The Syndemic Factors of Violence Exposure, Substance Use, and Mental Health Problems: Relationships to Sexual Risk Behaviors in HIV-Negative Young Men Who Have Sex With Men

Abstract

Objective:

A syndemic of violence exposure, substance misuse, and mental health problems (i.e., depression, anxiety, posttraumatic stress disorder) is associated with increased unprotected anal sex and number of sexual partners in young men who have sex with men (YMSM). However, few studies have examined multiple forms of violence—including childhood abuse, intimate partner violence, and community violence—and identified which factors are significant predictors of HIV risk for YMSM when controlling for all other factors. Accordingly, this study examined the additive and independent influence of different forms of violence, substance misuse, and mental health problems associated with sexual risk behaviors in YMSM who used HIV prevention services.

Method:

A convenience sample of 168 (97 Black or multiethnic Black, 71 white) YMSM ages 18–34 completed computer-assisted personal interviews assessing syndemic factors and HIV risk behaviors. We conducted Spearman correlations and negative binomial regressions to describe syndemic relationships and identify the significant independent predictors of HIV risk.

Results:

A syndemic of violence exposure, substance use, and mental health problems was observed in the sample. Also, participants with depression, drug abuse in the clinical range, and polydrug use reported significantly higher frequencies of sexual risk behaviors. Violence exposure did not uniquely predict sexual risk.

Conclusions:

Depression, drug abuse, and polydrug use should be targets for HIV prevention among YMSM using HIV prevention services.

Although HIV infection rates have declined recently in the United States, men who have sex with men (MSM) remain disproportionately affected by the disease (Centers for Disease Control and Prevention [CDC], 2019). Young MSM (YMSM) between the ages of 13 and 34 years account for the greatest number of new HIV infections in the MSM population, with Black YMSM bearing the highest burden of disease (CDC, 2018). Engaging in sexual risk behaviors (e.g., unprotected anal sex or anal sex with multiple partners in the absence of preexposure prophylaxis and treatment as prevention) remains the leading HIV transmission mode for YMSM (CDC, 2018), and several studies have demonstrated that exposure to violence, substance misuse, and mental health problems are independently associated with increased sexual risk behaviors in this population (Agnew-Brune et al., 2019; Duncan et al., 2018; Freeman et al., 2011; Halkitis et al., 2018; Koblin et al., 2006; Mutchler et al., 2012; Ristuccia et al., 2018).

Other recent studies have shown that violence exposure, substance misuse, and mental health problems act synergistically to increase HIV risk in vulnerable populations, including YMSM (Halkitis et al., 2013; Lee et al., 2020; Maiorana et al., 2021; Mustanski et al., 2017). These synergistic relationships are conceptualized as an HIV syndemic (Singer, 1994). Although many studies have examined the syndemic relationship between mental health problems (i.e., depression, anxiety, and posttraumatic stress disorder [PTSD]), substance misuse, violence, and HIV risk in YMSM, most have narrowly conceptualized violence, defining violence as childhood sexual abuse, intimate partner violence (IPV), or exposure to community-level violence and omitting other more common forms of child maltreatment, such as physical and emotional abuse. A more comprehensive examination of violence exposure among YMSM that includes childhood maltreatment—along with the identification of the other syndemic factors that are uniquely significantly associated with sexual risk behaviors among YMSM—may inform development and improvement of targeted HIV prevention programs.

Literature Review

Violence Exposure and HIV Risk Behaviors in Young Men Who Have Sex with Men

Several types of violence exposure have been investigated as risk factors for HIV, including histories of family violence (e.g., childhood sexual and physical abuse), IPV, and community-level violence (e.g., being shot or mugged). From a socioecological perspective, being a member of a sexual and/or racial minority group in the United States is associated with increased risk for violence exposure (Merrick et al., 2018), which predisposes individuals to increased health risks, including behaviors that place them at higher risk for sexually transmitted infections. For example, studies examining the influence of childhood sexual abuse have consistently demonstrated that MSM are more likely to have experienced childhood sexual abuse than their heterosexual peers and that MSM who reported histories of childhood sexual abuse were more likely to be infected with HIV and to engage in unprotected sex than those without sexual abuse histories (Lloyd & Operario, 2012). Moreover, a greater severity of childhood sexual abuse is associated with higher sexual risk for MSM. A study by Welles and colleagues (2009) found that experiencing more frequent childhood sexual abuse was significantly associated with having a greater number of sexual partners and more frequent unprotected anal sex. Regarding physical abuse, a study by Schilder et al. (2014) found that experiencing physical abuse as a child was associated with increased HIV risk among YMSM. However, the literature examining the relationship between physical and emotional child abuse, including the severity of these forms of abuse, and HIV risk in MSM is limited.

To date, most studies exploring the relationship between violence exposure and HIV risk factors have used samples of adult MSM with a wide range of ages, with little research examining violence exposure among YMSM specifically. However, the research regarding HIV sexual risk and violence among YMSM indicates a significant relationship between experiencing IPV and increased HIV sexual risk (Koblin et al., 2006; Mustanski et al., 2012) as well as between exposure to community violence and HIV sexual risk in Black YMSM living with HIV (Quinn et al., 2016).

