Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Aug 1.
Published in final edited form as: Prev Sci. 2023 Jun 16;24(6):1102–1114. doi: 10.1007/s11121-023-01561-x

Comparing factors associated with increased stimulant use in relation to HIV status using a machine learning and prediction modeling approach

Cheríe S Blair 1, Marjan Javanbakht 2, W Scott Comulada 3,4, Robert Bolan 5, Steven Shoptaw 6, Pamina M Gorbach 1,2, Jack Needleman 3
PMCID: PMC10795486  NIHMSID: NIHMS1950898  PMID: 37328629

Abstract

Stimulant use is an important driver of HIV/STI transmission among men who have sex with men (MSM). Evaluating factors associated with increased stimulant use is critical to inform HIV prevention programming efforts. This study seeks to use machine learning variable selection techniques to determine characteristics associated with increased stimulant use and whether these factors differ by HIV status. Data from a longitudinal cohort of predominantly Black/Latinx MSM in Los Angeles, California was used. Every 6 months from 8/2014–12/2020, participants underwent STI testing and completed surveys evaluating the following: demographics, substance use, sexual risk behaviors, and last partnership characteristics. Least absolute shrinkage and selection operator (lasso) was used to select variables and create predictive models for an interval increase in self-reported stimulant use across study visits. Mixed-effects logistic regression was then used to describe associations between selected variables and the same outcome. Models were also stratified based on HIV status to evaluate differences in predictors associated with increased stimulant use. Among 2,095 study visits from 467 MSM, increased stimulant use was reported at 20.9% (n=438) visits. Increased stimulant use was positively associated with unstable housing (adjusted [a]OR 1.81; 95% CI 1.27–2.57), STI diagnosis (1.59; 1.14–2.21), transactional sex (2.30; 1.60–3.30), and last partner stimulant use (2.21; 1.62–3.00). Among MSM living with HIV, increased stimulant use was associated with binge drinking, vaping/cigarette use (aOR 1.99; 95% CI 1.36–2.92), and regular use of poppers (2.28; 1.38–3.76). Among HIV-negative MSM, increased stimulant use was associated with participating in group sex while intoxicated (aOR 1.81; 95% CI 1.04–3.18), transactional sex (2.53; 1.40–2.55), and last partner injection drug use (1.96; 1.02–3.74). Our findings demonstrate that lasso can be a useful tool for variable selection and creation of predictive models. These results indicate that risk behaviors associated with increased stimulant use may differ based on HIV status and suggest that co-substance use and partnership contexts should be considered in the development of HIV prevention/treatment interventions.

Keywords: substance use, men who have sex with men, HIV, stimulants

INTRODUCTION

Stimulant use is substantially higher among men who have sex with men (MSM) compared to the U.S. general population (Centers for Disease Control and Prevention [CDC], 2019; Substance Abuse and Mental Health Services Administration [SAMHSA], 2019). Consumption of stimulants among MSM frequently occurs within social and sexual contexts, such as night clubs, bath houses, circuit parties, and sex clubs (Drumright et al., 2006; Giorgetti et al., 2017). Stimulant use, particularly methamphetamine, is a prominent driver of HIV transmission among MSM - thought to be driven by risky sexual behaviors with serodiscordant partners (Freeman et al., 2011; Ostrow et al., 2009). Despite significant advances in the field of biomedical HIV prevention, HIV transmission among MSM who use stimulants remains one of the predominant factors contributing to the ongoing HIV epidemic (Swartz & McCarty-Caplan, 2018). As such, understanding factors that are related to stimulant use and HIV risk is important to develop effective, targeted interventions for this key population.

Factors contributing to HIV transmission among MSM differ by HIV status. For HIV-negative individuals, HIV acquisition is predominantly through condomless receptive anal intercourse with sexual partners living with HIV (Baggaley et al., 2010). For MSM living with HIV (MWH), risk for transmitting HIV to sexual partners is driven by condomless anal intercourse (particularly insertive) and HIV viral load, with higher levels of HIV viremia associated with increased risk of HIV transmission (LeMessurier et al., 2018). Given the differences that exist regarding mechanisms of HIV transmission, it is important to understand stimulant use and risk behavior within the context of HIV status, given the impact that these distinctions may have on HIV transmission dynamics.

While stimulant use is linked to sexual risk behaviors, evidence suggests that stimulant use prevalence and patterns of use may differ according to HIV status (Centers for Disease Control and Prevention [CDC], 2019; Halkitis, Fischgrund, et al., 2005; Hood et al., 2018; Substance Abuse and Mental Health Services Administration [SAMHSA], 2019). MWH have a higher likelihood of using stimulants to avoid unpleasant emotions, social pressures, and conflict, compared to their HIV-negative counterparts (Halkitis, Green, et al., 2005). Among MWH, stimulants may be used as an avoidance coping strategy to deal with an HIV diagnosis (Halkitis et al., 2008). Stimulants may also be used to dispel potential anxieties or fears associated with sexual activity among MWH, resulting in MWH using stimulants for sexual reasons compared to HIV-negative MSM who tend to report using stimulants socially (Halkitis, Fischgrund, et al., 2005). These stimulant use patterns may influence sexual risk behaviors, as it has been suggested that stimulants may have a greater impact on frequency of condomless intercourse among MWH compared those who are HIV-negative (Halkitis et al., 2008). For example, longitudinal analysis of a cohort of MSM in the U.S. revealed that MSM who used stimulants and underwent HIV seroconversion had higher frequency of risky sexual behaviors and stimulant use than those who remained HIV-negative (Swartz & McCarty-Caplan, 2018).

These differences in stimulant use patterns highlight that the factors which are driving the ongoing stimulant and HIV epidemics among MSM are nuanced and complex. Determining whether the contexts that surround stimulant use patterns differ based on HIV status represents an important next step in understanding how stimulant use contributes to HIV transmission dynamics within the sexual networks of MSM and to the development of effective interventions. However, data explicitly evaluating differential factors contributing to ongoing stimulant use are limited, with most studies either including HIV status as a covariate or evaluating stimulant use within the context of a cohort comprised exclusively of MWH or HIV-negative MSM (Chartier et al., 2009; Colyer et al., 2020; Fletcher et al., 2020; Freeman et al., 2011). This study seeks to bridge this gap by evaluating differences in factors associated with increased stimulant use among a diverse cohort of MSM and examining how they differ according to HIV status. This analysis utilized machine learning techniques, specifically least absolute shrinkage and selection operator (lasso), to select variables and create predictive models to evaluate factors associated with increased stimulant use. Sub-analyses stratifying participants based on HIV status were conducted to evaluate differences in predictors associated with increased stimulant use. To the best of our knowledge, this will be one of the first studies to use machine learning techniques to select predictors associated with increased stimulant use and to compare differences according to HIV status. Findings from this analysis will provide important information on whether contexts surrounding increases in stimulant use patterns differ according to HIV status, which can be used to inform future HIV prevention programming efforts.

