Modeling the Impact of Social Discrimination and Financial Hardship on the Sexual Risk of HIV Among Latino and Black Men Who Have Sex With Men

Abstract

Objectives. We examined the impact of social discrimination and financial hardship on unprotected anal intercourse with a male sex partner of serodiscordant or unknown HIV status in the past 3 months among 1081 Latino and 1154 Black men who have sex with men (MSM; n = 2235) residing in Los Angeles County, California; New York, New York; and Philadelphia, Pennsylvania.

Methods. We administered HIV testing and a questionnaire assessing 6 explanatory variables. We combined traditional mediation analysis with the results of a path analysis to simultaneously examine the direct, indirect, and total effects of these variables on the outcome variable.

Results. Bivariate analysis showed that homophobia, racism, financial hardship, and lack of social support were associated with unprotected anal intercourse with a serodiscordant or sero-unknown partner. Path analysis determined that these relations were mediated by participation in risky sexual situations and lack of social support. However, paths between the explanatory variable and 2 mediating variables varied by participants’ serostatus.

Conclusions. Future prevention research and program designs should specifically address the differential impact of social discrimination and financial hardship on lack of social support and risky sexual situations among Latino and Black MSM.

Men who have sex with men (MSM) continue to make up the most disproportionately affected risk group for HIV/AIDS, representing more than half of all people living with HIV/AIDS and more than 60% of all new HIV infections in the United States.^1–3 HIV seroprevalence among Latino and Black MSM is particularly troubling, with estimates ranging from 17% to 46%.^4–6

Several theories have attempted to explain—at the individual level—racial and ethnic differences in HIV prevalence and infection rates among MSM.⁷ Although the HIV behavioral research literature has documented important associations among psychosocial variables, substance use, and sexual risk for HIV, little evidence has suggested a strictly behavioral basis for racial and ethnic disparities in HIV seroprevalence and seroincidence.^8,9

Several investigators^10–14 have begun examining macro- or social-level factors that may contribute to disparities in HIV rates among Latino and Black MSM. This important shift in focus away from individual-level and mainly behavioral correlates of sexual risk can ultimately expand available prevention opportunities to include those for which the locus of intervention is not the individual.¹⁵ For example, experiences of social discrimination (self-reported experiences of homophobia and racism) and financial hardship have been shown to be associated with heightened risk for HIV infection among Latino and Black MSM^16–18 and have led to interventions that range from social marketing initiatives to community mobilization efforts designed to educate the public and instigate social action.^19–22 However, few investigators have explored the precise psychosocial or situational mechanisms through which experiences of social discrimination affect the risk of HIV acquisition and transmission. One exception worth noting is Diaz et al.,¹⁷ who developed a 10-item scale that measured participation in contexts, settings, and situations that, according to Latino MSM focus-group participants, made it difficult to practice safe sex. In their study, difficult situations included sex in interpersonal relationships of unequal power, sex under the influence of drugs or alcohol, and sex in someone else's home. Mays et al.¹¹ pointed out the importance of possible links among social inequality, culture, social support, and interpersonal relationships in HIV-related risk. However, no previous study has examined mediating factors linking social discrimination and financial hardship with the sexual risk for HIV acquisition and transmission among HIV-positive and HIV-negative Latino and Black MSM across multiple large urban areas.²³

The Brothers y Hermanos research project was a multisite study funded by the Centers for Disease Control and Prevention to identify the sociocultural, psychosocial, and behavioral predictors of HIV infection among Latino and Black MSM. Using respondent-driven sampling techniques, we administered HIV testing and a survey questionnaire to assess experiences of racism, homophobia, financial hardship, lack of social support, situations that made using condoms during sex more difficult—such as exchange sex, having sex in a casual partner's home, or having a more masculine sex partner—and recent unprotected anal intercourse with a male sex partner of serodiscordant or unknown HIV status (hereinafter termed serodiscordant or unknown-status unprotected anal intercourse).