Substance Use and HIV Risk in Young Men Who Have Sex with Men

Use of certain substances, such as amphetamines and alcohol, can result in disinhibition and cognitive impairment that may lead individuals to engage in sexual risk behaviors. Use of specific substances has also been consistently identified with increased sexual risk among YMSM (Freeman et al., 2011; Halkitis et al., 2018; Mutchler et al., 2012; Newcomb & Mustanski, 2014; Newcomb et al., 2011), but the strength of the relationship has differed by the type of substance. For example, evidence regularly demonstrates significant, positive relationships between both methamphetamine and ecstasy use and sexual risk (Freeman et al., 2011; Mutchler et al., 2012), whereas findings regarding the relationship between alcohol use and sexual risk have been less consistent. Two studies reported that alcohol use was positively associated with increased HIV sexual risk (Moeller et al., 2014; Newcomb & Mustanski, 2014); other research found no significant relationship between the two (Pollock et al., 2012). Further, there is some evidence that age may moderate the relationship between alcohol use and HIV sexual risk, with alcohol use only a significant risk factor for increased HIV sexual risk among younger YMSM (Newcomb & Mustanski, 2014).

Mental Health and HIV Risk in Young Men Who Have Sex with Men

Finally, depression, anxiety, and PTSD have been identified as factors that contribute to HIV risk behaviors in YMSM. For example, YMSM who experience depression may be less motivated to engage in health-promoting behaviors such as condom use due to feelings of apathy or hopelessness associated with depression (Allgower et al., 2001). Moreover, YMSM with anxiety symptoms may engage in unprotected sex as a way of briefly escaping their anxious feelings (Yi et al., 2010). In fact, a study using a probability sample of MSM recruited through cluster random sampling of gay community-identified neighborhoods found that MSM with the greatest number of depressive symptoms were significantly more likely to engage in unprotected sex than those with the fewest symptoms (Fendrich et al., 2013). Moreover, Agnew-Brune and colleagues (2019) found that YMSM from the National Behavioral Health Surveillance study were more likely to engage in condomless anal intercourse as their depression and anxiety symptoms increased. Similarly, two studies demonstrated that experiencing a greater number of PTSD symptoms—such as reexperiencing and hyperarousal—was associated with a greater likelihood of engaging in unprotected sex (O’Cleirigh et al., 2013; Radcliffe et al., 2011).

Syndemic Model

Several researchers have used the term syndemicic to describe the co-occurrence of violence exposure, substance misuse, and mental health problems and their relationships to HIV risk among MSM. The syndemic model was first proposed by Singer (1994) to explain the relationships between substance use, violence, and AIDS among people of color living in an urban environment. This initial syndemic model—referred to in the literature as SAVA (Singer, 1994, 2009)—posits that problems of substance abuse, violence, and AIDS are interrelated to the extent that, in certain populations, it is nearly impossible to understand one problem in the absence of the others. The SAVA syndemic model was later adapted to include four psychosocial problems (polydrug use, depression, partner abuse, and childhood sexual abuse) for application to a large sample of MSM (Stall et al., 2003); a slightly different set of psychosocial problems (substance use, psychological distress, partner violence, and sexual assault) was applied to YMSM (Mustanski et al., 2007). More recently, the SAVA syndemic model has been adapted and tested in numerous studies of MSM, including a sample of African American MSM (O’Leary et al., 2014), a global sample of MSM (Santos et al., 2014), and an ethnically diverse sample of adult MSM in New York City (Safren et al., 2018).

Fewer recent studies using the syndemic model have been conducted among samples of YMSM. However, recent syndemic studies of HIV risk in YMSM have used similar syndemic factors: substance use, mental health problems, and violence exposure (Mustanski et al., 2017). Yet, studies vary in the operationalization of these factors, especially regarding violence exposure. For example, a recent study by da Silva and colleagues (2020) used a measure of community violence to capture the violence exposure component of a syndemic affecting young Black men and transgender women who have sex with men. Recent syndemic studies of HIV risk in young Latino MSM did not include community violence but did assess for IPV and childhood sexual abuse (Blashill et al., 2020; Lee et al., 2020). Other studies have focused on bullying and/or IPV (Maiorana et al., 2021; Turpin et al., 2020), and one study omitted violence in its syndemic analyses (Halkitis et al., 2015). One of the more comprehensive operationalizations of violence included victimization or threats of violence because of one’s sexual orientation, IPV, and childhood sexual abuse (Mustanski et al., 2017).

Although all the aforementioned studies have provided valuable insight into the relationships between various syndemic factors and HIV risk among diverse samples of YMSM, no known syndemic studies of HIV risk among YMSM have included all types of child maltreatment (i.e., sexual, physical, and emotional abuse), IPV, and community-level violence exposure. Assessing all these forms of violence exposure across individuals in the same sample contributes to a more complete understanding of the level of violence experienced by YMSM and allows us to test for the unique contribution of each form of violence to HIV sexual risk behaviors.

To determine the syndemic relationship between co-occurring risk factors and HIV risk behaviors, most studies create an additive index. Few studies have begun to identify which specific syndemic factors are uniquely associated with greater HIV risk behaviors after controlling for the other factors in YMSM, particularly among those who are HIV-negative and use HIV prevention services. Untangling the syndemic factors that comprise the index to identify those factors that are independently significantly associated with HIV risk may deepen our understanding of potential HIV prevention targets for YMSM by identifying which specific mental health, substance use, and violence-related mechanisms of HIV risk to address. Therefore, the current study explored the following research questions:

1. To what extent do Black and white YMSM experience exposure to violence, substance use, and mental health problems?

2. Is there a syndemic effect of violence exposure, substance misuse, andmental health problems on HIV risk behaviors among Black and white YMSM who use HIV prevention services?

