In the United States, an HIV prevalence rate of 21.6 of 100,000 is estimated for 13-year to 14-year olds, with an escalating rate in the slightly older adolescent, emerging, and young adult segments of the population: 37.9 (15-19 years old); 120.3 (20-24 years old); 239.4 (25-29 years old).1 Among men, the majority of cases are attributed to same-sex behavior.
Epidemiological findings further indicate that between 2001 and 2007, HIV infections have increased among young men who have sex with men (YMSM) in New York City, specifically among those men 13-29 years of age.2 Moreover, these data have shown that rates of infection between 2001 and 2007 have increased in greatest proportion among black/African American and Latino/Hispanic men who have sex with men (MSM); in 2006, these 2 groups accounted for more than 90% of new diagnoses among MSM under 20 years.2 A study of adult MSM in New York City found the HIV prevalence was 28.6%, with the proportion of black men reporting an HIV-positive serostatus (43.1%) nearly twice that of white MSM (19.7%) or Hispanic MSM (25.7%).3 On a national level between 2005 and 2008, the greatest increase in new HIV infections among MSM have occurred in YMSM4 and is manifested along the entire YMSM age spectrum including adolescence (ages 13-17), emerging adulthood (ages 18-24), and young adulthood (ages 25-29), developmental epochs informed by the work of Arnett.5,6
Paradoxically, Although HIV prevalence has been consistently higher among black and Latino YMSM, some studies have found potentially risky sex7-10 and drug-using behaviors7,9 to be comparable or higher in white YMSM than in other ethnic groups. Some have proposed that despite the elevated rates of risk-taking in white YMSM, the high rate of HIV infection among black MSM is due to the likelihood that these men engage in sex with partners of their own race where prevalence is higher.7,8 Among Asian and Pacific Islander YMSM, incidence of HIV remains relatively low.4
For YMSM, HIV prevalence and various risk factors for HIV increase significantly with age.11,12 Harawa et al.7 found that HIV prevalence increased from 1.7% among men 15-16 years old to 10.5% among participants 21-22 years old. Mustanski et al12 found that although age was not significant for sexual risk taking, the odds of having an HIV-positive serostatus increased with age. Further, Rosario et al13 detected in a sample 14-21 years that age was positively related to more sexual encounters in the prior 6 months. Sifakis et al8 demonstrated history of UAI to increase with age. Previous studies have explored the HIV prevalence and sexual risk-taking behaviors of YMSM, but few have explored these behaviors during formative adolescent years. This study seeks to add to the literature by describing the sexual risk taking behaviors of YMSM 13-29 years old with a particular emphasis on onset of sex and its potential impact on subsequent sexual activity. Specifically, we (1) describe the same-sex onset in a sample of YMSM ages 13-29; (2) consider the lifetime and recent sexual partnering of YMSM; (3) document episodic level sexual risk taking of YMSM with same-sex main and casual partners and consider these behaviors in relation to concordance of participant and partner age, race/ethnicity, and HIV status; and (4) delineate differences in all sexual behavior markers by developmental stage and race/ethnicity.
Procedures and Participants
“Project Desire” was a cross-sectional survey of a nonprobability sample of 580 gay, bisexual, and other YMSM in New York City, 13-29 years old. Eligibility criteria included being biologically male and male-identified, having had sex with a man, and being between 13-29 years old. Because of issues pertaining to parental consent, the data were collected in 2 phases; in the first phase for those 18-29 years and in the latter those 13-17 years.
Participants 18-29 years old were recruited over 90 days during the summer of 2008. Research staff employed an active sampling procedure at various diverse venues throughout the 5 boroughs, including several large community events, social venues, bars, dance clubs, and public spaces, parks, street corners. Different venues throughout the city were used as a means of diversifying the sample. Recruitment was completed over 75 hours across all 5 boroughs of New York City. For the 18-year-old to 29-year-old cohort, surveys and consent were conducted on touch-screen personal digital assistant devices in real time at the venues where they were recruited, using ForAllSurveys survey software. This system allows greater respondent privacy and removes barriers to honest responding, such as embarrassment, feedback from facial expressions of the interviewer, and other social influences.14
Recruitment of the 13-year-old to 17-year-old cohort took place in first quarter of 2009. These participants were recruited through active recruitment at lesbian, gay, and bisexual service providers, street recruitment, and passively through participant referral. Potential participants were provided with a palm card with information about the study and a phone number to call to be screened for eligibility. Eligible participants were required to complete both participant and parent/guardian consent, which they brought to the research center where they completed the survey.