METHODS

Data Source and Study Procedures

Data for this analysis came from the Men Who Have Sex with Man and Substance Use Cohort at UCLA Linking Infections, Noting Effects (mSTUDY; U01 DA036267), an ongoing longitudinal cohort designed to evaluate the impact of substance use on HIV transmission. The cohort consists of a group of racially/ethnically diverse MSM who are living with or are at high-risk for HIV. Methods have been previously described (Aralis et al., 2018). Briefly, the cohort consists of predominantly Black and Latinx MSM, half with active substance use at enrollment. Participants were recruited to include half MWH and half HIV-negative MSM by design. HIV-negative MSM were recruited from a community-based university research clinic, and MWH were recruited from a community-based organization that provides clinical and community resources for the lesbian, gay, and transgender community in Los Angeles. Inclusion criteria for the cohort were: 1) 18–45 years old at the time of study enrollment, 2) born male, 3) condomless anal intercourse with a man in past 6 months (if HIV-negative). Study enrollment began in August 2014, and recruitment is ongoing to replace loss to follow-up. To date, 577 MSM have been enrolled. This analysis consists of visits that occurred from August 2014 (study inception) to December 2020 where participants provided self-reported data on stimulant use in last 6 months.

Study visits occurred every 6 months. At each visit, participants underwent STI testing, clinician interview, and completed a computer-assisted self-interview survey that collected information on the following: demographics, substance use, mental health, and sexual behaviors. Rectal and pharyngeal swabs as well as urine samples were collected at each visit and tested for gonorrhea/chlamydia (GC/CT) with nucleic acid amplification testing (Aptima Combo 2, GenProbe, San Diego, CA). Blood samples were collected for syphilis testing using rapid plasma reagin (RPR) with confirmatory testing via the Treponema pallidum particle agglutination test (TPPA). Infectious syphilis (i.e., primary, secondary, or early latent) was defined following positive test results through confirmation from the local health department and using the Centers for Disease Control and Prevention determination (Workowski & Bolan, 2015). Study personnel assisted with notifying participants of their STI testing results and facilitated linkages to care for positive results. The study was reviewed and approved by the Office of Human Research Participant Protection (OHRPP) at the University of California, Los Angeles.

Statistical Analysis

Measures

The purpose of this analysis was to utilize machine learning for variable selection and prediction to determine factors associated with increased stimulant use. We also sought to evaluate if predictors associated with increased stimulant use differed by HIV status. Our increased stimulant use outcome variable was constructed from the question “In the last 6 months, how often did you use [drug]?”. Possible stimulants included methamphetamine, cocaine powder, and crack cocaine. Response options included: “Daily”, “Weekly”, “Monthly”, “Less often than monthly”, “Once”, and “Never”. Methamphetamine, cocaine powder, and crack cocaine were combined into one composite “stimulants” variable. We chose to create a composite stimulants variable given the well-established link of stimulant use with sexual risk behavior and HIV transmission (Gamarel et al., 2015; Hojilla et al., 2018; Swartz & McCarty-Caplan, 2018). A lag variable was created, indicating whether there was an increase in reported stimulant use compared to the prior visit (e.g., reporting using stimulants “once” at Tn followed by “weekly” use at Tn+1), which was binary (no increased stimulant use or yes increased stimulant use).

Predictors for inclusion in lasso models were selected based on whether the variable had a conceptually relevant relationship with stimulant use based on the literature (Supplemental Tables 1 and 2). Variables included participant demographics as well as the following constructs: housing instability (Glick et al., 2018), history of incarceration (Anderson-Carpenter et al., 2017), intimate partner violence (Wu et al., 2014), depression (Javanbakht et al., 2020), substance use (Patterson et al., 2005), sexual behaviors (Semple et al., 2010), and last sexual partner characteristics (Wray & Monti, 2019).

For substance use variables, participants were asked the question “In the last 6 months, how often did you use [drug]?”. Potential drug options included: fentanyl, heroin, prescription opiates, marijuana, and poppers. Response options included: “Daily”, “Weekly”, “Monthly”, “Less often than monthly”, “Once”, and “Never”. Given relatively low prevalence of reported opiate use and overlap between types of opiates used in this sample, fentanyl, heroin, and prescription opiates were combined into one composite opiates variable. Regular drug use was a binary variable defined as reporting daily or weekly use in the past 6 months (e.g., for regular opiate use, yes = reporting “weekly” or “daily” opiate use in past 6 months; no = reporting “monthly”, “less often than monthly”, “once”, or “never” in past 6 months). Transactional sex was defined as the participant giving or receiving money, drugs, and/or a place to stay in exchange for anal sex in the past 3 months.

Creation of Lasso Models

Lasso models were created for variable selection and development of predictive models. Lasso regression selects predictors by fitting models using all possible predictors and shrinking the regression coefficients of predictors that do not sufficiently contribute to error minimization to zero, thus eliminating them from the model (Tibshirani, 2011). Lasso was selected over traditional statistical models because the regularization methods used in lasso promote sparse models that minimize standard errors and improve interpretability of models (Tibshirani, 1996). In comparison to lasso, traditional statistical models are prone to overfitting, resulting in models with low bias but high variance, which may result in inaccurate predictions (Hastie et al., 2015). The dataset was randomly split 50/50 into a testing and training dataset (Stata Corp LLC., 2019). All potential variables (Supplemental Table 1) were included in initial lasso models. All models controlled for age, race/ethnicity, and HIV status. Models were fit on the training dataset using ten-fold cross-validation and ordered based on the magnitude of the tuning parameter (λ), i.e., a parameter to control the degree to which regression coefficients are shrunk towards zero to obtain suitable model fit (Hastie et al., 2009). The model with the value of λ that minimized the out-of-sample prediction error was identified and cross-validation plots were created to ensure that λ was minimized. Goodness of fit (GOF) and model performance were evaluated over a grid of λ values within one standard error (SE) of the minimal value of λ. Models were evaluated using the testing and training datasets by 1) deviance and deviance ratios, 2) area under the receiver operating curve (AUC) using mixed-effects logistic regression models, and 3) Matthews Correlation Coefficient (MCC). The lasso model within 1 SE of the minimal value of λ with the lowest deviance and deviance ratio (Hastie et al., 2015), highest AUC (Lasko et al., 2005), MCC closest to an absolute value of 1 (Chicco & Jurman, 2020), and had the most consistent indices between the training and testing datasets was selected as the final lasso model. To evaluate whether predictors associated with increased stimulant use differed by HIV status, the dataset was also stratified by HIV status and lasso models were created as above.

For the final models selected, descriptive statistics (frequency, percentage, median, interquartile range [IQR]) of the predictors selected by lasso models were calculated. Chi-square and Kruskal-Wallis tests were used to evaluate whether the distribution of predictors differed based on increased stimulant use. Mixed-effects logistic regression analyses were used to calculate unadjusted and adjusted odds ratios using variables selected from the final lasso model, using increased stimulant use as the outcome variable. Complete case analysis was used (n=2,095/2,676 visits), and all analyses were conducted using Stata 16.1 (StataCorp, College Town, TX).

RESULTS

Entire Cohort

The sample consisted of 2,095 visits across 467 participants. Increased stimulant use was reported at 20.9% (n=438) of visits (Table 1). Median age was 33 years (IQR 28–40; range 18–50) and 53.8% (n=1,126) of visits were completed by MWH. Almost half of study visits were completed by Latinx participants (50.1% of study visits; n=1,049), followed by Black (38.9%; n=814), White (6.5%; n=136), and other racial/ethnic groups (4.6%; n=96). Participants who reported increased stimulant use reported a higher prevalence of unstable housing, unemployment, cannabis use and binge drinking, regular opiate use, and last partner substance use (specifically, stimulants and ecstasy), compared to those without increased stimulant use. Positive STI testing occurred more frequently during visits with increased stimulant use (23.7%; n=104/334) compared to visits without increased stimulant use (13.8%; n=228/1,429).

