combination of club drugs with other substances. This study also (c) builds upon the gaps mentioned previously by including a time component in our analyses. Findings from this study can inform targeted public health strategies accounting for unique challenges related to minority stress.

2. Materials and Methods

2.1. Study Design and Sample Characteristics

This is a secondary analysis of data from the Project 18 (P18) Cohort Study, a longitudinal study of a young, racially and ethnically diverse sample of SMM in NYC. There have been two waves of recruitment: Wave 1 (2009) and Wave 2 (2014). Eligibility for this study required participating in Wave 2. We limited our sample to Wave 2 participants to ensure the use of more recent data and because participants were 22–23 years old rather than 18–19, which is important given the legal drinking age of 21 and our focus on substance use behaviors. Approximately half of those in Wave 1 continued into Wave 2. Follow-up occurred every 6 months for three years. More details of P18 are described elsewhere (Halkitis et al., 2021; LoSchiavo et al., 2021; Martino et al., 2021; Stults et al., 2021).

Wave 2 recruitment inclusion required being 22 or 23 years old, assigned male at birth, reported sex with another man in the past six months, self-reported HIV negative or unknown serostatus, and place of residence being NYC.

Participants were recruited using non-probability sampling in 2014 from social media sites, NYC streets and parks, college campuses and dormitories, gay-identified events, and community centers. For this sample, 73.9% were recruited using social networking websites and dating apps. Snowball sampling was used where participants could refer other individuals to the study.

Individuals were screened over the phone to confirm eligibility. If criteria were met, individuals had an appointment at the Center for Health, Identity, Behavior and Prevention Studies at the New York University Washington Square Campus. At this visit, individuals confirmed they met age requirements by providing photo identification, completed a brief screening tool, and provided written consent to participate. After consenting to an oral HIV antibody test, baseline measurement was collected with the OraQuick Advance^® rapid HIV ½ antibody test.

Wave 2 of recruitment opened in 2014, and 274 participants (41.2% of the Wave 1 cohort) were retained from Wave 1. Further, 391 new participants were enrolled, providing a total sample of 665 participants. Then, 528 participants (79.4%) were retained at the first 6-month follow-up, and 416 participants (62.6%) were retained until the end of the study at the 36-month follow-up. Ethical approval for this study was gained from the Institutional Review Boards at New York University (#10-6802) and written informed consent from each participant was obtained prior to participation.

Participants self-reported race, ethnicity, sexual orientation, and income completed an audio computer-assisted self-interview (ACASI) assessment at baseline. For sensitive information, there is evidence that ACASI minimizes skipping of questions (Metzger et al., 2000). Race and ethnicity were categorized as Hispanic, Asian/non-Hispanic, Black/non-Hispanic, White/non-Hispanic, Multi-racial/non-Hispanic, or Other/non-Hispanic. Perceived family SES was measured using a 5-Item Likert scale (“Lower”, “Lower Middle”, “Middle”, “Upper Middle”, “Upper”). Individuals were able to self-identify sexual orientation from the following choices: “heterosexual or straight”, “gay or lesbian”, and “bisexual”. Subjects who specified “heterosexual or straight” were excluded from analysis.

2.2. Stressor Indicators

Stressor indicators were chosen based upon those used in a prior P18 cohort study (Tran et al., 2023). Discrimination was measured using a 5-item scale from P18. These five items include several indicators of discrimination. Three indicators are binary, such as: “In the last 6 months, have you been discriminated against by anyone at your current workplace(s)” and “In the last 6 months, have you been discriminated against by anyone at your current school”. The other two indicators have a range to indicate discrimination frequency or severity. For example, “Which of the following reasons do you believe you have been discriminated against … (select all that apply)” will have a score of 0 if no reasons have been selected or 4 if all four reasons have been selected. Responses to all items are summed such that a higher score indicates greater discrimination.

Gay-related stigma (hereafter referred to as stigma) was measured with the 4-Item Gay Related Stigma Scale, Revised (Berger et al., 2001; Wright et al., 2007) which measures stigma related to gay identity. An example item from this scale includes “I have stopped socializing with some people because of their reactions of my being gay/bisexual”. Items are scored on a 4-point Likert scale ranging from 1 (strongly disagree) to 4 (strongly agree). Responses are summed such that higher scores indicate greater gay-related stigma.