In this study, we examined the associations among experiences of social discrimination, financial hardship, and serodiscordant or unknown UAI among Latino and Black MSM living in Los Angeles County, California; New York, New York; and Philadelphia, Pennsylvania. Moreover, we assessed whether specific variables—namely, lack of social support and situations that make using condoms during sex more difficult—mediated these hypothesized associations. We examined whether

1. Experiences of social discrimination, financial hardship, and lack of social support would each be positively associated with serodiscordant or unknown-status unprotected anal intercourse;

2. Experiences of social discrimination and financial hardship would be positively associated with lack of social support;

3. Experiences of social discrimination, financial hardship, and lack of social support would each be associated with reports of being in situations that make using condoms during sex more difficult; and

4. Participation in risky sexual situations would mediate the associations between social discrimination, financial hardship, and lack of social support and serodiscordant or unknown-status unprotected anal intercourse.

METHODS

From May 2005 through April 2006, we recruited Latino men in Los Angeles County and New York and Black men in New York and Philadelphia for a cross-sectional interview study of adult male residents who reported having had sex with a man in the past 12 months. Participation was open to men who were HIV-positive, HIV-negative, or of unknown serostatus. Transgender men and women were not eligible to participate.

We used respondent-driven sampling, a form of chain-referral sampling, to recruit participants at each site.^24,25 Past studies have used respondent-driven sampling to enroll hard-to-reach or hidden populations such as injection drug users, ecstasy users, and Latino gay men.^26,27 A total of 2235 MSM enrolled in the study—516 Latino and 614 Black MSM in New York, 540 Black MSM in Philadelphia, and 565 Latino MSM in Los Angeles County. Detailed study methods are reported elsewhere.²⁸

Study Procedures

Enrollment and study implementation took place at several locations in each city. The choice of location was the participants’, and choices ranged from the administrative offices of local health departments implementing the study to partnering community-based organizations that served the target populations. After participants were screened for eligibility and their written informed consent was obtained, they completed an audio computer-assisted self-interview questionnaire. The questionnaire was available in English and Spanish versions and took approximately 45 minutes to complete.

All participants, except those who disclosed during screening that they had previously been diagnosed as HIV-positive, were then tested for HIV antibodies using a rapid or standard oral fluid HIV antibody test (OraQuick Advance; OraSure Technologies, Inc., Bethlehem, PA). All participants received pre- and posttest counseling.

At the end of data collection and the HIV testing session, each participant was offered an opportunity to recruit others into the study. Those who agreed were given a brief training on strategies for referring members of their social network to the study. Participants earned an additional $15 to $20 for each eligible person (up to 3) recruited. All participants, regardless of whether they agreed to be recruiters, were reimbursed $50 to cover any out-of-pocket expenses related to study participation. The study protocol was approved by the institutional review boards at the Centers for Disease Control and Prevention and at each study site.

Measures

Experiences of social discrimination.

We modified an existing scale to measure experiences of homophobia and racism in the study population.¹⁷ We used 5 questions to assess recent experiences of homophobia in which the participant felt uncomfortable or threatened because of his sexual orientation or gender nonconformity (e.g., “In the past 12 months, how often have you been made fun of or called names because people thought you were homosexual or not manly enough?” and “In the past 12 months, how often have you had to act more manly than usual in order to be accepted?”). Participants answered these items using 5-point response options: never (1), once (2), 2–3 times (3), 4–7 times (4), and 8 or more times (5). A factor analysis indicated that all 5 items loaded onto a single factor and had good internal consistency (Cronbach's α = .83). We obtained a total score by summing those items with higher scores indicating more experiences of homophobia.

We used another 5 questionnaire items to measure recent experiences of racism in which a participant felt uncomfortable or rejected because of his race or ethnicity (e.g., “In the past 12 months, how often have you thought that you have been treated rudely or unfairly because of your race or ethnicity?” and “In the past 12 months, how often have you thought that you have had trouble finding a male lover or boyfriend because of your race or ethnicity?”). Participants answered these items using the same 5-point response options as described for homophobia. A factor analysis indicated that all 5 items loaded onto a single factor with acceptable internal consistency (Cronbach's α = .73). We obtained a total score by summing the items with higher scores indicating more frequent experiences of racism.

Experiences of financial hardship.

We included 3 items in the questionnaire to measure participants’ socioeconomic status and level of financial hardship. We collected self-reported annual income, employment status, and information on how often a participant had run out of money (e.g., “During the past 12 months, how many times did you run out of money for your basic necessities like rent or food?”). We used a single item that asked how often a participant had run out of money because it reflected the most variation within the study sample. Response options for this item were never (1), once (2), twice (3), and 3 times or more (4).