Table 1:

Participant characteristics, substance use, sexual risk behavior, and last partner characteristics, stratified by whether increased stimulant use was reported at mSTUDY visits 8/2014–12/2020 (N=2,095 visits)

No Increased Stimulant Use (n=1,657) Increased Stimulant Use (n=438) p-value
n (%) n (%)
Age (median, IQR) 33 (28–40) 33 (28–39) 0.76
Race/Ethnicity
 White 104 (6.3%) 32 (7.3%) 0.13
 Black 665 (40.1%) 149 (34.0%)
 Latinx 812 (49.0%) 237 (54.1%)
 Other 76 (4.6%) 20 (4.6%)
HIV
 Negative 791 (47.7%) 178 (40.6%) 0.008
 Living with HIV 866 (52.3%) 260 (59.4%)
Employment status a
 Employed 1,327 (80.1%) 311 (71.0%) <0.001
 Unemployed 330 (19.9%) 127 (29.0%)
Unstable housing a
 No 1,454 (87.7%) 326 (74.4%) <0.001
 Yes 203 (12.3%) 112 (25.6%)
Substance Use
Binge drinking a
 Never 965 (58.2%) 207 (47.3%) <0.001
 Monthly or less 540 (32.6%) 166 (37.9%)
 Weekly/daily 152 (9.2%) 65 (14.8%)
Vaping/Cigarette use a
 No 1,226 (74.0%) 247 (56.4%) <0.001
 Yes 431 (26.0%) 191 (43.6%)
Cannabis use a
 No 890 (53.7%) 157 (35.8%) <0.001
 Weekly or less frequent 434 (26.2%) 144 (32.9%)
 Daily 333 (20.1%) 137 (31.3%)
Regular opiate use a
 No 1,606 (96.9%) 416 (95.0%) 0.048
 Yes 51 (3.1%) 22 (5.0%)
Sexual Risk Behavior
Sexually transmitted infection
 No 1,429 (86.2%) 334 (76.3%) <0.001
 Yes 228 (13.8%) 104 (23.7%)
Anal intercourse while intoxicated a
 No 915 (55.2%) 132 (30.1%) <0.001
 Yes 742 (44.8%) 306 (69.9%)
Transactional sex a
 No 1,484 (89.6%) 311 (71.0%) <0.001
 Yes 173 (10.4%) 127 (29.0%)
Last Partner Characteristics
Last partner was an unknown person
 No 1,502 (90.6%) 376 (85.8%) 0.003
 Yes 155 (9.4%) 62 (14.2%)
Last partner used stimulants
 No 1,292 (78.0%) 225 (51.4%) <0.001
 Yes 365 (22.0%) 213 (48.6%)
Last partner used ecstasy
 No 1,591 (96.0%) 398 (90.9%) <0.001
 Yes 66 (4.0%) 40 (9.1%)
Open in a new tab
a

Last 6 months; IQR = Interquartile range

Lasso models selected predictors associated with constructs surrounding financial insecurity, substance use, sexual risk behaviors, and last partnership characteristics (AUC=0.75). In adjusted analysis (Table 2), transactional sex associated the most highly with increased stimulant use, with MSM who reported transactional sex having over twice the odds of reporting increased stimulant use (aOR 2.30; 95% CI 1.60–3.30) compared to those who did not report transactional sex. Having a last partner who used stimulants had higher odds of increased stimulant use (aOR 2.21; 95% CI 1.62–3.00) compared to participants whose last partner did not use stimulants. Positive STI testing was associated with 1.59 times higher odds (95% CI 1.14–2.21) of increased stimulant use, compared to negative STI testing. Increased stimulant use was also positively associated with unstable housing, vaping/cigarette use, cannabis use, regular opiate use, anal intercourse while intoxicated, and having a last partner who was anonymous, compared to participants who did not report those behaviors.

Table 2:

Unadjusted and adjusted odds ratios of factors associated with increased reported stimulant use at mSTUDY visits (N=2,095)

OR (95% CI) p-value aOR (95% CI) p-value
Age 1.00 (0.98–1.02) 0.97 1.01 (0.98–1.03) 0.68
Race/Ethnicity
 White Ref -- Ref --
 Black 0.77 (0.40–1.50) 0.44 1.08 (0.57–2.06) 0.80
 Latinx 1.11 (0.58–2.12) 0.76 1.55 (0.82–2.91) 0.17
 Other 0.87 (0.35–2.11) 0.75 1.30 (0.55–3.05) 0.55
HIV
 Negative Ref -- Ref --
 Living with HIV 1.44 (1.03–2.02) 0.035 1.30 (0.93–1.81) 0.12
Employment status a
 Employed Ref -- Ref --
 Unemployed 1.49 (1.09–2.02) 0.011 1.13 (0.82–1.55) 0.45
Unstable housing a
 No Ref -- Ref --
 Yes 2.74 (1.94–3.86) <0.001 1.81 (1.27–2.57) 0.001
Substance Use
Binge drinking a
 Never Ref -- Ref --
 Monthly or less 1.55 (1.15–2.08) 0.004 1.59 (1.17–2.15) 0.003
 Weekly/daily 2.04 (1.30–3.22) 0.002 1.76 (1.13–2.75) 0.013
Vaping/Cigarette use a
 No Ref -- Ref --
 Yes 2.29 (1.69–3.10) <0.001 1.69 (1.24–2.31) 0.001
Cannabis use a
 No Ref -- Ref --
 Weekly or less frequent 2.22 (1.59–3.08) <0.001 1.82 (1.31–2.54) <0.001
 Daily 2.87 (1.98–4.16) <0.001 2.24 (1.55–3.25) <0.001
Regular opiate use a
 No Ref -- Ref --
 Yes 2.07 (1.06–4.04) 0.032 1.00 (0.50–2.01) 1.00
Sexual Risk Behavior
Sexually transmitted infection
 No Ref -- Ref --
 Yes 2.09 (1.51–2.88) <0.001 1.59 (1.14–2.21) 0.006
Anal intercourse while intoxicated a
 No Ref -- Ref --
 Yes 3.01 (2.27–3.98) <0.001 1.64 (1.21–2.20) 0.001
Transactional sex a
 No Ref -- Ref --
 Yes 4.17 (2.97–5.85) <0.001 2.30 (1.60–3.30) <0.001
Last Partner Characteristics
Last partner was unknown person
 No Ref -- Ref --
 Yes 1.89 (1.27–2.83) 0.002 1.54 (1.03–2.32) 0.037
Last partner used stimulants
 No Ref -- Ref --
 Yes 3.55 (2.69–4.69) <0.001 2.21 (1.62–3.00) <0.001
Last partner used ecstasy
 No Ref -- Ref --
 Yes 2.45 (1.43–4.20) 0.001 1.25 (0.71–2.18) 0.44
Open in a new tab
a

Last 6 months

Note: Bold indicates p-value <0.05

Stratified by HIV Status

Among MWH, the sample consisted of 1,199 study visits across 242 participants. Increased stimulant use was reported at 22.9% (n=274) of visits, and median age was 36 years (IQR 31–41) (Supplemental Table 3). Compared to lasso models containing the entire cohort and those restricted to HIV-negative participants, lasso tended to select constructs surrounding polysubstance use among models restricted to MWH (AUC=0.71). In adjusted analyses (Table 3), increased stimulant use was positively associated with unstable housing (aOR 2.25; 95% CI 1.45–3.51), vaping/cigarette use (1.99; 1.36–2.92), using poppers regularly (2.28 1.38–3.76), and transactional sex (2.33; 1.48–3.65) and was negatively associated with reporting that one’s last partner was a regular/main partner (0.70; 0.49–0.99), compared to MWH who did not report those characteristics/behaviors.