Internalized homophobia was measured using the 4-Item Internalized Homophobia Scale (Thiede et al., 2003). Items are scored on a 4-point Likert scale ranging from 1 (strongly disagree) to 4 (strongly agree). Specific items in this scale assess self-rejection, discomfort, shame, or guilt of one’s sexual identity. Responses are summed such that higher scores indicate greater internalized homophobia.

Perceived stress was measured using the 10-item Perceived Stress Scale (Cohen et al., 1983). Items are on a 5-point Likert scale ranging from 0 (never) to 4 (very often). Items in this scale assess feelings of being overwhelmed, having a lack of control in one’s life, or feeling like things are “piling up” in one’s life. Responses are summed such that higher scores indicate greater perceived stress. The means and standard deviations of discrimination, stigma, internalized homophobia, and perceived stress can be found in

Table A1. Variable Means and Standard Deviations.

Variable	Mean	Standard Deviation
Stigma	5.86	2.68
Discrimination	0.25	0.63
Internalized Homophobia	7.03	7.71
Perceived Stress	12.66	14.98

of Appendix A.

2.3. Substance Use Indicators

Substance use behavior was collected with 30-day Timeline Followback (TLFB) assessments which utilize a calendar-based measure to ascertain use during the 30 days preceding the interview. Participants were asked to report days they used the following substances: alcohol, marijuana, inhalant nitrates, powder cocaine, methylenedioxymethamphetamine (MDMA/ecstasy), gamma-hydroxybutyrate (GHB), ketamine, crack cocaine, heroin, rohypnol, methamphetamine, and pharmaceuticals without a prescription: PDE-5 inhibitors (i.e., Cialis^®, Levitra^®, Viagra^®), opioid pain relievers (e.g., Percocet^®, Oxycontin^®), benzodiazepines (e.g., Valium^®, Xanax^®), and ADHD drugs/stimulants (e.g., Adderall^®, Ritalin^®, Concerta^®). Based on disaggregated data, summary drug use variables were created by totaling the number of days that participants reported alcohol to intoxication, club drugs, and using at least one club drug and at least one other substance (poly club drug days). We defined club drug use as use of any of the following substances: methamphetamine, MDMA/ecstasy, ketamine, cocaine, and GHB. We defined poly club drug use as using at least one of these club drugs (methamphetamine, MDMA/ecstasy, ketamine, cocaine, and GHB) and at least one additional non-club drug substance.

2.4. Mental Health Indicators

Three mental health outcomes were measured: anxiety, depression, and PTSD.

First, the 21-Item Beck Anxiety Inventory (BAI), which consists of a Likert-type scale ranging from 0 (not at all) to 3 (severely), was used to assess the severity of anxiety symptoms experienced (Beck et al., 1988). Items in the BAI explore symptoms such as dizziness, shortness of breath, heart palpitations, and fear.

Second, the 21-item Beck Depression Inventory, 2nd Edition (BDI-II), which consists of a Likert-type scale ranging from 0 (not at all) to 3 (severely), was used to assess the severity of depressive symptoms experienced (Beck et al., 1996). Items in the BDI-II explore mood-related symptoms such as sadness, hopelessness, and worthlessness, as well as behavioral symptoms like changes in sleep and physical symptoms such as fatigue.

Finally, the 17-item PTSD Checklist, which consists of a Likert-type scale ranging from 1 (not at all) to 5 (extremely), was used to assess the severity of post-traumatic stress disorder symptoms. Items in this checklist explore recurring traumatic memories, distressing dreams, avoidance, emotional numbness, and feelings of detachment from others (Blanchard et al., 1996). Validity and reliability of these measures on P18 data have been published and discussed elsewhere (Ompad et al., 2016).