All participants were provided a monetary incentive. The protocol was approved by the Institutional Review Boards of New York University and the New York City Department of Health and Mental Hygiene.
All data were self-reported by study participants and utilized a skip logic built into the administration of surveys.
Participants reported age, race/ethnicity, HIV status, and sexual orientation. We grouped participants into 3 age groups in accordance with the developmental literature5,6 as follows: 13-year to 17-year olds (adolescents), 18-year to 24-year olds (emerging adults), and 24-year to 29-year olds (young adults).
Sexual Onset With Male Partners
We asked participants whether they had engaged in any of the following behaviors with other men: mutual masturbation oral, receptiveanal, and insertive anal intercourse as the age at which they first engaged in these behaviors with men.
Lifetime and Recent Sexual Partners
Participants were asked to indicate the number of men with whom they have had sex in their lifetimes. Sex was defined as mutual masturbation, oral, or anal intercourse. Participants also were asked the number of different male partners (including main and causal partners) with whom they had sex in the 3 months before assessment (recent partners). One item also assessed the number of recent female recent partners with whom the participant had sex. With females, sex was defined as mutual masturbation, oral, vaginal, or anal intercourse.
Episodic Sexual Behavior With Main and Causal Male Partners
We gathered data for the most recent sexual episode with main and casual partners. Participants reported whether they had engaged in unprotected oral, receptive anal, and insertive anal intercourse and their sexual partner's HIV serostatus, race/ethnicity, and age.
Descriptive statistics and frequencies were calculated for the key variables, and χ2 tests of independence were undertaken to determine relations between age groups and demographic characteristics (race/ethnicity, sexual orientation, self-reported HIV status). We then examined the relations between age of sexual onset with other men by age group and race ethnicity using analysis of variance with post hoc comparisons and analysis of covariance when controlling for age of the participants. Similar analyses by race we have undertaken separately within each of the 3 age groups. Number of lifetime and recent male sexual partners by age and race/ethnicity were examined using nonparametric Kruskal-Wallis tests because these variables were highly skewed. For these analyses, a total of 21 cases were removed for those indicating no sexual activity with a man in their lifetimes. In addition, a small number of cases (n = 12) in the analyses for age at first sexual onset were removed for those reporting age 4 years or less as being likely response errors. Analysis of episodic level unprotected sexual behavior data in relation to age group and race/ethnicity were undertaken using χ2 tests of independence, and associations between participant and partner age were examined using Spearman rank correlations, whereas associations between participant and partner race/ethnicity and HIV status were undertaken using χ2 tests and computation of odds ratios.
We recruited 580 YMSM for the study. The data for 1 participant was deleted, as more than 50% of his responses were missing as were 21 cases of individuals reporting no sex with men in their lifetimes, yielding a final analytic sample of 558. Of these 558 men, 7.0% (n = 39) were adolescents, 60.4% (n = 337) were emerging adults, and 32.6% (n = 182) were young adults.
We detected a relation between age group and race [χ2 (8) = 18.94, P = 0.02] with a higher proportion of mixed/other men and a lower proportion of white and Asian/Pacific Islander (API) men in the adolescent cohort. Age group and sexual orientation also were related [χ2 (4) = 11.20, P = 0.02] with a higher proportion of adolescents identifying as bisexual and men in the 2 older age groups. Last, there was a relation between HIV status and age group [χ2 (4) = 12.33, P = 0.02], with men in the older age groups more likely to report a positive HIV status. Sample characteristics are shown further in Table 1.
Among the YMSM, sexual orientation was related to race/ethnicity [χ2 (8) = 17.29, P = 0.03], with white (81.3%, n = 87) and Latino (77.7%, n = 129), men more likely to identify as gay than bisexual than black (70.9%, n = 107) men. Self-reported HIV status differed by age group [χ2 (4) = 12.33, P = 0.02], with 11% (n = 20) young adults reporting a seropositive status as compared with 5.6% (n = 19) of emerging adults and 2.6% (n = 1) of the adolescents; and by race/ethnicity [χ2 (8) = 28.54, P < 0.001], with more black (11.3%, n = 17), Latino (7.7% n = 13), and mixed race/other men (11.1%, n = 7) indicating a seropositive status, 8% (n = 13) of Latino men and less than 2% of white (n = 2) and API (n = 1) men.