Lack of social support.

We modified published scales to measure participants’ perceived availability of functional social support.²⁹ In a factor analysis, 5 items on the scale (i.e., “There is no one I can talk to about the important decisions in my life,” “I often feel isolated and alone,” “No one really understands my most private worries and fears,” “I feel no one respects who I am,” and “There is no one I can depend on to lend me $50 if I needed it for an emergency”) loaded onto a single factor describing a lack of emotional or informational support from others. Other items on the scale that measured the different dimensions of social support (i.e., tangible, positive social interaction, affection) were not as highly correlated with our dependent variable, serodiscordant or unknown-status unprotected anal intercourse. Participants responded on a 4-point scale ranging from strongly disagree (1) to strongly agree (4). We used the total score across the 5 items (Cronbach's α = .81), with higher scores indicating less social support.

Risky sexual situations.

We developed 5 questionnaire items based on the work by Diaz et al.¹⁷ in this area, which assessed men's involvement in situations hypothesized to make using condoms during sex difficult. These items assessed the occurrence during past 3 months of having anal sex in exchange for drugs, money, or a place to stay; using crystal methamphetamine, cocaine, crack, heroin, or ecstasy; having sex in a casual partner's home; having a partner who was more masculine than the participant; and using alcohol or drugs before or during sex.^30–34 We summed responses to these individual items so that a negative response was coded as 0 and an affirmative response was coded as 1. The scale total, which ranged from 0 to 5, was categorized into a 4-level continuous variable with the reported number of risky sexual situations expressed as none (0), 1 (1), 2 (2), and 3–5 (3).

Serodiscordant or unknown-status unprotected anal intercourse (dependent variable).

We assessed whether participants had engaged in any unprotected anal intercourse in the past 3 months with any male sex partners who were of serodiscordant or unknown HIV status. This dependent variable was used in a previous analysis of this study population to signify the highest risk for potential HIV acquisition and transmission.³⁵ We also restricted our analysis to participants who reported that they were either HIV-negative (n = 1013) or HIV-positive (n = 865) on the basis of their most recent HIV test results. Thus, we excluded 357 participants (16%) from the sample because they had not previously been tested, had not received their last test results, or had received an indeterminate test result. Participants were dichotomized as having had serodiscordant or unknown-status unprotected anal intercourse in the past 3 months if the participants (1) were HIV-negative and had unprotected anal intercourse with at least 1 partner whose serostatus was HIV-positive or unknown or (2) were HIV-positive and had unprotected anal intercourse with at least 1 partner whose serostatus was HIV-negative or unknown.

Data Analysis

To understand the associations between experiences of social discrimination and serodiscordant or unknown-status unprotected anal intercourse, we compared our results from a traditional mediation analysis approach³⁶ with results from a path analysis.³⁷ By using both techniques in tandem, we were able to corroborate our findings and to examine possible differences in our hypothesized model by racial and ethnic group more closely. In our traditional mediation analysis approach, we used χ² and t-test statistics to examine bivariate associations between our explanatory variables (social discrimination, lack of social support, risky sexual behavior) and our dependent variable, serodiscordant or unknown-status unprotected anal intercourse in the past 3 months. We then tested our hypothesized mediation model, using linear and logistic regression analyses to examine whether social discrimination, financial hardship, and lack of social support were associated with serodiscordant or unknown-status unprotected anal intercourse by way of participation in risky sexual situations.

We used MPlus version 5 software³⁸ to perform our path analysis, thus combining all the steps involved in the traditional mediation modeling approach to simultaneously examine the direct, indirect, and total effects of the explanatory variables on our dependent variable (serodiscordant or unknown-status unprotected anal intercourse in the past 3 months). Although drawing causal conclusions on the basis of these cross-sectional data is inappropriate, we established the theoretical order of the relations, or pathways, to the endogenous or dependent variable before analysis. The final model presents coefficients associated with the hypothesized path between each explanatory variable and the mediation and outcome variables. Statistically significant coefficients with P < .05 suggest support for a hypothesized path in the model.