Table 3:

Unadjusted and adjusted odds ratios of factors associated with increased reported stimulant use at mSTUDY visits for participants living with HIV (N=1,199 visits)

OR (95% CI) p-value aOR (95% CI) p-value
Age 0.97 (0.94–1.00) 0.07 0.99 (0.96–1.02) 0.34
Race/Ethnicity
 White Ref -- Ref --
 Black 0.78 (0.33–1.82) 0.56 1.28 (0.56–2.91) 0.56
 Latinx 1.02 (0.45–2.32) 0.96 1.54 (0.69–3.43) 0.29
 Other 0.54 (0.19–1.50) 0.23 0.96 (0.35–2.63) 0.93
Unstable housing a
 No Ref -- Ref --
 Yes 2.91 (1.88–4.48) <0.001 2.25 (1.45–3.51) <0.001
Substance Use
Binge drinking a
 Never Ref -- Ref --
 Monthly or less 1.70 (1.16–2.48) 0.006 1.63 (1.12–2.38) 0.011
 Weekly/daily 2.02 (1.05–3.88) 0.035 1.87 (1.00–3.51) 0.05
Vaping/Cigarette use a
 No Ref -- Ref --
 Yes 2.06 (1.41–3.00) <0.001 1.99 (1.36–2.92) <0.001
Regular poppers use a
 No Ref -- Ref --
 Yes 3.29 (2.01–5.40) <0.001 2.28 (1.38–3.76) 0.001
Regular opiate use a
 No Ref -- Ref --
 Yes 1.40 (0.55–3.59) 0.48 0.80 (0.29–2.20) 0.67
Sexual Risk Behavior
Having a regular partner a
 No Ref -- Ref --
 Yes 1.74 (1.20–2.51) 0.003 1.38 (0.95–2.02) 0.09
Transactional sex a
 No Ref -- Ref --
 Yes 3.28 (2.13–5.05) <0.001 2.33 (1.48–3.65) <0.001
Last Partner Characteristics
Last partner was regular/main partner
 No Ref -- Ref --
 Yes 0.71 (0.50–1.00) 0.05 0.70 (0.49–0.99) 0.043
Last partner used ecstasy
 No Ref -- Ref --
 Yes 1.52 (0.73–3.19) 0.27 1.28 (0.60–2.75) 0.52
Open in a new tab
a

Last 6 months

Note: Bold indicates p-value <0.05

Among HIV-negative participants, the sample consisted of 912 study visits across 228 participants. Increased stimulant use was reported at 18.2% (n=166) of study visits and median age was 30 years (IQR 26–36). HIV-negative participants with increased stimulant use reported lower income and education levels as well as higher frequency of unemployment, unstable housing, cannabis use, sexual risk behaviors, and last partner substance use (e.g., alcohol, poppers, and injection drug use) compared to those who did not report increased stimulant use (Supplemental Table 4). Compared to lasso models containing the entire cohort as well as those restricted to MWH, lasso models including HIV-negative participants tended to include more sexual risk behaviors and last partner substance use (AUC=0.76). In adjusted analysis (Table 4), increased stimulant use was positively associated with unstable housing (aOR 1.94; 95% CI 1.09–3.45), cannabis use, transactional sex (2.53; 1.40–4.55), group sex while intoxicated (1.81; 1.04–3.18), and having a last sexual partner who injected drugs (1.96; 1.02–3.74) and was negatively associated with higher levels of education, compared to HIV-negative participants who did not endorse those characteristics/behaviors.

Table 4:

Unadjusted and adjusted odds ratios of factors associated with increased reported stimulant use at mSTUDY visits for HIV-negative participants (N=912 visits)

OR (95% CI) p-value aOR (95% CI) p-value
Age 1.00 (0.96–1.04) 0.93 1.01 (0.97–1.05) 0.73
Race/Ethnicity
 White Ref -- Ref --
 Black 0.98 (0.30–3.20) 0.97 1.01 (0.34–3.00) 0.99
 Latinx 1.86 (0.58–5.94) 0.29 1.54 (0.53–4.49) 0.43
 Other 1.11 (0.23–5.37) 0.90 1.05 (0.24–4.56) 0.94
Education
 Less than high school Ref -- Ref --
 High school 0.30 (0.13–0.71) 0.006 0.40 (0.17–0.91) 0.03
 More than high school 0.33 (0.15–0.75) 0.008 0.41 (0.19–0.92) 0.03
Annual income
 Less than $10,000 Ref -- Ref --
 $10,000-$30,000 0.71 (0.43–1.16) 0.17 0.90 (0.54–1.50) 0.68
 More than $30,000 0.53 (0.29–0.99) 0.047 0.64 (0.33–1.23) 0.18
Employment status a
 Employed Ref -- Ref --
 Unemployed 1.78 (1.08–2.94) 0.024 1.40 (0.81–2.42) 0.23
Unstable housing a
 No Ref -- Ref --
 Yes 3.06 (1.78–5.27) <0.001 1.94 (1.09–3.45) 0.025
Substance Use
Cannabis use a
 No Ref -- Ref --
 Weekly or less frequent 2.25 (1.32–3.82) 0.003 2.15 (1.26–3.66) 0.005
 Daily 2.73 (1.51–4.96) 0.001 2.00 (1.11–3.59) 0.02
Sexual Risk Behavior
Sexual partner concurrency a
 No Ref -- Ref --
 Yes 1.70 (1.11–2.62) 0.016 1.18 (0.75–1.87) 0.47
Anal intercourse while intoxicated a
 No Ref -- Ref --
 Yes 2.54 (1.60–4.02) <0.001 1.59 (0.96–2.62) 0.07
Group sex while intoxicated a
 No Ref -- Ref --
 Yes 3.73 (2.26–6.15) <0.001 1.81 (1.04–3.18) 0.037
Transactional sex a
 No Ref -- Ref --
 Yes 5.02 (2.88–8.76) <0.001 2.53 (1.40–4.55) 0.002
Last Partner Characteristics
Last partner used alcohol
 No Ref -- Ref --
 Yes 1.77 (1.09–2.88) 0.021 1.31 (0.81–2.13) 0.28
Last partner used poppers
 No Ref -- Ref --
 Yes 1.70 (1.00–2.91) 0.05 1.31 (0.77–2.23) 0.33
Last partner injected drugs
 No Ref -- Ref --
 Yes 2.90 (1.54–5.47) 0.001 1.96 (1.02–3.74) 0.043
Open in a new tab
a

Last 6 months

Note: Bold indicates p-value <0.05

DISCUSSION

In this analysis of a diverse cohort of MSM in Los Angeles, California, increased stimulant use was positively associated with unstable housing, transactional sex, polysubstance use, STIs, and sexual risk behavior. However, constructs correlated with increased stimulant use in lasso models differed when the sample was stratified by HIV status. Among MWH, polysubstance use was highly correlated with increased stimulant use. However, among HIV-negative participants, last partnership characteristics and sexual risk behaviors were correlated with increased stimulant use. This analysis is among the first to utilize lasso for variable selection and creation of predictive models to evaluate factors that are associated with increased stimulant use. Our approach demonstrates that machine learning techniques can be a useful and efficient tool to assist with selecting relevant predictors from datasets with large amounts of potential variables to create conceptually relevant models. Using lasso for variable selection allowed us to evaluate differences in predictors that were associated with increased stimulant use according to HIV status. These findings provide an important next step in understanding the disproportionate effect that the complicated stimulant use epidemic has on certain MSM subpopulations and could be used to inform future HIV prevention interventions.