3. Which of these factors are significant predictors of HIV risk behaviors when controlling for other syndemic factors and demographics in this population?

Method

Participants

Participants were 168 YMSM recruited from four community-based AIDS service organizations that provide HIV prevention and testing services in St. Louis and Kansas City, MO. Participants were ages 18–34 years with a mean age of 25.1 (SD = 3.7). For analyses, bisexual, queer, and pansexual were recoded into one category because of small cell sizes, as suggested by Flanders (2017) since they are considered nonmonosexual, as opposed to the monosexual identity of gay. The ethnic composition of the sample was 51.5% (n = 86) Black and 44.3% (n = 74) white; 4.2% (n = 7) identified as more than one ethnicity (i.e., two identified as Black and white; two as Black, white, and First Nations; one as Black and Latino; one as Black, Latino, Native American, and white; and one as other). For comparative analyses by ethnicity, participants who identified as multiethnic or other were collapsed with participants who identified as Black to allow for comparisons between ethnic majority (i.e., white) and minority (nonwhite) participants. This Black and multiethnic Black (BMEB) categorization is supported by recent research demonstrating that both white and Black individuals in the United States are likely to engage in hypodescent, in which any person whose ethnicity includes Black is categorized as Black by others (Ho et al., 2017). Table 1 presents demographic characteristics of the total sample by ethnicity.

Table 1.

Participant Demographics for Total Sample and by Ethnicity

Demographic Variable	Total Sample	BMEB MSM	White MSM
	n (%)	n (%)	n (%)
Ethnicity
Black	90 (53.57)	–	–
White	71 (42.26)	–	–
More than one ethnicity	7 (4.17)	–	–
Sexual identity
Gay	132 (78.57)	66 (68.04)	66 (92.96)
Bisexual	32 (19.05)	28 (28.87)	4 (5.63)
Queer, pansexual	4 (2.38)	3 (3.09)	1 (1.41)
Employment
Full time	122 (72.62)	68 (70.10)	54 (76.06)
Part time	32 (19.05)	17 (17.53)	15 (21.03)
Unemployed	14 (8.33)	12 (12.37)	2 (2.82)
Education level
Did not complete high school	2 (1.19)	2 (2.06)	0 (0)
High school or GED	26 (15.48)	21 (21.65)	5 (7.04)
Some college	64 (38.10)	41 (42.27)	23 (32.39)
Associates/technical degree	16 (9.52)	6 (6.19)	10 (14.08)
Bachelor’s degree	47 (27.98)	23 (23.71)	24 (33.08)
Graduate school/post college	13 (7.74)	4 (4.12)	9 (12.68)
Income (monthly)
$0–$499	15 (9.04)	10 (10.42)	5 (7.14)
$500–$999	26 (15.66)	13 (13.54)	13 (18.57)
$1,000–$1,499	32 (19.28)	23 (23.96)	9 (12.68)
$1,500–$1,999	25 (15.06)	11 (11.46)	14 (20.00)
$2,000–$2,499	27 (16.27)	12 (12.50)	15 (21.43)
$2,500–$2,999	13 (7.83)	9 (9.38)	4 (5.71)
$3,000+	28 (16.87)	18 (18.75)	10 (14.29)
City
St. Louis, MO	124 (74.25)	75 (77.32)	49 (69.01)
Kansas City, MO	44 (26.19)	22 (22.68)	22 (30.99)
Age (years) M(SD)	25.01 (3.58)	24.81 (3.77)	25.27 (3.31)

Note. BMEB = Black and multiethnic Black; MSM = men who have sex with men.

Study Procedures

To participate in the study, participants were required to (a) identify as male; (b) be between the ages of 18 and 34 years; (c) identify as white, Black, or multiethnic including Black; (d) report engaging in anal or oral sex with another man in the last 3 months; and (e) have a negative or unknown serostatus. Participants were recruited through conversations with the interviewer in partner agency waiting rooms, announcements during group HIV prevention programs, and by referral from HIV prevention and testing staff. Approximately 2.8% of participants screened for eligibility declined to participate in the study.

Those who consented to participate completed a 15- to 20-minute computer-assisted personal interview in a private room with the interviewer, who was available to answer clarifying questions that arose for the participant. The interviewer was not able to see participant responses, increasing participant privacy and likelihood of providing accurate results. REDCap (Research Electronic Data Capture) software was used to collect and manage data (Harris et al., 2009). Participants received a $15 gift card for their time. The study was approved by the institutional review board at Washington University in St. Louis.

Measures

HIV Risk Behaviors

Participants reported the number of unprotected anal sex (UAS) occasions and number of male sexual partners in the past 3 and 12 months. In this study, UAS was defined as anal sex without the use of condoms or preexposure prophylaxis (PrEP). Count measures using 3- and 12-month timeframes were selected based on studies of participant recall that found that retrospective periods of 3 or 12 months yield the most accurate reports of past sexual behaviors, and that frequencies are preferred over dichotomous measures of condomless sex (Noar et al., 2006; Schroder et al., 2003).