We examined our hypotheses for Latino and Black MSM separately to allow for variation by racial or ethnic group in the mediation analysis. In addition, to examine whether participants’ perceived HIV status moderated the hypothesized relations, we stratified our path models by HIV status. Finally, we computed goodness-of-fit statistics to show how well the hypothesized paths fit the data for each racial and ethnic group, stratified by HIV status.

RESULTS

The study included 1081 Latino and 1154 Black MSM (n = 2235). Table 1 presents demographic and other characteristics of study participants. Fifty-seven percent of Latinos and 9% of Blacks were born outside of the United States. Latinos as a whole were younger than were Blacks (median age = 32 and 43 years, respectively), and more Latinos (58%) than Blacks (34%) reported that they were employed full or part time.

TABLE 1—

Characteristics of Latino and Black Men Who Have Sex With Men: The Brothers y Hermanos Study; Los Angeles County, CA; New York, NY; and Philadelphia, PA; 2005–2006

Characteristics	Latino MSM (n = 1081), No. (%)	Black MSM (n = 1154), No. (%)
Age, y
18–29	432 (40)	158 (14)
30–39	339 (31)	239 (21)
≥ 40	307 (29)	755 (65)
Highest education completed
< high school graduate	243 (22)	269 (23)
High school graduate or GED	461 (43)	555 (48)
Technical school graduate or associate's degree	204 (20)	215 (19)
≥ 4-y college degree	170 (15)	114 (10)
Employment
Full time	328 (30)	150 (13)
Part time	298 (28)	248 (22)
Unemployed	268 (25)	477 (41)
Unable to work (disabled)	172 (16)	257 (22)
Retired	13 (1)	18 (2)
Annual income before taxes, $
< 5000	317 (30)	396 (35)
5000–9999	232 (22)	269 (24)
10 000–19 999	218 (21)	229 (20)
20 000–29 999	128 (12)	111 (10)
≥ 30 000	152 (15)	122 (11)
Born in United States
Yes	460 (43)	1054 (91)
No	620 (57)	100 (9)
HIV status (based on testing done in study)
HIV-positive	416 (39)	608 (53)
HIV-negative	661 (61)	540 (47)
On antiretroviral therapy in past 3 mo (among HIV-positive or aware men)
Yes	256 (69)	313 (65)
No	113 (31)	169 (35)

Note. GED = general equivalency diploma; MSM = men who have sex with men. Because of missing data, column totals for each listed characteristic do not always equal the total sample size for each race/ethnic group. The sample size was n = 2235.

Both groups had similar distributions on education and annual income: 416 Latinos (39%) and 608 Blacks (53%) were HIV-positive, of whom two thirds reported receiving antiretroviral therapy in the past 3 months.

Bivariate Analysis

Table 2 presents scale means and percentages of Latino and Black MSM who reported experiencing social discrimination, financial hardship, or lack of social support or having participated in risky sexual situations by the outcome variable of interest. We found associations between each of the explanatory variables and the dependent variable. Men who engaged in serodiscordant or unknown-status unprotected anal intercourse reported higher mean scores for experiences of racism or homophobia and a higher frequency of financial hardship than men who did not. The only exception was Black MSM, for whom we found no difference in frequency of financial hardship between men who engaged in serodiscordant or unknown-status unprotected anal intercourse and men who did not.

TABLE 2—

Associations Among the Explanatory Variables and Unprotected Anal Intercourse With a Male Sex Partner of Serodiscordant or Unknown HIV Status for Men Who Have Sex With Men: The Brothers y Hermanos Study; Los Angeles County, CA; New York, NY; and Philadelphia, PA; 2005–2006