2.5. Analytic Plan

Our analytic plan follows the methodology of a recently published P18 paper (Tran et al., 2023). Descriptive statistics were used to summarize the distribution and data missingness and measure reliability of the variables. Substance use and mental health as count data appeared to have higher-than-expected observations of zero values. Therefore, zero-inflated mixed effects Poisson or negative binomial regression models were applied to analyze the association of risk factors with the outcome of interest (Atkins & Gallop, 2007). The choice of either a Poisson or negative binomial regression model was determined based on overdispersion tests and comparison of model fit indices.

To account for the repeated measures within the data, all models included random intercepts at the participant level to model within-person correlation over time. The time component was modeled as a categorical fixed effect to account for non-linear change across visits. Covariates in the model were either time-varying (i.e., perceived stress, discrimination, stigma, internalized homophobia) or time-invariant (i.e., baseline characteristics such as race and ethnicity).

We ran separate models for each outcome, and all covariates were included in each model. We conducted residual diagnostics to ensure that the chosen model adequately captured the data characteristics, to evaluate whether the model assumptions were met, and to check for overdispersion. Visual diagnostics, such as Q-Q plots and residual plots, were used to assess if the residuals were appropriately distributed. Although we did not conduct a variance inflation factor analysis, we did not observe unusually large standard errors or model convergence issues in our multi-variable models, suggesting severe multi-collinearity was unlikely. Model fit statistics can be found in

Table A2. Variable Means and Standard Deviations of Table 2.

Outcome	Model	Akaike Information Criterion	Bayesian Information Criterion	Log-Likelihood Value
Alcohol to Intoxication	Zero-Inflated Poisson	12,216.83	12,307.07	−6093.414
	Zero-Inflated Negative Binomial	13,349.94	13,440.18	−6659.968
Club Drug	Zero-Inflated Poisson	4773.443	4869.699	−2370.722
	Zero-Inflated Negative Binomial	5235.961	5326.200	−2602.980
Poly Club Drug	Zero-Inflated Poisson	2648.805	2745.023	−1308.402
	Zero-Inflated Negative Binomial	*	*	*

*: Indicates model does not converge.

and

Table A3. Variable Means and Standard Deviations of Table 3.

Outcome	Model	Akaike Information Criterion	Bayesian Information Criterion	Log-Likelihood Value
Anxiety	Zero-Inflated Poisson	17,732.41	17,816.64	−8852.204
	Zero-Inflated Negative Binomial	10,905.57	10,995.82	−5437.787
Depression	Zero-Inflated Poisson	17,342.54	17,426.77	−8657.272
	Zero-Inflated Negative Binomial	12,607.62	12,697.87	−6288.811
PTSD	Zero-Inflated Poisson	17,732.41	17,816.64	−8852.204
	Zero-Inflated Negative Binomial	15,962.13	16,052.38	−7966.066

of Appendix A.

3. Results

The sample was diverse in terms of sociodemographic variables (

Table 1. Baseline Sample Characteristics of the P18 Wave 2 Study Sample (n = 528).

	n (%)
Age, Years
Mean (SD)	22.5 (0.613)
Race
Asian/Multi-racial/Native American/Other non-Hispanic	92 (17.4%)
Black non-Hispanic	135 (25.6%)
Hispanic/Latino	165 (31.3%)
White non-Hispanic	136 (25.8%)
Sexual Identity
Bisexual	78 (14.8%)
Gay	436 (82.6%)
Missing	14 (2.7%)
Educational Attainment
Junior high school diploma	12 (2.3%)
High school diploma or GED	240 (45.5%)
Associate degree	57 (10.8%)
Bachelor’s degree	212 (40.2%)
Graduate degree	6 (1.1%)
Missing	1 (0.2%)
Income
USD 0 to USD 9999	226 (42.8%)
USD 10,000 to USD 19,999	107 (20.2%)
USD 20,000 to USD 34,999	101 (21.0%)
USD 35,000 to USD 54,999	47 (8.9%)
USD 55,000 to USD 74,999	12 (2.2%)
USD 75,000 to USD 99,999	4 (0.8%)
USD 100,000 and over	2 (0.4%)
Missing	19 (3.6%)

). For each separate mental health outcome, discrimination, internalized homophobia, and perceived stress were significantly associated with a more severe mental health score. For PTSD, there was also a positive association with stigma (β = 0.050; 95% CI −0.0003, 0.101; p < 0.1). For all mental health outcomes and after controlling for covariates, Black non-Hispanic participants had, on average, the least severe mental health outcomes. Being below the NYC poverty line was associated with more severe anxiety and PTSD but not depression. Compared to baseline, anxiety scores showed no significant change at later visits. Depression and PTSD showed significant increases at several subsequent visits, with depression having the largest increase at the fourth visit (β = 0.221, p < 0.0001). These results are presented in

Table 2. Zero-Inflated Negative Binomial Regression Model Results of Mental Health Outcomes, Stressors, and Baseline Covariates.