Violence Exposure

The current study assessed for severity of child abuse, IPV, and experiencing and witnessing community violence. Experiences of child abuse before the age of 18 were measured using three subscales of the Childhood Trauma Questionnaire (i.e., physical, sexual, and emotional abuse; Bernstein et al., 1994). Responses were on a 5-point Likert-type scale that ranged from “never true” to “very often true.” Intimate partner violence in the last 12 months was assessed using the IPV-GBM Scale (Stephenson & Finneran, 2013), which includes five subscales: physical and sexual (e.g., partner punched or hit you), monitoring (e.g., demanded access to your cell phone), controlling (e.g., prevented you from seeing your friends), HIV-related (e.g., lied about his HIV status), and emotional (e.g., call you fat or ugly). Responses on a 6-point Likert-type scale ranged from “never” to “more than 5 times.” The neighborhood domain of the Centers for Disease Control and Prevention M2 Victimization Scale (Dahlberg et al., 2005) was used to measure experiencing and witnessing community violence, including being physically threatened, attacked, robbed, or harassed, or seeing any of those things happen in one’s neighborhood. Participants responded on a 4-point Likert scale that ranged from “never” to “often.” Internal consistency reliability was adequate across all measures and ranged from α = .72 to α = .93.

Mental Health Problems

Depression, anxiety, and PTSD were assessed using the Patient Health Questionnaire-9 (Kroenke & Spitzer, 2002), Generalized Anxiety Disorder Questionnaire (Spitzer et al., 2006), and PTSD Checklist—Civilian Version (Blanchard et al., 1996), respectively. Scores at or above the clinical cutoff for significant depression, anxiety, and PTSD were coded as 1, and those below were coded as 0. Internal consistency reliability for all mental health measures in the current study was adequate and ranged from α = .82 to α = .89.

Substance Use

The Alcohol Use Disorders Identification Test (Babor et al., 2001) was used to assess hazardous drinking, and the Drug Abuse Screening Test 10-item version (Yudko et al., 2007) was used to assess drug abuse in the current study. The internal consistency reliability for the current study was high for the Alcohol Use Disorders Identification Test at α = .84 and adequate for the Drug Abuse Screening Test at α = .74. Participants were also asked about lifetime use (yes/no) of alcohol, marijuana, cocaine/crack, hallucinogens, heroin, inhalants, sedatives, stimulants, tranquilizers, and prescription pain relievers without physician’s orders (PhenX Toolkit; Hamilton et al., 2011). We created a lifetime polydrug use variable that coded those who reported 2 or more substances in their lifetime as 1 and those who had not as 0.

Syndemic Index

We created an index of syndemic factors by dummy coding the 12 risk factors of HIV risk behaviors assessed and then summing the number of factors for a total index score. This method has been used in previous studies that examined the relationship between similar syndemics and HIV risk in MSM (e.g., Pantalone et al., 2018; Tsai & Burns, 2015). The 12 risk factors included five related to violence exposure (i.e., physical, emotional, and sexual abuse; IPV; experiencing and witnessing community violence); three relating to substance use (i.e., lifetime polydrug use, hazardous drinking, drug abuse); and three related to mental health problems (i.e., depression, anxiety, and PTSD symptoms). The three abuse subscale scores were dichotomized using the clinical cutoff scores, where 0 = none to low abuse and 1 = low to moderate or greater levels of abuse. The IPV and community violence variables were recoded as 0 = did not experience and 1 = did experience. Lifetime polydrug use was coded as 0 = has not used two or more drugs in one’s lifetime and 1 = used two or more drugs in one’s lifetime. Hazardous drinking, substance abuse, and the three mental health variables were recoded as 0 = score below clinical cutoff and 1 = at or above clinical cutoff. The index resulted in an actual range of scores of 0–11; however, for analyses, scores equal to or greater than 9 were collapsed due to small sizes for 10 cells (n = 3), 11 (n = 2), and 12 (n = 0).

Data Analysis

Frequencies were used to describe the extent to which YMSM in the sample engaged in HIV risk behaviors, were exposed to different forms of violence, used substances, and experienced mental health problems. We used bivariate analyses to examine correlations between HIV risk behaviors and violence exposure, substance misuse, and mental health problems, as well as to test for differences in outcome variables by ethnicity, sexual orientation, and recruitment city. Bivariate analyses included Spearman correlations, Wilcoxon rank sum, chi-square tests, and independent samples t-tests.

Negative binomial regressions were used for multivariable analyses in which the number of unprotected sex occasions was the dependent variable; we used truncated negative binomial regressions for multivariable analyses in which the number of male sex partners was the dependent variable. Truncated negative binomial regressions were used to account for the absence of zeros in the distribution of the male sex partner dependent variable. To improve model fit and reduce the likelihood of Type 1 errors, HIV risk behavior variables were Winsorized at the 95th percentile prior to computing negative binomial regressions (Barnett & Lewis, 1984).

For syndemic analyses, we regressed the syndemic index on each HIV risk behavior, controlling for ethnicity, to determine the effect of the syndemic index on HIV risk behavior for the full sample regardless of self-reported ethnic identity. In other multivariable models used to identify predictors of HIV risk behavior, all violence exposure, substance use, and mental health variables that were significantly associated with any HIV risk behavior (p < .05) at the bivariate level were included in all negative binomial regression models. Limiting the predictor variables to only those that were significantly associated with HIV risk behaviors at the bivariate level allowed us to reduce the number of independent variables in these models. This data reduction technique was important given the current study’s modest sample size. Using the same independent variables in all four negative binomial regression models also provided consistency across models and allowed comparison of significant predictors by HIV risk behavior.

Results

To determine if the two recruitment sites differed in participant demographics and key dependent variables, we conducted chi-square and independent samples t-tests. Results indicated that there were no significant differences according to ethnicity, employment, education level, income, sexual identity, or age. Likewise, results of a series of Wilcoxon rank sum tests revealed that the number of UAS occasions and number of male sexual partners in the last 3 and 12 months did not significantly differ by city. Thus, it was not necessary to include recruitment city as a control variable in our analyses.