	Risky UAI Among Latino MSM			Risky UAI Among Black MSM
Explanatory Variables	Yes	No	Test statistic (P)	Yes	No	Test statistic (P)
Experiences of discrimination, mean
Homophobia (Cronbach's α = .83)	9.72	8.26	4.40 (<.001)	9.43	8.13	4.65 (<.001)
Racism (Cronbach's α = .73)	8.23	7.24	3.86 (<.001)	8.31	7.64	2.97 (.003)
Financial hardship: ran out of money for basic needs (single item), %	65	51	13.28 (<.001)	77	75	0.568 (.45)
Lack of social support (Cronbach's α = .81), mean	11.71	10.16	4.89 (<.001)	11.69	11.09	2.26 (.02)
Potentially risky sex situations (dichotomized and scale items)
Had anal sex for drugs, money, or a place to stay, %	24	8	39.96 (<.001)	42	10	153.10 (<.001)
Used illicit drugs in past 3 mo, %	41	23	26.05 (<.001)	52	40	14.01 (<.001)
Sex in someone else's home, %	33	22	10.94 (<.001)	40	21	42.56 (<.001)
Had a partner who was more masculine, %	35	22	16.41 (<.001)	38	17	59.87 (<.001)
Used alcohol or drugs before or during sex, %	46	17	78.28 (<.001)	63	30	100.16 (<.001)
Sum of potentially risky sexual situations (sum of 5 items)	1.78	0.93	9.72 (<.001)	2.37	1.18	14.11 (<.001)

Note. MSM = men who have sex with men; UAI = unprotected anal intercourse. Risky UAI = UAI with partner of serodiscordant or unknown HIV status in the past 3 months.

In addition, men engaging in serodiscordant or unknown-status unprotected anal intercourse reported greater lack of social support. They also reported participating more frequently in risky sexual situations than men who did not report serodiscordant or unknown-status unprotected anal intercourse.

Multivariate Analysis

Traditional mediation analysis.

We used traditional mediation analysis to test 3 of our hypothesized associations among Latino and Black MSM combined:

1. Experiences of social discrimination and financial hardship would predict lack of social support;

2. Experiences of social discrimination, financial hardship, and lack of social support would each predict participation in risky sexual situations; and

3. Potentially risky sexual situations would mediate the associations between social discrimination, financial hardship, and lack of social support and serodiscordant or unknown-status unprotected anal intercourse in the past 3 months.

To test our first hypothesis, we entered experiences of homophobia, racism, and financial hardship as predictors of lack of social support into a multiple linear regression equation. The results supported hypothesis 1; experiences of discrimination and financial hardship predicted lack of social support (homophobia: B = 0.50; P < .001, racism: B = 0.51; P = .001, financial hardship: B = 0.75; P < .001). To test the second hypothesis, we entered experiences of social discrimination, financial hardship, and lack of social support as predictors of participation in risky sexual situations. Lack of social support (B = 0.03; P < .001), together with experiences of homophobia (B = 0.12; P = .001), racism (B = 0.11; P = .006), and financial hardship (B = 0.14; P < .001) were positively associated with participation in potentially risky sexual situations. Finally, to test whether risky sexual situations mediated the associations among social discrimination, financial hardship, and lack of social support on serodiscordant or unknown-status unprotected anal intercourse in the past 3 months, we entered all variables into a logistic regression equation as predictors of the dependent variable. The addition of the hypothesized mediation variable (participation in risky sexual situations) to the full model reduced the direct associations of all of the explanatory variables except for homophobia. Our finding did not meet the criteria of Baron and Kenny³⁶ for complete mediation because experiences of homophobia retained a small direct effect on unprotected anal intercourse with a male partner of serodiscordant or unknown HIV status (odds ratio = 1.2; 95% confidence interval = 1.1, 1.4; P = .006). In addition to using a combined sample to test our 3 hypotheses, we also ran these regression models stratifying by racial and ethnic group and perceived HIV serostatus. We found some group differences and thus pursued our path analyses examining each racial and ethnic group separately and stratifying by perceived HIV serostatus (i.e., HIV-negative and HIV-positive).

Path analysis.

On the basis of the findings from the traditional mediation analysis, we examined a path model of serodiscordant or unknown-status unprotected anal intercourse for each of the racial and ethnic groups. Within each group, we compared HIV-positive and HIV-negative men by allowing each serostatus stratum to have a different path coefficient (Figures 1 and 2).

FIGURE 1—

Associations among social discrimination, financial hardship, and unprotected anal intercourse (UAI) with a male sex partner of serodiscordant or unknown HIV status among HIV-positive and HIV-negative Latino men who have sex with men: the Brothers y Hermanos study; Los Angeles County, CA; New York, NY; and Philadelphia, PA; 2005–2006.

Note. Path coefficients are listed for both HIV-positive (listed first) and HIV-negative (listed second) Latino men who have sex with men.