	Anxiety		Depression		PTSD
Predictor	Estimate	s.e.	Estimate	s.e.	Estimate	s.e.
Stigma	0.010	0.027	0.024	0.023	0.050 *	0.026
Discrimination	0.240 *	0.123	0.175 *	0.104	0.211 *	0.128
Internalized Homophobia	0.058 ****	0.016	0.067 ****	0.014	0.078 ****	0.016
Perceived Stress	0.039 ****	0.007	0.017 ****	0.006	0.040 ****	0.07
Race/Ethnicity (Ref = White/non-Hispanic)
Asian/Multi-racial/Native American/Other non-Hispanic	−0.343 *	0.182	−0.103	0.146	−0.031	0.146
Black non-Hispanic	−1.040 ****	0.175	−0.489 ****	0.138	−0.481 ****	0.139
Hispanic/Latino	−0.361 **	0.164	−0.112	0.130	−0.058	0.130
Education (Ref = Less than Bachelor’s Degree)
Completed Bachelor’s Degree	0.060	0.132	−0.175	0.106	−0.179 *	0.106
Completed Graduate Degree	−0.456 *	0.255	−0.734 ****	0.209	−0.629 ***	0.212
Sexual Identity (Ref = Bisexual)
Gay	0.074	0.153	0.113	0.125	−0.051	0.125
Baseline Income (Ref = Above NYC Poverty Line)
Below NYC Poverty Line	0.217 *	0.165	0.103	0.102	0.296 ***	0.102
Visit (Relative to Baseline)
Visit 1	−0.092 ***	0.035	0.145 **	0.057	0.167 ***	0.064
Visit 2	−0.030	0.034	−0.021	0.057	−0.023	0.066
Visit 3	−0.017	0.036	0.123 **	0.057	0.063	0.066
Visit 4	−0.048	0.036	0.221 ****	0.055	0.235 ****	0.065
Visit 5	−0.051	0.037	0.138 **	0.058	−0.007	0.066
Visit 6	−0.013	0.037	0.172 ***	0.057	0.077	0.065

Significance: * = p < 0.1, ** = p < 0.05, *** = p < 0.01, **** = p < 0.001.

.

For separate models of alcohol to intoxication days (ATI), club drug days, and poly club drug days (at least one day of club drug use and at least one day using another illicit substance), perceived stress was the only stressor associated with an increase in the average days of each outcome. Discrimination was associated with a decrease in club drug days (β = −0.483; 95% CI −0.655, −0.311; p < 0.001) and poly club drug days (β = −0.526; 95% CI −0.725, −0.327; p < 0.001). For all outcomes, Black non-Hispanic was associated with the lowest average days of substance use. Club drug use did not significantly change over time, but poly club drug use showed a significant decrease at each visit (p < 0.001). These results are presented in

Table 3. Zero-Inflated Poisson Regression Model Results of Substance Use Outcomes, Stressors, and Baseline Covariates.