Extent of Violence Exposure, Substance Use, and Mental Health Problems

A majority of the sample reported experiencing violence. Approximately 64% of the sample reported at least one type of childhood abuse at or above the low-to-moderate level. The most frequently reported form of abuse reported was emotional (48%), followed by physical (43%), and sexual (29%). Additionally, 58% reported experiencing some type of intimate partner violence (e.g., being hit or kicked by partner, being closely monitored or controlled by partner) in the last year, 35% experienced community violence, and 46% witnessed community violence. For each form of violence experienced, participants received one point on the syndemic index.

Regarding substance use, 36% of participants engaged in lifetime polydrug use, and 19% reported polydrug use in the last 30 days. The drug most used by participants was marijuana (69% lifetime, 49% last 30 days), followed by hallucinogens (20% lifetime, 7% last 30 days), cocaine (17% lifetime, 8% last 30 days), and other stimulants (16% lifetime, 8% last 30 days). Moreover, 35% scored at or above the clinical cutoff for disordered drinking, and 10% scored at or above the clinical cutoff for drug abuse. Finally, 39% of the sample reported at least mild depression symptoms, 32% reported at least mild anxiety symptoms, and 27% scored at or above the clinical cutoff for PTSD.

Syndemic Relationship With HIV Risk Behaviors

As shown in Table 2, most of the YMSM in the current study reported experiencing at least one syndemic factor, and only 7% (n = 11) reported never experiencing any of the syndemic factors. Results of negative binomial regressions indicated a significant relationship between the number of syndemic factors experienced and HIV sexual risk behaviors, with greater numbers of syndemic factors predicting higher frequency of UAS and a greater number of male sexual partners in the last 3 and 12 months. For each additional syndemic factor experienced, the incidence rate for the number of UAS occasions in the last 3 months increased approximately 1.12 times (incidence rate ratio [IRR] = 1.12, p < .05). Results were similar for UAS occasions in the last 12 months, for which the incidence rate also increased by 1.12 times for every increase in the number of syndemic factors experienced (IRR = 1.12, p < 05). Similarly, as the number of syndemic factors increased by one, the incidence rate for the number of male sex partners increased by 1.11 times in the last 3 months (IRR = 1.11, p < .01) and 1.15 times in the last 12 months (IRR = 1.15, p < .001). Table 3 provides a summary of the syndemic analysis results.

Table 2.

Frequencies of Number of Syndemic Factors Endorsed (N = 168)

Number of Syndemic Factors Endorsed	n	%
0	11	6.55
1	11	6.55
2	22	13.10
3	26	15.38
4	19	11.31
5	26	15.48
6	17	10.12
7	14	8.33
8	8	4.76
9+	14	8.33

Table 3.

Negative Binomial Regression of Number of Syndemic Factors of HIV Risk Behaviors

HIV Risk Behavior	Regression Coefficient	SE	95% CI	IRR	Goodness of Fit
UAS-3 months					χ2 (163, n = 166) = 161.087, p = .298
Number of syndemic factors	0.111	0.046	[.020, .202]	1.117*
Ethnicity	−0.513	0.225	[−.955, −.072]	0.598*
UAS-12 months					χ2 (163, n = 166) = 164.016, p = .461
Number of syndemic factors	0.112	0.054	[.006, .219]	1.119*
Ethnicity	−0.434	0.251	[−.926, .057]	0.648
Number of Sex Partners—3 months					N/A
Number of syndemic factors	0.105	0.035	[.036, .174]	1.111**
Ethnicity	−0.290	0.171	[−.625, .046]	0.748
Number of sex partners—12 months					N/A
Number of syndemic factors	0.142	0.040	[.064, .220]	1.152***
Ethnicity	−0.492	0.188	[−.861, −.123]	0.611**

Note. The goodness-of-fit statistic is not provided for truncated negative binomial regression models in SAS. CI = confidence interval; UAS= unprotected anal sex; IRR = incidence rate ratio.

^*p < .05.

^**p < .01.

^***p < .001.

Relationships Between Violence Exposure, Substance Use, Mental Health Problems, and HIV Risk Behaviors

Bivariate results showed significant, positive associations between UAS and experiencing IPV, polydrug use, and depression and anxiety symptoms. Participants’ number of male sex partners was also positively and significantly associated with the aforementioned variables, as well as hazardous drinking, drug abuse, and PTSD symptoms. Relationships between violence exposure variables, mental health problems, substance misuse, and HIV risk behaviors are presented in Table 4.

Table 4.

Spearman Correlations Between Syndemic Factors and HIV Risk Behaviors

Variable	UAS—3mo	UAS—12mo	Num. Prts.—3mo	Num. Prts.—12mo
Emotional abuse	0.04	0.06	0.01	0.03
Physical abuse	−0.01	0.01	−0.11	−0.05
Sexual abuse	−0.08	0.03	0.01	0.06
IPV	0.07	0.16*	0.01	0.01
CV-E	−0.02	0.02	0.08	0.02
CV-W	0.09	0.09	−0.01	0.02
Depression	0.01	0.15*	0.21**	0.26***
Anxiety	0.11	0.16*	0.14	0.19*
PTSD	0.10	0.13	0.27***	0.27***
Drug abuse	0.08	0.04	0.29***	0.28***
Polydrug use—lifetime	0.17*	0.23**	0.32***	0.37***
Drug abuse	0.08	0.04	0.29***	0.28***
Hazardous drinking	0.03	0.11	0.14	0.19*

Note. UAS = unprotected anal sex; mo = months; num. prts. = number of male sexual partners; IPV = intimate partner violence; CV-E = community violence experienced; CV-W = community violence witnessed; PTSD = posttraumatic stress disorder.