*P < .05.

FIGURE 2—

The associations among social discrimination, financial hardship, and unprotected anal intercourse (UAI) with a male sex partner of serodiscordant or unknown HIV status among HIV-positive and HIV-negative Black men who have sex with men: the Brothers y Hermanos study; Los Angeles County, CA; New York, NY; and Philadelphia, PA; 2005–2006.

Note. Path coefficients are listed for both HIV-positive (listed first) and HIV negative (listed second) Black men who have sex with men.

*P < .05.

The direct and indirect effects among experiences of social discrimination, financial hardship, lack of social support, participation in risky sexual situations, and discordant or unknown-status unprotected anal intercourse were invariant for Latino and Black MSM with 3 exceptions: (1) The path between homophobia and lack of social support was statistically significant for all groups but not for Latino HIV-positive men, (2) a statistically significant direct path remained between lack of social support and discordant or unknown-status unprotected anal intercourse for HIV-negative Latino MSM but not for other men, and (3) a statistically significant path remained between homophobia and discordant or unknown-status unprotected anal intercourse for HIV-positive Black MSM but not for other men.

Our path models showed 2 other differences between HIV-positive and HIV-negative men in each racial and ethnic group. First, we observed a statistically significant path between homophobia and risky sexual situations among Latino and Black HIV-positive MSM. This was not the case for HIV-negative men. Second, the path between racism and risky sexual situations was statistically significant for Latino and Black HIV-negative men but not for HIV-positive men.

When examining the goodness-of-fit statistics, we determined that the path models that best fit the observed data for both Latino (χ² test of model fit = 0.040; P = .001; root-mean-square error of approximation = 0.000; comparative fit index = 1.000) and Black MSM (χ² test of model fit = 1.068; df = 2; P = .586; root-mean-square error of approximation = 0.000; comparative fit index = 1.000).³⁹

DISCUSSION

We have documented the important role that social discrimination and financial hardship play in shaping the sexual risk for HIV acquisition and transmission among Latino and Black MSM. Using mediation analysis, we found that (1) experiences of social discrimination (racism and homophobia) and financial hardship were associated with lack of social support; (2) discrimination, financial hardship, and lack of social support were each associated with participation in risky sexual situations; and (3) risky sexual situations mediated the associations among discrimination, financial hardship, and lack of social support and recent serodiscordant or unknown-status unprotected anal intercourse for the combined sample of Latino and Black MSM. However, when we used a path analysis to model the impact of social discrimination on serodiscordant or unknown-status unprotected anal intercourse, all hypothesized associations were not equivalent for HIV-positive men and HIV-negative men. Although the overall model fit our data, we found 2 important differences between Latino and Black MSM in the impact of specific forms of discrimination on our outcome of interest. First, the impact of homophobia on serodiscordant or unknown-status unprotected anal intercourse was different for HIV-positive Black MSM compared with HIV-positive Latino MSM and appeared to be influenced only in part by their participation in risky sexual situations. After accounting for the mediating effects of risky sexual situations, a direct association remained between experiences of homophobia and discordant or unknown-status unprotected anal intercourse among HIV-positive Black MSM. However, the path between homophobia and discordant or unknown-status unprotected anal intercourse among HIV-positive Latino MSM was fully mediated by risky sexual situations.

Second, the association between lack of social support and discordant or unknown-status unprotected anal intercourse could be explained only in part by its association with risky sexual situations. For example, a direct association remained between lack of social support and serodiscordant or unknown-status unprotected anal intercourse only among HIV-negative Latino MSM.