Onset of Sexual Behaviors
Both current age and race were found to be related to the onset of sexual behavior, with the former due to the enrollment criterion of sexual activity for the study, resulting in men within the adolescent group indicating a younger age of sexual onset in general and for oral intercourse (OI), receptive anal intercourse (RAI), and insertive anal intercourse. After controlling for current age (ANCOVA), the age of any sexual onset varied significantly by race [F (4, 537) = 3.19, P = .01]. This differences also was noted with regard to specific acts: as was the age at which participants first engaged in specific acts: OI [F (4, 525) = 3.40, P = 0.01], RAI [F (4, 433) = 4.62, P = 0.001], and IAI [F (4, 446) = 3.22, P = 0.01]. In all cases, black or Latino men reported an earlier onset than white or API men. No effect was detected for mutual masturbation. Means for age at first sex by race are shown in Table 2. In each of the analyses (age at first sex, OI, RAI, and IAI), participant age was a significant covariate (P < 0.001) indicative of both the adolescent and emerging adult groups reporting an earlier onset of sexual activity than the young adult group: any sex (14.05 vs. 14.35 vs. 16.21); mutual masturbation (14.50 vs. 14.79 vs. 16.56); OI (14.40 vs. 15.34 vs. 17.22); RAI (14.72 vs. 16.77 vs. 18.65); and IAI (15.16 vs. 17.08 vs 19.88)
Analyses of age of initiation of sexual activity by race were also conducted separately within each age group to overcome the potential impact of the observed relationship between age group and race (Figs. 1-3). Differences by race emerged within the adolescent emerging adult groups.
Within the adolescent group, significant differences by race remained for age at first sexual contact [F (4, 34) = 3.87, P = 0.01), OI [F (4, 34) = 4.11, P < 0.01), and RAI [F (4, 27) = 5.28, P < 0.003). In all 3 cases, black adolescents reported significantly later sexual initiation compared with Latino (15.0 years vs. 13.2 years, P = 0.001; OI: 15.5 years vs. 13.6 years, P < 0.001; RAI: 16.1 years vs. 14.1 years, P = 0.001) and mixed/other race (15.0 years vs. 13.7 years, P = 0.02; OI: 15.5 years vs. 14.0 years, P = 0.01; RAI: 16.1 years vs. 14.0 years, P < 0.001) adolescents. Black adolescents also reported first engaging in RAI at a significantly older age than white adolescents (16.1 years vs. 13 years, P = .03). Furthermore, in the adolescent sample, the reported age of first sexual activity was similar between black and API men (15.0 years vs. 15.5 years; OI: 15.5 years vs. 15.5 years; RAI: 16.1 years vs. 16.0 years).
Within the emerging adult group, significant (or marginally significant) differences by race also remained for age at first sex [F (4,324) = 2.25, P = 0.06], mutual masturbation [F (4, 287) = 2.32, P = .06], OI [F (4, 315) = 2.81, P = 0.03], and RAI [F (4, 256) = 3.44, P < 0.01]. Latino men reported engaging in these sexual activities at a significantly younger age compared with white men (13.9 years vs. 15.1 years, P = 0.02; mutual masturbation: 14.2 years vs. 15.4 years, P = 0.03; OI: 14.7 years vs. 15.8 years, P = 0.04), API men (13.9 years vs. 15.5 years, P = 0.02; mutual masturbation: 14.2 years vs. 15.5 years, P = 0.04; OI: 14.7 years vs. 16.5 years, P < 0.01), and mixed race men (mutual masturbation: 14.2 years vs. 15.7 years, P = 0.03; OI: 14.7 years vs. 16.1 years, P = 0.04). Both black and Latino men were significantly (or marginally significant) younger than white men (black: 16.4 years vs. 17.4 years, P = 0.07; Latino: 16.3 years vs. 17.4 years, P = 0.04). They were also younger than API men (black: 16.4 years vs. 18.3 years, P < 0.01; Latino: 16.3 years vs. 18.3 years, P < 0.01) when they first engaged in RAI.
Lifetime and Recent Partners
As anticipated, nonparametric analyses of lifetime number of male sexual partners showed significant differences by age group [Kruskal-Wallis χ2 (2) = 58.25, P < 0.001] with young adults reporting more lifetime partners (Md = 20) than either emerging adults (Md = 10, P < 0.001) or adolescents (Md = 6, P < 0.001). The number of lifetime male partners differed significantly by race (Kruskal-Wallis χ2 (4) = 10.16, P = .04) with white reporting Md = 20 as compared with Md = 11 for Latinos, and Md = 10 for all other groups, though this was most likely due to indirect effects of the association between age group and race, with a higher proportion of white men being older. Those with a gay identity reported more lifetime partners than those identifying as bisexual (43.36 vs. 12.59; F (2, 555) =6.34, P = 0.002).