^*p < .05.

^**p < .01.

^***p < .001.

When compared with white YMSM, BMEB YMSM in this sample reported significantly less UAS in the last 3 months and significantly fewer male sex partners in the last 3 and 12 months. Additionally, a higher level of education was significantly associated with a greater number of male sex partners in the last 3 and 12 months. There were no significant differences in frequency of HIV risk behavior by income, employment, age, or sexual orientation. Results of bivariate associations between HIV risk behaviors and participant demographics are summarized in Table 5.

Table 5.

Spearman Correlations Between Demographics and HIV Risk Behaviors

Variable	UAS-3mo	UAS-12mo	Num. Prts.–3mo	Num. Prts.–12mo
Ethnic self-identity (0 = white, 1 = Black)	−0.16*	−0.13	−0.20*	−0.27***
Age	0.07	0.04	0.07	0.07
Sexual orientation (0 = bisexual/pan/queer; 1 = gay)	0.03	0.01	−0.00	0.06
Employment (0 = part time/unemployed; 1 = full time)	0.06	0.10	0.00	0.01
Education	−0.11	−0.09	0.22*	0.22*
Income	−0.07	−0.06	0.08	0.04
City (0 = St. Louis, MO; 1 = Kansas City, MO)	−0.04	−0.02	−0.03	−0.03

Note. UAS = unprotected anal sex; mo = months; num. prts. = number of male sexual partners.

^*p < .05.

^**p < .01.

^***p < .001.

All variables that were significantly associated at the bivariate level with the number of unprotected anal sex occasions and number of male sexual partners were included in four multivariable negative binomial regression models to determine which syndemic factors were the strongest predictors of HIV risk behaviors in this sample.

Predictors of Unprotected Anal Sex

Pearson chi-square goodness-of-fit test values for the models predicting UAS in the last 3 [χ²(156, N = 166) = 164:88, p = .53] and 12 months [χ²(156, N = 166) = 157:05, p = .46] were nonsignificant, indicating good model fit. Results of multivariable analyses showed that there were no significant predictors of UAS in the last 3 months. However, results indicated that participants who scored at or above the clinical cutoff for depression reported significantly higher rates of UAS in the last 12 months (IRR = 1.86, p = .05) than those with minimal or no depression. No other variables significantly predicted UAS in the last 12 months. A summary of all negative binomial regression results is found in Table 6.

Table 6.

Negative Binomial Regression: Syndemic Predictors of Unprotected Anal Sex in the Last 3 and 12 Months (N = 166)

HIV Risk Behavior	Independent Variables	Coefficient	SE	95% Cl	IRR	p-value	Goodness of Fit
UAS—3 months	IPV	0.163	0.249	[−0.325, 0.652]	1.178	.512	χ2 (156, n = 166) = 164.875, p = .528
	Depression	0.192	0.295	[−0.386, 0.770]	1.212	.515
	Anxiety	0.070	0.332	[−0.582, 0.722]	1.073	.833
	PTSD	0.383	0.325	[−0.253, 1.019]	1.466	.238
	Hazardous drinking	0.059	0.253	[−0.438, 0.555]	1.060	.817
	Drug abuse severity	−0.091	0.406	[−0.887, 0.704]	1.260	.822
	Polydrug use	0.231	0.289	[−0.335, 0.797]	0.913	.424
	Ethnicity*	−0.540	0.258	[−1.046, −0.033]	0.583	.038
	Education	−0.094	0.108	[−0.306, 0.117]	0.911	.388
UAS—12 months	IPV	0.429	0.270	[−0.100, 0.958]	1.536	.112	χ2 (156, n = 166) = 157.049, p = .463
	Depression*	0.620	0.317	[−0.001, 1.240]	1.858	.050
	Anxiety	−0.259	0.382	[−1.007, 0.489]	0.772	.497
	PTSD	0.252	0.371	[−0.476, 0.980]	1.287	.497
	Hazardous drinking	−0.095	0.283	[−0.650, 0.459]	0.909	.176
	Drug abuse severity	−0.378	0.443	[−1.245, 0.490]	0.685	.736
	Polydrug use	0.420	0.310	[−0.188, 1.027]	1.521	.394
	Ethnicity	0.584	0.177	[−1.132, 0.054]	0.582	.075
	Education	−0.137	0.118	[−0.368, 0.095]	0.872	.248

Note. CI = confidence interval; IRR = incidence rate ratio; UAS = unprotected anal sex; IPV = intimate partner violence; PTSD = posttraumatic stress disorder.

^*p < .05.

^**p < .01.

^***p < .001.

Predictors of Number of Male Sexual Partners

We conducted truncated negative binomial regression models to identify significant predictors of the number of male sexual partners. Because the SAS procedure used to compute these models (PROC FMM) does not provide goodness-of-fit statistics, they are not reported here. When controlling for all other variables in the models, lifetime polydrug use (IRR = 1:54, p = .02) and scoring at or above the clinical cutoff for drug abuse (IRR = 1.67, p = .03) were significantly associated with greater incidence rates of male sexual partners in the last 3 months. Those who reported using multiple drugs over their lifetime also had a significantly higher incidence rate of number of male sexual partners in the last 12 months than those who did not report using multiple drugs (IRR = 2:01, p = .001). Table 7 provides a summary of all truncated negative binomial regression results.