Unstable housing and transactional sex were consistently selected across all models, suggesting that these variables were highly correlated with increased stimulant use regardless of HIV status. Compared to non-sexual minorities, sexual minorities disproportionately experience unstable housing, often due to homophobia, rejection, and abuse that forces them from their homes (Baams et al., 2019; Romero et al., 2020). Furthermore, stimulants are often used as form of coping with stressful feelings associated with being unstably housed as well as a means of survival (Johnson & Chamberlain, 2008). For example, stimulants may be utilized to stay awake to protect belongings, facilitate social interaction with others, or as an alternative to psychiatric medications (Bungay et al., 2006). Unstably housed MSM may also use stimulants to obtain a sense of belonging, to bond with others, or due to perceived social norms (Barman-Adhikari et al., 2016; Johnson & Chamberlain, 2008). Additionally, transactional sex has a strong association with unstable housing and can be used as a mechanism to obtain financial support or shelter among unstably housed individuals (Mimiaga et al., 2009). Transactional sex may also occur for the purpose of obtaining drugs (Javanbakht et al., 2019). These findings underscore the importance of addressing the underlying factors that often drive the interdependent relationship between unstable housing, stimulant use, and transactional sex. Specifically, these results suggest the potential utility of interventions, such as contingency management, designed to reduce these barriers through linkages to financial or community resources in exchange for not using substances (Tracy et al., 2007).

In addition to socioeconomic disadvantage, having a partner who used stimulants and engaging in anal intercourse while intoxicated were positively associated with increased stimulant use. These findings highlight the unique social and sexual contexts where stimulants are frequently consumed by MSM, such as sex clubs, circuit parties, and bath houses (Drumright et al., 2006; Giorgetti et al., 2017; Reback et al., 2004). Stimulants are often used by MSM in sexual settings to obtain sexual partners, increase libido, augment sexual stamina, and for disinhibition (A Bourne et al., 2015; Weatherburn et al., 2017). However, sexualized stimulant use can impair decision making and lead to risk behaviors, such as increased number of casual partners and impaired condom negotiation (Berry et al., 2020; Hoenigl et al., 2016). Due to these contexts, sexualized stimulant use is independently associated with HIV/STIs and is an important driver of HIV/STI transmission within the sexual networks of MSM who use stimulants (Lai et al., 2020; Reback & Fletcher, 2018). As increased stimulant use was also associated with having an STI, a known risk factor for HIV transmission, these findings underscore the importance of coordinated public health efforts that incorporate treatment of comorbid stimulant use into HIV/STI treatment and prevention interventions.

Among MWH, polysubstance use (e.g., binge drinking, smoking, regular opiate use, and regular popper use) correlated highly with increased stimulant use. Polysubstance use among MWH may be used as a coping mechanism related to an HIV diagnosis, HIV-related stigma, or depressive symptoms (Earnshaw et al., 2020; Glynn et al., 2019). Substance use may also be used as an avoidant coping strategy to mitigate stress associated with being a sexual minority (Mereish et al., 2017). This minority stress may be exacerbated by the stigma of living with HIV, potentially resulting in increased substance use. This consideration is highlighted in a study by Jerome et. al., where MWH who used substances reported higher levels of distressing emotions related to an undesirable self-image and daily stressors than their HIV-negative counterparts (Jerome et al., 2009). The need for external validation due to a negative self-image may cause MWH to engage in substance use to feel more desirable and to form connections with others in both social and sexual contexts (Edelman et al., 2016). MWH may use substances for social inclusion among groups where substance use is socially accepted and due to fear that they may be excluded from these groups if they do not engage in substance use (W. Hawkins et al., 2019). Furthermore, the normalization of substance use within social circles may perpetuate continued substance use and serve as a potential barrier to reductions in consumption, particularly if the individual perceives that their social network would not be supportive of their desire to stop using substances (Edelman et al., 2016). Understanding these contexts and drivers of substance use among MWH is particularly important from a public health standpoint given the well-established connection between substance use and sexual risk behavior (Hegazi et al., 2017), which was further supported by our findings demonstrating that increased stimulant use was associated with transactional sex and having a last partner that was a non-primary partner. In addition to increased sexual risk behavior, substance use is also associated with antiretroviral therapy nonadherence, further reinforcing the contribution of substance use to ongoing HIV transmission within certain MSM subpopulations (Socias & Milloy, 2018). Collectively, these findings highlight the importance of interventions designed to improve peer support, reduce HIV-related stigma, and the value of integrating substance use treatment into the HIV care continuum.

In contrast to MWH, increased stimulant use correlated highly with socioeconomic status, sexual risk behaviors (e.g., having group sex or anal intercourse while intoxicated), and last partner substance use among HIV-negative participants. These findings are consistent with data demonstrating that stimulant use is highly prevalent within sexual contexts and associated with increased individual-level sexual risk behaviors among MSM who use stimulants (Loza et al., 2020; Shoptaw & Reback, 2007). However, our results suggest that partnership dynamics may influence stimulant use patterns or vice versa. It is possible that MSM who use stimulants may seek out partners with similar patterns of substance use or that their partners’ substance use may influence their own behaviors (Derrick et al., 2019; Shariati et al., 2017). Alternatively, HIV-negative MSM may tend to use stimulants within sexualized contexts where substance use is more common and where they are more likely to encounter partners who also engage in sexualized substance use, such as circuit parties, bath houses or sex clubs (Adam Bourne et al., 2015; Fulcher et al., 2019). Beyond individual-level risk behavior, substance use within sexual partnerships has been associated with sexual risk behaviors, such as condomless anal intercourse with serodiscordant partners (Brown et al., 2017). Furthermore, within stable partnerships, partnership-level substance use may influence couples’ sexual behavior and decision-making surrounding risk mitigation strategies, such as sexual agreements and whether those agreements are broken (Mitchell et al., 2014). These findings highlight the sexual contexts in which substances are used among HIV-negative MSM and further support the extant literature indicating that stimulant use likely plays a substantial role in HIV seroconversion and STI transmission within such subpopulations that use stimulants (Halkitis et al., 2006; Hoenigl et al., 2016). As partnership dynamics likely influence sexualized substance use and subsequent sexual risk behaviors, these results suggest the potential utility of sexual partnership-based interventions that combine substance use treatment with HIV prevention.

Limitations

Our findings must be considered within the context of limitations. As a machine learning variable selection technique, lasso aids interpretation by selecting a distinct subset of predictors but tends to exclude correlated variables from models, potentially leading to misspecification and omitted variables bias, relative to ridge regression and other techniques. However, we accounted for this by conducting sensitivity analyses to ensure that correlated variables were not spuriously deleted (results not shown). It is also important to note that, while lasso selected variables that were highly correlated with our outcome, we are unable to make causal inferences from our models. As this was a secondary data analysis, our study was constrained to measures that were contained in the dataset, resulting in potential omitted variables bias and measurement bias. This consideration is particularly relevant as certain constructs regarding sexual risk behaviors and substance use were not captured within the dataset, such as partnership dyadic characteristics and contexts/settings in which substance use and sexual activities took place, which should be considered in the interpretation of our results and represent an important area of future research. We presented confidence intervals to aid interpretation of the regression coefficients but suggest caution making statistical inferences in the context of variable selection (Lockhart et al., 2014). As increased stimulant use was reported in 20.9% of visits, AUC may be overestimated due to imbalances in our data. However, we sought to overcome this potential limitation by utilizing multiple metrics to assess model performance when selecting our models, such as the MCC which is robust to class imbalance and asymmetry (Chicco & Jurman, 2020). Finally, as this cohort comprises a diverse sample of MSM with high rates of substance use, this limits the generalizability of our findings to other subpopulations of MSM.