The participants reported an average of 6.08 (SD = 15.20, Md = 2) total male partners (primary/main + casual) in the past 3 months, with 10.8% (n = 60) indicating no recent male sexual partners. Total recent male partners varied by age group [χ2 (2) = 11.17, P < 0.01] with young adults reporting significantly more partners than either emerging adults or adolescents (3 partners vs. 2 partners for each, P < 0.05 and P = 0.001, respectively). Bo differences were noted across sexual orinetaion.
In addition, the YMSM reported an average of 1.25 (SD = 4.43, Md = 0) female partners in the same time frame. The majority (77.8% n = 434) indicated no recent female partners. No differences were detected by age but a difference emerged for race/ethnicity [χ2 (4) = 11.24, P = 0.02]. Subanalyses indicated that black YMSM reported more female partners than Latino, white, API, and mixed race/other YMSM (all P < 0.05). On average black men reported 2.37 (SD = 6.44, Md = 0, 42% > group Md) female partners as compared with 1.20 (SD = 1.20, Med = 0, 14% > group Md) among Latinos and an average of less than 1 (Md = 0, 16% or less > group Md) for white, API, and mixed/race/other YMSM.
Episodic Sexual Behaviors
In the prior 3 months, 45.5% (n = 254) reported having a boyfriend. Of these 254 men, 53.1% (n = 135) described themselves as currently in this relationship at the time of assessment. Having a main male partner did not vary significantly by developmental group. We considered the sexual behaviors that took place during the most recent sexual encounter with a main partner. unprotected oral intercourse (UOI), unprotected receptive anal intercourse (URAI) and unprotected insertive anal intercourse (UIAI), and mutual masturbation did not vary significantly with age group or by race/ethnicity.
There was a strong positive correlation between the participants age and their partner's age (Spearman, r = .63, P < 0.001). This relation was also evident within race/ethnicity for black men (Spearman r = .54, P < 0.001), Latino men (Spearman r = .64, P < 0.001), and white men (Spearman r = 0.59, P < 0.001), but not API men (Spearman r = .27, P > 0.05). The participant's self-reported HIV status was related to his partner's HIV status with 89.3% (n = 176) of HIV-negative participants, 50.0% (n = 10) of HIV-positive participants, and 72.7% (n = 8) of unsure/never tested participants having seroconcordant partners [χ2 (4) = 110.71, P < 0.001].
Analyses also demonstrated that participants were generally likely to have a main partner who was a member of his own race/ethnicity [χ2 (16) = 159.35, P < 0.001]. Analyses excluded those men with mixed or other racial status. Of the 4 main groups with 70.1% (n = 54) of black men having black partners, 53.7% (n = 44) of Latino men having Latino partners, and 66.7% (n = 28) of white men having white partners; however only 34.8% (n = 8) of API men reported having API partners. A comparison of black men with main partners to non-black men (excluding mixed race/other) with main partners, indicated that black men were 2.38 [95% confidence interval (CI): 1.26 to 4.52] times more likely to report a same race partner, whereas API men were 3.33 times less likely (odds ratio = 0.30, 95% CI: 0.12 to 0.75) than non-API men to report a same race main partners. Unity was achieved in the comparisons of the white and Latino men with main partners.
At the time of assessment, 66.1% (n = 368) of the men had engaged in sex with a casual partner in the last 3 months. None of the unprotected sexual behaviors within the most recent episode (UOI, URAI, UIAI) varied significantly with age group. We conducted analyses of sexual behaviors by race/ethnicity and found that UOI was related to race/ethnicity [χ2 (4) = 9.93, P = 0.04]; white men (52.7%, n = 39) were more likely to engage in UOI as compared with black men [33.3%, n = 31; χ2 (1) = 6.35, P = 0.01]. A trend toward significance was achieved for URAI [χ2 (4) = 7.85 P = 0.09]. Subanalyses indicate a significant finding, with black men more likely than white men to report URAI with a casual partner [χ2 (1) = 3.77, P = 0.05], specifically, 20.4% (n = 19) of black men reported URAI with a casual partner, as compared with 9.5% (n = 7) of white men. Similarly, black men were more likely to report URAI than Latinos [χ2 (1) = 5.29, P = 0.02; 20.4% vs. 9.1%].