	Alcohol to Intoxication		Club Drug		Poly Club Drug
Predictor	Estimate	s.e.	Estimate	s.e.	Estimate	s.e.
Stigma	0.020	0.013	0.023	0.025	0.060 *	0.031
Discrimination	−0.312 ****	0.054	−0.483 ****	0.087	−0.526 ****	0.101
Internalized Homophobia	−0.006	0.008	−0.034 **	0.016	−0.053 ***	0.018
Perceived Stress	0.021 ****	0.002	0.012 ****	0.003	0.009 ***	0.003
Race/Ethnicity (Ref = White/non-Hispanic)
Asian/Multi-racial/Native American/Other non-Hispanic	−0.629 ****	0.148	−0.221	0.310	−0.024	0.227
Black non-Hispanic	−0.654 ****	0.138	−1.500 ****	0.307	−0.493 **	0.247
Hispanic/Latino	−0.371 ***	0.130	−0.129	0.274	−0.492 **	0.198
Education (Ref = Less than Bachelor’s Degree)
Completed Bachelor’s Degree	−0.121	0.106	−0.116	0.228	−0.494 ***	0.172
Completed Graduate Degree	−0.229	0.212	−0.726	0.457	−1.020 ***	0.357
Sexual Identity (Ref = Bisexual)
Gay	0.210 *	0.126	−0.359	0.261	−0.054	0.195
Baseline Income (Ref = Above NYC Poverty Line)
Below NYC Poverty Line	−0.371 ****	0.102	−0.373	0.218	0.016	0.162
Visit (Relative to Baseline)
Visit 2	−0.143 *	0.086	−0.023	0.070	−0.273 ****	0.078
Visit 4	−0.121	0.089	−0.067	0.072	−0.288 ****	0.079
Visit 6	−0.142	0.091	−0.050	0.075	−0.339 ****	0.082

Significance: * = p < 0.1, ** = p < 0.05, *** = p < 0.01, **** = p < 0.001.

.

The final aim (

Table 4. Zero-Inflated Poisson Regression Model Results of Substance Use outcomes and Mental Health Predictors, controlling for race/ethnicity, income, education.

	Alcohol to Intoxication		Club Drug		Poly Club Drug
	Estimate	s.e.	Estimate	s.e.	Estimate	s.e.
Anxiety	0.003 *	0.001	−0.001	0.002	0.001	0.002
Depression	0.003 *	0.002	0.009 ****	0.002	0.011 ****	0.002
PTSD	0.001	0.002	0.005 **	0.002	0.008 ****	0.002

Significance: * = p < 0.1, ** = p < 0.05, **** = p < 0.001.

) was to study if days of ATI, club drug days, and poly club drug days separately were predicted by anxiety, depression, and PTSD severity, after controlling for education, race/ethnicity, and baseline income. For ATI days, depression (β = 0.003; 95% CI −5.0 × 10⁻⁰⁵, 0.006; p < 0.1) and anxiety (β = 0.003; 95% CI −0.0004, 0.006; p < 0.1) were significant predictors. For club drug days, depression (β = 0.009; 95% CI 0.004, 0.013; p < 0.001) and PTSD (β = 0.005; 95% CI 0.003, 0.009; p < 0.05) were significant predictors. Likewise, depression (β = 0.011; 95% CI 0.006, 0.015; p < 0.001) and PTSD (β = 0.008; 95% CI 0.004, 0.012; p < 0.001) also predicted poly club drug days. Anxiety was not a significant predictor of club drug days or poly club drug days.

4. Discussion

4.1. Mental Health Outcomes

All stressors except stigma were associated with each mental health outcome, which is expected as stress is one of the most significant predictors of mental health disorders (Jackson et al., 2001; Kroenke et al., 1997) due to the physiologic effect of chronic stress on brain function and individual allostatic load (McEwen, 2007). Stress is highly associated with anxiety (Konstantopoulou et al., 2020), depression (Slavich & Irwin, 2014), and PTSD (Shalev, 2009; Y. Sun et al., 2021).

In our analyses, stigma only significantly predicted PTSD severity. This finding is still important as stigma is a major barrier for seeking mental health treatment (Cascalheira & Smith, 2018). Internalized homophobia was positively associated with each mental health outcome. This is found in the literature, even using other internalized homophobia scales (Igartua et al., 2003). Negative thoughts about identity are known to lead to chronic psychological distress and decreased self-esteem (Igartua et al., 2003).

Our findings indicate the differences of the magnitude in which perceived stress, stigma, discrimination, and internalized homophobia contribute to the severity of anxiety, depression, and PTSD over time in a young, diverse population. Discrimination had the largest impact on the severity of all mental health outcomes. This may be because of the external nature of discrimination, which is less controllable by the individual compared to the other more internal-facing stressors (Patterson et al., 1990), providing fewer avenues for the individual to cope with or influence the impact of discrimination. Future research should examine the impact of the inclusion of additional distal stressors, as these chronic stressors may provide additional insights into long-term mental health outcomes.