Table 7.

Truncated Negative Binomial Regression: Syndemic Predictors of the Number of Male Sex Partners in Last 3 and 12 Months (N = 168)

HIV Risk Behavior	Independent Variables	Coefficient	SE	95% CI	IRR	p-value
NMS—3 months	IPV	0.107	0.169	[−0.224, 0.437]	1.125	.528
	Depression	0.317	0.187	[−0.050, 0.684]	1.373	.091
	Anxiety	−0.073	0.204	[−0.472, 0.326]	0.929	.719
	PTSD	0.305	0.197	[−0.082, 0.692]	1.357	.122
	Hazardous drinking	−0.013	0.165	[−0.337, 0.310]	0.987	.936
	Drug abuse severity*	0.514	0.242	[0.040, 0.989]	1.673	.034
	Polydrug use*	0.433	0.180	[0.080, 0.786]	1.542	.016
	Ethnicity	−0.002	0.166	[−0.328, 0.324]	0.998	.990
	Education	0.081	0.065	[−0.046, 0.208]	1.084	.211
NMS—12 months	IPV	0.214	0.189	[−0.158, 0.585]	1.238	.259
	Depression	0.359	0.212	[−0.056, 0.774]	1.431	.090
	Anxiety	0.005	0.238	[−0.346, 0.471]	1.005	.984
	PTSD	0.283	0.218	[−0.145, 0.711]	1.327	.196
	Hazardous drinking	0.119	0.181	[−0.236, 0.473]	1.126	.513
	Drug abuse severity	0.322	0.296	[−0.257, 0.901]	1.380	.276
	Polydrug use**	0.698	0.201	[0.304, 1.091]	2.009	.001
	Ethnicity	−0.165	0.187	[−0.531, 0.201]	0.848	.376
	Education	0.111	0.073	[−0.032, 0.254]	1.118	.128

Note. CI = confidence interval; IRR = incidence rate ratio; IPV = intimate partner violence; NMS = number of male sex partners; PTSD = posttraumatic stress disorder.

^*p < .05.

^**p < .01.

^***p < .001.

Discussion

Consistent with previous studies, findings from our syndemic analysis indicate that the incidence of HIV risk behaviors increased as the number of syndemic factors experienced increased (Blashill et al., 2020; da Silva et al., 2020; Mustanski et al., 2007) among BMEB and white YMSM using HIV prevention services. These findings suggest that reducing or treating the negative consequences of these syndemic risk factors may lower HIV risk in this population.

Overall, findings from negative binomial regression models indicate that the significant predictors of HIV risk behaviors vary by type of behavior. Specifically, a greater number of depression symptoms was significantly associated with engaging in more frequent UAS, and lifetime polydrug use and drug abuse were significantly associated with a greater number of sex partners in YMSM who use prevention services. The unique significant risk factors for each HIV risk behavior have been little studied among YMSM. However, in one study of HIV risk behaviors among injection drug users in Russia, Abdala and colleagues (2012) found that two distinctive HIV risk behaviors—UAS and sex with multiple partners—were predicted by unique variables. Findings from the current study suggest a similar dynamic between the syndemic factors of mental health (i.e., depression) and substance use (i.e., polydrug use and drug abuse) and HIV risk behaviors in YMSM.

The current study found that YMSM using HIV prevention services who scored at or above the clinical cutoff for depression reported significantly more UAS occasions; this finding is consistent with results from previous research (Agnew-Bruneet al., 2019; Koblin et al., 2006). Depression may affect HIV risk behaviors in several ways. Individuals with depressive symptoms can experience decreased motivation to initiate or maintain healthy behaviors (Allgöwer et al., 2001), which may include condom or PrEP use when engaging in anal sex. Also, YMSM who are depressed may attempt to ameliorate their symptoms by seeking out unprotected sex, a behavior that some YMSM associate with increased feelings of pleasure and emotional intimacy (Golub et al., 2012). Moreover, longer periods of depression may exacerbate HIV risk through repeated use of risk behaviors to cope with symptoms or due to decreased motivation for self-care.

Additionally, in the current study, polydrug use and drug abuse predicted a greater number of male sexual partners. These findings align with previous research on predictors of multiple male sex partners (Halkitis et al., 2011) and other HIV risk behaviors in MSM (Halkitis et al., 2018; McCarty-Caplan et al., 2014). Polydrug use may indicate a high-risk personality profile that is typified by high levels of sensation seeking and negative self-perception and low levels of impulse control (Patterson et al., 2005). Alternatively, other scholars suggest that MSM engage in polydrug use to cope with loneliness or psychological distress (McCarty-Caplan et al., 2014). Finally, social and environmental context may play a substantial role in polydrug use and drug abuse. Historically, YMSM have congregated in physical (e.g., bars) and virtual (e.g., geosocial networking applications) venues where substance use is common and accepted. In this type of environment, YMSM may experiment with different drugs being used by other MSM in the environment, leading to risk behaviors such as condomless sex or sex with multiple partners (Halkitis et al., 2011).