Our research has in part corroborated past studies of social discrimination and HIV risk among MSM^40–42 by demonstrating the impact of homophobia on serodiscordant or unknown-status unprotected anal intercourse for both HIV-positive and HIV-negative Latino and Black MSM as mediated by lack of social support and participation in risky sexual situations. For example, social support may have moderated the deleterious effects of homophobia, racism, and financial hardship on risky sexual situations by encouraging the avoidance of such situations.¹³ However, how racism and financial hardship separately and differentially influenced the sexual risk for HIV acquisition and transmission among HIV-positive and HIV-negative Latino and Black MSM remains understudied and not well understood.²³ Some investigators have shown that race-based sexual stereotyping may result in segregation of sexual networks that in turn creates heightened risk for HIV transmission because of Black MSM's higher HIV seroprevalence.⁴⁰ More specifically, if overall HIV prevalence is higher in African American communities and Black MSM's sexual partnering choices are influenced by social and sexual segregation—thus limiting their choices of sexual partners to other African Americans—their risk for HIV exposure is potentially greater than that of men whose sexual partnering choices are not constrained by racial segregation. This point is supported by previous research that revealed differences in interracial versus intraracial sexual mixing between African Americans and Whites that may help to explain the disproportionate sexually transmitted infection rates among African Americans.^43,44 Alternatively, or in tandem, HIV-negative Black MSM's risk of acquiring HIV may be increased if race-based stereotypes create prescribed sexual roles during anal intercourse that compromise a man's capacity or intention to practice safer sex.⁴⁵

Finally, we may not have captured mediating factors in this study that might better explain the differential associations we found. For example, homophobia may influence risky sexual situations for HIV-positive men because of the stigma associated with HIV, which is at times conflated with the stigma associated with homosexuality. This stigma might present a particular challenge for HIV-positive men who may not publicly self-identify as gay or homosexual but for whom an HIV-positive serostatus may be interpreted as a gay signifier.⁴⁶ Similarly, lack of social support may be influenced by social network characteristics that may or may not reinforce a person's capacity to tactically weigh the risks and benefits of sexual choices and health choices.⁴⁷

Limitations

This study had several limitations that are important to note. First, the respondent-driven sampling method we used to recruit study participants may not have resulted in a representative sample of all Latino and Black MSM at the study sites. However, our goal for this analysis was to examine the hypothesized associations within these critically under-researched populations, and respondent-driven sampling allowed us to obtain a large sample of Latino and Black MSM that is among the largest and most diverse ever recruited in the United States. Second, we likely did not include all potential mediating variables when modeling the impact of social discrimination on the sexual risk for HIV acquisition and transmission. For example, research on the impact of racism on health often involves multilevel analyses that include structural-level or institutional factors that move beyond individually reported subjective experiences of discrimination.⁴⁸ Nevertheless, we included a range of variables not previously studied with both HIV-positive and HIV-negative Latino and Black MSM. Third, although we used audio computer-assisted self-interview data collection methods to reduce social desirability bias, the participants’ behavioral data were self-reported and therefore reliant on the participants’ recall and candor. Fourth, because these data are cross-sectional and subject to temporal ambiguity, we could not definitively determine the directionality of statistical associations presented. Last, some of the statistically significant paths in our model were weak. With large samples, finding weaker associations is easier. However, irrespective of how weak, reporting associations between distal and dependent variables is still important because weak associations may simply reflect the limitations of available measures to capture the health impact of social phenomena such as racism or homophobia.

Conclusions

Our results underscore the need for more research designed to reveal psychosocial and structural mechanisms that heighten the risk for HIV acquisition and transmission among Latino and Black MSM. Experiences of racism, homophobia, and financial hardship are important foci of research, especially given the disproportionate HIV disease burden for these groups. Our findings also highlighted the importance of examining cross-group differences among MSM in general and HIV-positive and HIV-negative Latino and Black MSM in particular. Such differences have important implications for prevention program design in terms of the specific cofactors (mediators and predictors) of risk they target for a particular racial or ethnic group. Our study showed that, for HIV-positive Black MSM, addressing homophobia is important; for HIV-negative Latino MSM, addressing lack of social support is important. Our data, coupled with findings from prior research, illustrate more importantly why a one-size-fits-all approach to complex social issues is inappropriate and suggest the need to address racism, homophobia, and financial hardship as part of HIV prevention strategies differentially tailored for HIV-positive and HIV-negative Latino and Black MSM. The effects of discrimination may be buffered by (1) creating safe spaces in which HIV-positive and HIV-negative Latino and Black MSM can critically explore and discuss the impact of social discrimination on their physical and mental health;⁴⁹ (2) enhancing feelings of belonging and social support, particularly from family and friends;⁵⁰ (3) ensuring access to culturally competent and linguistically appropriate sexual health services, including HIV prevention information and education, care, and treatment; and (4) providing opportunities for social action, including volunteerism and activism.⁵¹