Conclusions

This study is among the first to utilize lasso for variable selection to evaluate factors associated with increased stimulant use among a diverse cohort of MSM. Our analysis adds to the literature by demonstrating that variables commonly collected in HIV and substance use research can be used to build models which predict stimulant use with a reasonable degree of accuracy. Furthermore, this study is among the first to explicitly evaluate differences in factors that may contribute to increased stimulant use based on HIV status. As engagement in healthcare may differ according to HIV status (Babel et al., 2021; Powers & Miller, 2015), characterizing differences in predictors of stimulant use based on HIV status is crucial toward the development of efficacious HIV/STI interventions that can be incorporated into HIV treatment and prevention efforts. Our findings demonstrate that increased stimulant use was positively associated with unstable housing and transactional sex regardless of HIV status. These results underscore the importance of designing HIV prevention interventions that address the underlying factors that often drive the interdependent relationship between unstable housing, stimulant use, and transactional sex which contribute to ongoing HIV/STI transmission among vulnerable MSM subpopulations. Specifically, strategies such as contingency management may prove beneficial in these populations (Tracy et al., 2007).

Our analysis revealed that polysubstance use was associated with increased stimulant use among MWH, whereas sexual risk behaviors, sexualized substance use, and last partner substance use were correlated with increased stimulant use among HIV-negative participants. These findings indicate that the underlying motivations and factors which contribute to stimulant use patterns among MSM likely differ based on HIV status and suggest that these distinctions should be considered in the design of HIV prevention and treatment interventions. Specifically, our results demonstrate the potential role of interventions that integrate substance use treatment into the HIV care continuum and reduce HIV-related stigma among MWH who use stimulants, such as educational programming, counseling, and linkages to support groups (Heijnders & Van Der Meij, 2006). Conversely, HIV-negative MSM who use stimulants may benefit from HIV prevention interventions that address sexualized substance use as well as sexual partnership-based interventions.

Supplementary Material

Supplemental Tables

Funding:

mSTUDY is funded by the National Institute on Drug Abuse (U01 DA036267). CSB was supported by the National Institute of Mental Health (T32 MH080634) and the National Institute on Drug Abuse (K23 DA054004). WSC, SS, and PMG were supported by the Center for HIV Identification, Prevention, and Treatment Services (CHIPTS; P30 MH058107).

Footnotes

Compliance with Ethical Standards:

Ethics approval: The study was reviewed and approved by the Office of Human Research Participant Protection (OHRPP) at the University of California, Los Angeles. The study was performed in accordance with the ethical standards as laid down in the Declaration of Helsinki.

Consent to participate: Written informed consent was obtained from all participants prior to enrollment in the mSTUDY.

Conflicts of Interest: The authors have no relevant financial or non-financial interests to disclose.