An interesting finding of this study was that violence exposure (i.e., childhood abuse, IPV, and community-level violence) was not an independent significant predictor of HIV sexual risk behaviors in the bivariate and multivariable models. It is possible that violence exposure indirectly influences HIV sexual risk behaviors, with depression, drug abuse, or polydrug use serving as the pathways by which violence influences HIV risk behaviors. Another noteworthy finding is the rate at which the YMSM in this study reported experiencing different forms of violence. For example, 48% of our sample reported experiencing childhood emotional abuse, and 43% reported physical abuse as a child, compared to approximately 11% and 28% of the general population, respectively (CDC & Kaiser Permanente, 2016). It may be that YMSM who engage in HIV prevention services have more psychological and social needs than those who do not. Yet, the rate of multiple forms of violence experienced by the sample warrants further investigation and may have implications for violence-related trauma treatment as a strategy to reduce depression and subsequent HIV risk behaviors among YMSM.

Finally, findings from the current study align with previous research on HIV-negative or mostly HIV-negative samples of MSM (Eaton et al., 2010; Taylor et al., 2012) and indicate that white YMSM engage in significantly greater frequencies of HIV risk behaviors than their BMEB counterparts. Although this association only remained significant in the multivariable model predicting UAS in the last 3 months, this finding suggests that there may be differences in HIV risk factors between white and BMEB YMSM. A recent qualitative study by Quinn (2019) applied an intersectional framework to the study of syndemics among Black gay and bisexual men and found that systems of oppression (e.g., racism, homonegativity, social inequality) may be unique syndemic factors for Black YMSM that warrant additional investigation. Studying the unique syndemic factors at the intersection of ethnicity and sexual orientation may help prevention service providers tailor interventions for specific communities.

Limitations

Several methodological limitations should be considered when interpreting findings of the current study. First, a convenience sample was recruited for this study, which limits the external validity of the results. Although results may not be generalizable to the broader population of Black and white YMSM, they are likely representative of YMSM who access HIV prevention services at AIDS service organizations in midsized, Midwestern urban areas. Additionally, the study relied on self-report from participants to collect data on sensitive topics. The desire to provide socially desirable answers may have influenced the responses of some participants. However, data collection procedures were chosen to minimize social desirability bias (Gnambs & Kaspar, 2015). Another limitation of the study was the need to collapse Black and multiethnic Black participants into one group due to the small sample of multiethnic Black participants. The study was similarly limited by needing to collapse participants who self-identified as bisexual, queer, pansexual, and “other” into one group, also due to sample sizes for those groups. Lastly, we did not collect data regarding the type of sex participants engaged in with each of their male sexual partners, nor the positioning (insertive or receptive) of their anal sex encounters, limiting the specificity of risk regarding that sexual behavior. However, a count of total sex partners has been used as an HIV risk behavior outcome in national studies of MSM sexual risk behaviors (Shadaker et al., 2017).

Implications for Future Research and Practice

Despite the growing evidence that depression and substance use are significant predictors of HIV sexual risk behaviors for YMSM, the majority of HIV prevention services for YMSM are not designed to address these problems with HIV-negative clients. One potential strategy for addressing the prevention needs of these clients in the AIDS service setting would be to integrate the Screening, Brief Intervention, and Referral to Treatment (SBIRT) approach (Substance Abuse and Mental Health Services Administration, 2017) to assess and provide a brief intervention and referral for those who have or are at risk of developing depression or a substance use disorder. SBIRT has been successfully integrated into primary care settings (Savage & Sanchez, 2016) and with young people seeking rapid HIV testing in an emergency department (Edelman et al., 2012). Although SBIRT has been shown to be acceptable and efficacious for a variety of populations, it has not been well tested among HIV-negative YMSM. Future research can test the feasibility, acceptability, and efficacy of implementing an SBIRT approach that is inclusive of depressive symptoms and lifetime polydrug use in AIDS service organizations that deliver HIV testing and prevention services for YMSM.

Further research is needed to fully understand the syndemic factors of violence exposure, substance misuse, and mental health, as well as how they interact to influence HIV risk behaviors. First, future syndemic research on YMSM may wish to stratify results by ethnicity. As argued by Quinn (2019), reporting results of a racially diverse sample may obscure the importance of unique syndemic factors that occur at the intersection of ethnicity and sexual orientation. Stratifying results could lead to the development of culturally tailored interventions. Additionally, future studies are needed to examine pathways among these factors, including the potential indirect effect of exposure to different types of violence on HIV risk behaviors through substance use or mental health problems. Previous studies examining the relationship between childhood sexual abuse and HIV risk behaviors in adult MSM populations found a significant indirect relationship between experiencing abuse and greater HIV risk, with substance use identified as the significant mediating factor (Mimiaga et al., 2009; Purcell et al., 2004). Furthermore, potential mediating models in which mental health problems serve as the mediator from violence exposure to HIV risk are consistent with Agnew and Brezina’s (2019) general strain theory, which posits that the relationship between negative stress, such as violence exposure, and engaging in risky behaviors is mediated by negative emotional states, such as depression. Testing similar mediating models in the current sample is beyond the scope of this study but may be an important path to pursue in future research that could further explicate intervention targets.

Conclusion

This study adds to the literature by identifying specific syndemic factors that can be targeted to enhance HIV prevention efforts with HIV-negative YMSM. Moreover, it increases our understanding of the extent to which YMSM experience multiple forms of child maltreatment and other forms of violence, engage in the use of multiple substances, and experience mental health problems. Given the high frequency of violence exposure, a trauma-informed approach to HIV prevention (including HIV testing) may be indicated. Use of SBIRT may help to assess for and respond to mental health, substance use, and violence exposure in this population at high risk for HIV.