Black non-Hispanic individuals reported less severe mental health symptomatology relative to white non-Hispanic individuals, consistent with past findings (Schraedley et al., 1999). Our specific findings may not be generalizable outside NYC, as the impact of race and ethnicity on mental health is dependent on the social environment an individual lives in (Wight et al., 2005). Lower severity of each mental health condition among Black non-Hispanic individuals may be due to strong familial and community social support (Watt & Sharp, 2002) and strong relationships and social support outside of the family context (Choi, 2002). Alternatively, this may be due to a stigma around mental health disclosure which is observed among Black youth (Fields-Oriogun et al., 2024).

Higher education attainment was associated with less severe anxiety, depression, and PTSD. Those who were below NYC’s poverty line on average had higher severity of each mental health condition, and the impact was highest in PTSD. There is evidence that both factors are bidirectional, such that those with more severe mental health conditions have less educational attainment and lower income and that decreased income contributes to the development of mental health conditions (Niemeyer et al., 2019; Ridley et al., 2020).

4.2. Substance Use Outcomes

We observe that stigma did not significantly impact alcohol to intoxication days but was significantly associated with a slight increase in poly club drug days. Future longitudinal studies should also assess the impact on alcohol use (not only alcohol to intoxication), as this result may be due to the difference in substance social acceptability.

Discrimination and internalized homophobia predicted a decrease in the number of days of alcohol to intoxication, club drug use, and poly club drug use. While the factors that lead to substance use are multi-faceted, the social aspect is a large component. There is evidence that stigma directly contributes to social avoidance (Corrigan & Watson, 2002; Manago, 2015). Among gay men, there is evidence that methamphetamine use is predicted by the level of engagement with the gay community (Isaiah Green & Halkitis, 2006; Martino et al., 2021; Prestage et al., 2007). Methamphetamine use among gay men is greater on the weekend (Halkitis et al., 2005b; Martino et al., 2021), indicative of the social nature of club drug use. Our findings may indicate that stigma, discrimination, and internalized homophobia lead to decreased social engagement and specifically decreased exposure to the club drug environment as a result of social withdrawal and, therefore, decreased club drug use. This is not to suggest that discrimination and internalized homophobia are protective factors, but they may emphasize the social nature of club drug use. Perceived stress was associated with an increase in alcohol to intoxication, club drug, and poly club drug days, in line with past studies which find that substance use is used as a coping mechanism for perceived stress (Kornely & Halfmann, 2020; Tavolacci et al., 2013).

For club and poly club drug days, Black non-Hispanic men had on average the fewest days of use. This is consistent with past studies, which found white men had the highest prevalence of any club drug use (Kelly et al., 2006) but also higher prevalence for each individual club drug. These findings may be due to white non-Hispanic participants on average having higher socioeconomic status and therefore more access to and ability to purchase club drugs. Alternatively, white non-Hispanic SMM may feel more welcomed in the club drug scene in general. Over the course of this longitudinal study, club and poly club drug use increases. The literature shows that club drug use increases throughout young adulthood and only begins to decrease once the individual emerges from young adulthood (Van Havere et al., 2009).

Depression predicted more days of alcohol to intoxication. Depression and PTSD both predicted more days of club drug and poly club drug use. Among club drug users, there is an association between club drug use and depression. More severe depression leads to more days of club drug and poly club drug use. Interestingly, there is emerging evidence of the ability for club drugs such as MDMA to be used by individuals with PTSD to decrease symptom severity (Marseille et al., 2022; Mitchell et al., 2021). Individuals with PTSD symptomatology may use different substances to relieve different PTSD symptoms. For example, some may choose to use club drugs such as amphetamine as a way to specifically relieve avoidance symptoms due to their nature as a stimulant (Basedow et al., 2021).