References

  1. Anderson-Carpenter KD, Fletcher JB, & Reback CJ (2017). Associations between methamphetamine use, housing status, and incarceration rates among men who have sex with men and transgender women. J Drug Issues, 47(3), 383–395. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Aralis HJ, Shoptaw S, Brookmeyer R, Ragsdale A, Bolan R, & Gorbach PM (2018). Psychiatric illness, substance use, and viral suppression among HIV-positive men of color who have sex with men in Los Angeles. AIDS Behav, 22(10), 3117–3129. 10.1007/s10461-018-2055-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Baams L, Wilson BDM, & Russell ST (2019). LGBTQ youth in unstable housing and foster care. Pediatrics, 143(3), e20174211. 10.1542/peds.2017-4211 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Babel RA, Wang P, Alessi EJ, Raymond HF, & Wei C. (2021). Stigma, HIV risk, and access to HIV prevention and treatment services among men who have sex with men (MSM) in the United States: a scoping review. AIDS Behav, 25(11), 3574–3604. 10.1007/s10461-021-03262-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Baggaley RF, White RG, & Boily MC (2010). HIV transmission risk through anal intercourse: systematic review, meta-analysis and implications for HIV prevention. Int J Epidemiol, 39(4), 1048–1063. 10.1093/ije/dyq057 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Barman-Adhikari A, Begun S, Rice E, Yoshioka-Maxwell A, & Perez-Portillo A. (2016). Sociometric network structure and its association with methamphetamine use norms among homeless youth. Soc Sci Res, 58, 292–308. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Berry MS, Bruner NR, Herrmann ES, Johnson PS, & Johnson MW (2020). Methamphetamine administration dose effects on sexual desire, sexual decision making, and delay discounting. Exp Clin Psychopharmacol. 10.1037/pha0000398 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Bourne A, Reid D, Hickson F, Torres-Rueda S, Steinberg P, & Weatherburn P. (2015). “Chemsex” and harm reduction need among gay men in South London. Int J Drug Policy, 26(12), 1171–1176. [DOI] [PubMed] [Google Scholar]
  9. Bourne A, Reid D, Hickson F, Torres-Rueda S, & Weatherburn P. (2015). Illicit drug use in sexual settings (‘chemsex’) and HIV/STI transmission risk behaviour among gay men in South London: findings from a qualitative study. Sex Transm Infect, 91(8), 564–568. [DOI] [PubMed] [Google Scholar]
  10. Brown RE, Turner C, Hern J, & Santos G-M (2017). Partner-level substance use associated with increased sexual risk behaviors among men who have sex with men in San Francisco, CA. Drug Alcohol Depend, 176, 176–180. 10.1016/j.drugalcdep.2017.02.016 [DOI] [PubMed] [Google Scholar]
  11. Bungay V, Malchy L, Buxton JA, Johnson J, MacPherson D, & Rosenfeld T. (2006). Life with jib: a snapshot of street youth’s use of crystal methamphetamine. Addict Res Theory, 14(3), 235–251. [Google Scholar]
  12. Centers for Disease Control and Prevention [CDC]. (2019). HIV infection risk, prevention, and testing behaviors among men who have sex with men - National HIV Behavioral Surveillance, 23 U.S. cities, 2017 (HIV Surveillance Special Report 22).
  13. Chartier M, Araneta A, Duca L, McGlynn LM, Gore-Felton C, Goldblum P, & Koopman C. (2009). Personal values and meaning in the use of methamphetamine among HIV-positive men who have sex with men. Qual Health Res, 19(4), 504–518. 10.1177/1049732309333018 [DOI] [PubMed] [Google Scholar]
  14. Chicco D, & Jurman G. (2020). The advantages of the Matthews correlation coefficient (MCC) over F1 score and accuracy in binary classification evaluation. BMC genomics, 21(1), 6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Colyer SP, Moore DM, Cui Z, Zhu J, Armstrong HL, Taylor M, … Lachowsky NJ (2020). Crystal methamphetamine use and initiation among gay, bisexual, and other men who have sex with men living with HIV in a treatment as prevention environment. Subst Use Misuse, 55(14), 2428–2437. 10.1080/10826084.2020.1833925 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Derrick JL, Wittkower LD, & Pierce JD (2019). Committed relationships and substance use: recent findings and future directions. Curr Opin Psychol, 30, 74–79. 10.1016/j.copsyc.2019.03.002 [DOI] [PubMed] [Google Scholar]
  17. Drumright LN, Patterson TL, & Strathdee SA (2006). Club drugs as causal risk factors for HIV acquisition among men who have sex with men: a review. Subst Use Misuse, 41(10–12), 1551–1601. 10.1080/10826080600847894 [DOI] [PubMed] [Google Scholar]
  18. Earnshaw VA, Eaton LA, Collier ZK, Watson RJ, Maksut JL, Rucinski KB, … Kalichman SC (2020). HIV stigma, depressive symptoms, and substance use. AIDS Patient Care STDS, 34(6), 275–280. 10.1089/apc.2020.0021 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Edelman EJ, Cole CA, Richardson W, Boshnack N, Jenkins H, & Rosenthal MS (2016). Stigma, substance use and sexual risk behaviors among HIV-infected men who have sex with men: a qualitative study. Prev Med Rep, 3, 296–302. 10.1016/j.pmedr.2016.03.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Fletcher JB, Clark KA, & Reback CJ (2020). Depression and HIV transmission risk among methamphetamine-using men who have sex with men. Addict Res Theory, 1–8. 10.1080/16066359.2020.1807960 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Freeman P, Walker BC, Harris DR, Garofalo R, Willard N, & Ellen JM (2011). Methamphetamine use and risk for HIV among young men who have sex with men in 8 US cities. Arch Pediatr Adolesc Med, 165(8), 736–740. 10.1001/archpediatrics.2011.118 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Fulcher K, Shumka L, Roth E, & Lachowsky N. (2019). Pleasure, risk perception and consent among group sex party attendees in a small Canadian urban centre. Cult Health Sex, 21(6), 650–665. 10.1080/13691058.2018.1508749 [DOI] [PubMed] [Google Scholar]
  23. Gamarel KE, Woolf-King SE, Carrico AW, Neilands TB, & Johnson MO (2015). Stimulant use patterns and HIV transmission risk among HIV-serodiscordant male couples. Journal of acquired immune deficiency syndromes (1999), 68(2), 147–151. 10.1097/QAI.0000000000000418 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Giorgetti R, Tagliabracci A, Schifano F, Zaami S, Marinelli E, & Busardò FP (2017). When “chems” meet sex: a rising phenomenon called “chemsex”. Curr Neuropharmacol, 15(5), 762–770. 10.2174/1570159×15666161117151148 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Glick SN, Burt R, Kummer K, Tinsley J, Banta-Green CJ, & Golden MR (2018). Increasing methamphetamine injection among non-MSM who inject drugs in King County, Washington. Drug Alcohol Depend, 182, 86–92. 10.1016/j.drugalcdep.2017.10.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Glynn TR, Llabre MM, Lee JS, Bedoya CA, Pinkston MM, O’Cleirigh C, & Safren SA (2019). Pathways to health: an examination of HIV-related stigma, life stressors, depression, and substance use. Int J Behav Med, 26(3), 286–296. https://link.springer.com/content/pdf/10.1007/s12529-019-09786-3.pdf [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Halkitis PN, Fischgrund BN, & Parsons JT (2005). Explanations for methamphetamine use among gay and bisexual men in New York City. Subst Use Misuse, 40(9–10), 1331–1345. 10.1081/JA-200066900 [DOI] [PubMed] [Google Scholar]
  28. Halkitis PN, Green KA, & Carragher DJ (2006). Methamphetamine use, sexual behavior, and HIV seroconversion. Journal of Gay & Lesbian Psychotherapy, 10(3–4), 95–109. 10.1300/J236v10n03_09 [DOI] [Google Scholar]
  29. Halkitis PN, Green KA, & Mourgues P. (2005). Longitudinal investigation of methamphetamine use among gay and bisexual men in New York City: findings from project BUMPS. J Urban Health, 82(1), i18–i25. 10.1093/jurban/jti020 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Halkitis PN, Mukherjee PP, & Palamar JJ (2008). Longitudinal modeling of methamphetamine use and sexual risk behaviors in gay and bisexual men. AIDS Behav, 13(4), 783–791. 10.1007/s10461-008-9432-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Hastie T, Tibshirani R, & Friedman J. (2009). The elements of statistical learning: data mining, inference, and prediction. Springer Science & Business Media. [Google Scholar]
  32. Hastie T, Tibshirani R, & Wainwright M. (2015). Statistical learning with sparsity: the lasso and generalizations. CRC press. [Google Scholar]
  33. Hegazi A, Lee MJ, Whittaker W, Green S, Simms R, Cutts R, … Pakianathan MR (2017). Chemsex and the city: sexualised substance use in gay bisexual and other men who have sex with men attending sexual health clinics. Int J STD AIDS, 28(4), 362–366. 10.1177/0956462416651229 [DOI] [PubMed] [Google Scholar]
  34. Heijnders M, & Van Der Meij S. (2006). The fight against stigma: An overview of stigma-reduction strategies and interventions. Psychology, Health & Medicine, 11(3), 353–363. 10.1080/13548500600595327 [DOI] [PubMed] [Google Scholar]