Our study benefits from a more nuanced measure of poly substance use compared to the existing literature, capturing days of use within the past 30 days rather than relying on past-month, past-12-months, or even lifetime prevalence of poly club drug use. Our approach provides a more sensitive assessment of substance use patterns over time, allowing for increased ability to find potential associations with psychosocial stressors. Unlike past-month or past-12-month prevalence measures, day-level frequency captures both the intensity of use as well as the continuity of substance use. These factors are crucial for understanding risk trajectories for substance use among sexual minority men.

4.3. Sociodemographic Predictors

Race/ethnicity had a significant relation to mental health outcomes, likely due to the interaction between race and socioeconomic status. However, one study found that, when parental education was included in a model, the importance of race/ethnicity for mental health significantly decreased (Kennard et al., 2006). Future studies should aim to include parental education in the model. Only baseline income was included in the models, but income may vary widely each year, and many individuals in this age group may receive financial support from families. Other dimensions of socioeconomic status should be explored.

The P18 data presented several challenges; mental health and substance use variables were skewed towards zero, a common issue with risk behaviors (Boulton & Williford, 2018; Counsell et al., 2011). Small sample size for those who identified as Asian, Native American, multi-racial, and other non-Hispanic limited the interpretability of the impact of race and ethnicity as these participants were grouped into a single “Asian/Multi-racial/Native American/Other non-Hispanic” category at baseline in the primary dataset.

Research Implications

The findings from our study have several important policy implications. Addressing mental health disparities among SMM requires tailored interventions that account for unique stressors. Policies aimed at reducing stigma and promoting mental health can help mitigate the negative impact of stressors and encourage help-seeking behaviors. Integrating mental health education into school curricula and workplace wellness programs can foster supportive environments and destigmatize existing mental health challenges.

Social support was not included in the analysis, which may partially mediate perceived stress and PTSD (Catabay et al., 2019). Policies that promote social support networks and community engagement can serve as protective factors against substance use disorders (Martino et al., 2021).

Therefore, we may overestimate the impact of various psychosocial stressors on mental health outcomes. Measures of resilience were not included in this dissertation and should be examined in future studies.

4.4. Limitations

For sample size considerations, we had one subgroup entitled “Asian/Multi-racial/Native American/Other non-Hispanic”; future results would be strengthened if these subgroups could be looked and individually and in greater detail to illuminate different patterns of stressors, mental health symptomatology, and substance use. Results may not generalize to other SMM despite the diverse sample present because this is a young urban sample in a specific social environment. Young adults are more likely to use club drugs in NYC than elsewhere. The social environment is crucial in driving club drug use (Johnston et al., 2008). Sampling may decrease generalizability because individuals were recruited to the study through social networks and other in-person venues. Socially isolated individuals who are at increased risk of poor mental health outcomes may be underrepresented in the sample. Finally, Wave 2 recruitment began in 2014, and future studies should examine more recently recruited individuals to identify any changing trends.

5. Conclusions

Among a diverse sample of SMM, psychosocial stressors significantly predicted anxiety, depression, and PTSD after controlling for baseline characteristics. Discrimination had the greatest impact on poor mental health. These stressors significantly impact the number of days of alcohol to intoxication, club drug use, and poly club drug use, however, the specific stressor affected whether there was an increase or decrease in the number of days (i.e., stigma predicted an increase in club drug use, whereas internalized homophobia predicted a decrease in club drug use).

These findings extend the current literature by examining the relationship between numerous psychosocial stressors, mental health outcomes, and substance use outcomes in a diverse cohort of SMM over a period of several formative years during young adulthood. These findings reinforce the public health implications that discrimination and stress have on driving substance use and mental health disparities among vulnerable populations.

Future longitudinal studies should be conducted in a different social setting to determine how generalizable these findings are outside of the NYC social environment. Parental education and measures of social support should also be included in future analyses to help understand the mechanisms between psychosocial stressors, substance use, and mental health.

Further, while mental health problems and substance use are important and challenging health problems in themselves, if left untreated or undertreated they can cause a wide range of other health conditions. Overall, intervening to reduce all forms of psychosocial stressors and continuing to help individuals with the treatment of mental health conditions and substance use can have a tremendous impact on reducing the growing health disparities among SMM.