  35. Hoenigl M, Chaillon A, Moore DJ, Morris SR, Smith DM, & Little SJ (2016). Clear links between starting methamphetamine and increasing sexual risk behavior: a cohort study among men who have sex with men. J Acquir Immune Defic Syndr, 71(5), 551–557. 10.1097/qai.0000000000000888 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Hojilla JC, Vlahov D, Glidden DV, Amico KR, Mehrotra M, Hance R, … Carrico AW (2018). Skating on thin ice: stimulant use and sub‐optimal adherence to HIV pre‐exposure prophylaxis. Journal of the International AIDS Society, 21(3), e25103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Hood JE, Buskin SE, Golden MR, Glick SN, Banta-Green C, & Dombrowski JC (2018). The changing burden of HIV attributable to methamphetamine among men who have sex with men in King County, Washington. AIDS Patient Care STDS, 32(6), 223–233. 10.1089/apc.2017.0306 [DOI] [PubMed] [Google Scholar]
  38. Javanbakht M, Ragsdale A, Shoptaw S, & Gorbach PM (2019). Transactional sex among men who have sex with men: differences by substance use and HIV status. J Urban Health, 96(3), 429–441. 10.1007/s11524-018-0309-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Javanbakht M, Shoptaw S, Ragsdale A, Brookmeyer R, Bolan R, & Gorbach PM (2020). Depressive symptoms and substance use: changes over time among a cohort of HIV-positive and HIV-negative MSM. Drug Alcohol Depend, 207, 107770. 10.1016/j.drugalcdep.2019.107770 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Jerome RC, Halkitis PN, & Siconolfi DE (2009). Club drug use, sexual behavior, and HIV seroconversion: a qualitative study of motivations. Subst Use Misuse, 44(3), 431–447. 10.1080/10826080802345036 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Johnson G, & Chamberlain C. (2008). Homelessness and substance abuse: which comes first? Australian Social Work, 61(4), 342–356. [Google Scholar]
  42. Lai HH, Kuo YC, Kuo CJ, Lai YJ, Chen M, Chen YT, … Yen YF (2020). Methamphetamine use associated with non-adherence to antiretroviral treatment in men who have sex with men. Sci Rep, 10(1), 7131. 10.1038/s41598-020-64069-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Lasko TA, Bhagwat JG, Zou KH, & Ohno-Machado L. (2005). The use of receiver operating characteristic curves in biomedical informatics. J Biomed Inform, 38(5), 404–415. 10.1016/j.jbi.2005.02.008 [DOI] [PubMed] [Google Scholar]
  44. LeMessurier J, Traversy G, Varsaneux O, Weekes M, Avey MT, Niragira O, … Rodin R. (2018). Risk of sexual transmission of human immunodeficiency virus with antiretroviral therapy, suppressed viral load and condom use: a systematic review. Can Med Assoc J, 190(46), E1350–E1360. 10.1503/cmaj.180311 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Lockhart R, Taylor J, Tibshirani RJ, & Tibshirani R. (2014). A significance test for the lasso. Ann Stat, 42(2), 413–468. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Loza O, Curiel ZV, Beltran O, & Ramos R. (2020). Methamphetamine use and sexual risk behaviors among men who have sex with men in a Mexico-US border city. Am J Addict, 29(2), 111–119. 10.1111/ajad.12985 [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Mereish EH, Goldbach JT, Burgess C, & DiBello AM (2017). Sexual orientation, minority stress, social norms, and substance use among racially diverse adolescents. Drug Alcohol Depend, 178, 49–56. 10.1016/j.drugalcdep.2017.04.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Mimiaga MJ, Reisner SL, Tinsley JP, Mayer KH, & Safren SA (2009). Street workers and internet escorts: contextual and psychosocial factors surrounding HIV risk behavior among men who engage in sex work with other men. J Urban Health, 86(1), 54–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Mitchell JW, Boyd C, McCabe S, & Stephenson R. (2014). A cause for concern: male couples’ sexual agreements and their use of substances with sex. AIDS Behav, 18(7), 1401–1411. 10.1007/s10461-014-0736-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Ostrow DG, Plankey MW, Cox C, Li X, Shoptaw S, Jacobson LP, & Stall RC (2009). Specific sex drug combinations contribute to the majority of recent HIV seroconversions among MSM in the MACS. J Acquir Immune Defic Syndr, 51(3), 349–355. 10.1097/QAI.0b013e3181a24b20 [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Patterson TL, Semple SJ, Zians JK, & Strathdee SA (2005). Methamphetamine-using HIV-positive men who have sex with men: correlates of polydrug use. J Urban Health, 82(1), i120–i126. 10.1093/jurban/jti031 [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Powers KA, & Miller WC (2015). Critical review: building on the HIV cascade: a complementary “HIV states and transitions” framework for describing HIV diagnosis, care, and treatment at the population level. J Acquir Immune Defic Syndr, 69(3), 341–347. 10.1097/qai.0000000000000611 [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Reback CJ, & Fletcher JB (2018). Elevated HIV and STI prevalence and incidence among methamphetamine-using men who have sex with men in Los Angeles County. AIDS Educ Prev, 30(4), 350–356. 10.1521/aeap.2018.30.4.350 [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Reback CJ, Larkins S, & Shoptaw S. (2004). Changes in the meaning of sexual risk behaviors among gay and bisexual male methamphetamine abusers before and after drug treatment. AIDS Behav, 8(1), 87–98. 10.1023/B:AIBE.0000017528.39338.75 [DOI] [PubMed] [Google Scholar]
  55. Romero AP, Goldberg SK, & Vasquez LA (2020). LGBT people and housing affordability, discrimination, and homelessness. [Google Scholar]
  56. Semple SJ, Strathdee SA, Zians J, & Patterson TL (2010). Social and behavioral characteristics of HIV-positive MSM who trade sex for methamphetamine. Am J Drug Alcohol Abuse, 36(6), 325–331. 10.3109/00952990.2010.505273 [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Shariati H, Armstrong HL, Cui Z, Lachowsky NJ, Zhu J, Anand P, … Moore DM (2017). Changes in smoking status among a longitudinal cohort of gay, bisexual, and other men who have sex with men in Vancouver, Canada. Drug Alcohol Depend, 179, 370–378. 10.1016/j.drugalcdep.2017.07.025 [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Shoptaw S, & Reback CJ (2007). Methamphetamine use and infectious disease-related behaviors in men who have sex with men: implications for interventions. Addiction, 102(s1), 130–135. 10.1111/j.1360-0443.2006.01775.x [DOI] [PubMed] [Google Scholar]
  59. Socias ME, & Milloy MJ (2018). Substance use and adherence to antiretroviral therapy: what is known and what is unknown. Curr Infect Dis Rep, 20(9), 36. 10.1007/s11908-018-0636-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Stata Corp LLC. (2019). Lasso reference manual. Stata Press. [Google Scholar]
  61. Substance Abuse and Mental Health Services Administration [SAMHSA]. (2019). 2018 National Survey on Drug Use and Health: lesbian, gay, and bisexual (LGB) adults. https://www.samhsa.gov/data/sites/default/files/reports/rpt23252/7_LGB_2020_01_14_508.pdf
  62. Swartz JA, & McCarty-Caplan D. (2018). A study of the longitudinal patterns of stimulant and amyl nitrite use and sexual behavior pre-and post-HIV seroconversion among MSM. AIDS Behav, 22(4), 1395–1409. https://link.springer.com/content/pdf/10.1007/s10461-017-2008-y.pdf [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Tibshirani R. (1996). Regression shrinkage and selection via the lasso. J R Stat Soc Series B Stat Methodol, 58(1), 267–288. [Google Scholar]
  64. Tibshirani R. (2011). Regression shrinkage and selection via the lasso: a retrospective. J R Stat Soc Series B Stat Methodol, 73(3), 273–282. 10.1111/j.1467-9868.2011.00771.x [DOI] [Google Scholar]
  65. Tracy K, Babuscio T, Nich C, Kiluk B, Carroll KM, Petry NM, & Rounsaville BJ (2007). Contingency management to reduce substance use in individuals who are homeless with co-occurring psychiatric disorders. Am J Drug Alcohol Abuse, 33(2), 253–258. 10.1080/00952990601174931 [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Hawkins, B. W, Armstrong HL, Kesselring S, Rich AJ, Cui Z, Sereda P, … Roth EA (2019). Substance use as a mechanism for social inclusion among gay, bisexual, and other men who have sex with men in Vancouver, Canada. Subst Use Misuse, 54(12), 1945–1955. 10.1080/10826084.2019.1621901 [DOI] [PMC free article] [PubMed] [Google Scholar]
  67. Weatherburn P, Hickson F, Reid D, Torres-Rueda S, & Bourne A. (2017). Motivations and values associated with combining sex and illicit drugs (‘chemsex’) among gay men in South London: findings from a qualitative study. Sex Transm Infect, 93(3), 203–206. [DOI] [PubMed] [Google Scholar]
  68. Workowski KA, & Bolan GA (2015). Sexually transmitted diseases treatment guidelines, 2015. MMWR Recomm Rep, 64(RR-03), 1–137. [PMC free article] [PubMed] [Google Scholar]
  69. Wray TB, & Monti PM (2019). Characteristics of sex events, partners, and motivations and their associations with HIV-risk behavior in a daily diary study of high-risk men who have sex with men (MSM). AIDS Behav, 24(6), 1851–1864. [DOI] [PMC free article] [PubMed] [Google Scholar]
  70. Wu E, El-Bassel N, McVinney LD, Hess L, Fopeano MV, Hwang HG, … Mansergh G. (2014). The association between substance use and intimate partner violence within Black male same-sex relationships. J Interpers Violence, 30(5), 762–781. 10.1177/0886260514536277 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Tables
NIHMS1950898-supplement-Supplemental_Tables.docx (40KB, docx)

RESOURCES