There was a positive association between participant age and partner age (Spearman r = 0.38, P < 0.001), although this relation is stronger among whites (Spearman r = 0.42, P < 0.01) and Latinos (Spearman r = 0.36, P < 0.001), than black men (Spearman r = 0.28, P = 0.01), and API men (Spearman's r = .33, P = .05). The participant's HIV status also was to his partner's HIV status [χ2 (4) = 22.62, P < 0.001], with 77% (n = 222) of HIV-negative men, 10.3% (n = 3) of HIV-positive men, and 57.0% (n = 9) of status unknown men choosing seroconcordant partners.
Analyses also demonstrated that participants were more likely to have a casual partner who was a member of his own race/ethnicity [χ2 (16) = 221.85, P < 0.001]. YMSM indicated a mixed or other race were excluded form analyses; of the others, 61.7% (n = 58) of black men having black partners, 46.5% (n = 53) of Latino men having Latino partners, and 77.3% (n = 58) of white men having white partners. Among API men, 31.7% (n = 13) reported API partners and 46.3% (n = 19) reported white partners. In terms of odds ratios, excluding those participants and partners of the mixed race/other category, black YMSM are 1.71 times more likely (95% CI: 1.01 to 2.60) than YMSM of other racial/ethnic groups to choose a same race casual sex partner, whereas whites are approximately 2.36 times more likely (95% CI: 1.29 to 4.33) to choose same race partners, and API men are approximately 4 times less likely (odds ratio = 0.28; 95% CI: 0.14 to 0.57) to choose same race partners; the odd ratio for Latinos approaches unity.
In recent years, epidemiological data in New York City and in other metropolitan areas across the United States have indicated an increase in the trajectory of new HIV infections in YMSM 13-29 years old and in particular among black and Latino YMSM.2-4 The findings of this cross-sectional behavioral investigation of 13-year-old to 29-year-old YMSM in New York City provides some insights into the current sexual behaviors of these young men and point to certain key risk patterns.
Although we recognize that our selection criteria may influence our findings on sexual onset, our results suggest that younger YMSM, grouped by developmental stages as adolescents and emerging adults, report sexual behaviors with another man at a younger age than the older YMSM in our cohort. Based on sheer probability, the onset of sexual behavior at a younger age provides a greater window of time for these young men to contract HIV. Across all of the behaviors we considered, adolescents reported a younger age of onset for sexual acts, including the most risky of HIV transmission behaviors, insertive and receptive anal intercourse.15 Of particular interest is our finding that young black and Latino men report onset of sexual behaviors with another man at a younger age than their white or API counterparts, which provides one possible contributing factor to these groups' disproportionate rise in new HIV diagnoses.
YMSM across all racial/ethnic groups reported and equivalent number of recent male sexual partners suggesting that sheer number of partners may be insufficient in explaining the rates of infection among black and Latino YMSM. Moreover, young adult YMSM reported more recent partners than their younger peers suggesting further that number of partners may fail to explain the rates of infection noted in adolescent and emerging adult black and Latino YMSM. The black YMSM in our sample indicated more recent female partners than any other racial/ethnic group (and a greater likelihood of identifying as bisexual), a finding which may provide a window to understanding the disproportionate prevalence of infection of black women as compared with all other women in the United States.1
Separate from behavior and with regard to sexual orientation, it should be noted that the YMSM identifying as gay reported more male sexual partners in their lifetimes than those who identified bisexual, but this relation was confounded by age, with older men more likely to gay identity. However, there were no differences between groups with regard to recent male sexual partners. A fully developed gay identity among YMSM emerges over time,16 may bestow protective effects in terms of well being,16 and moreover, a significant gay presence may lead to decreased sexual risk taking.17 However, the realization of these developmental points occur within a sociocultural context negative contexts could inhibit the formation of strong ties with other gay men and cause the YMSM to be at risk for depression and sexual risk taking.16 These ideas are indirectly supported by our work which demonstrate more risk taking among the black YMSM in our sample, who are more likely than whites to be in the adolescent group and also more likely to identify as bisexual than their white peers. However, the role of sexual identity formation as a protective factor for HIV is not fully established, and it has been noted the realization and disclosure of sexual identity at a younger age among MSM has been associated with an increased likelihood of depressive symptoms and an HIV-positive serostatus later in life.18
The episodic sexual behaviors of the participants in the study with male partners provide yet another avenue in which to consider recent epidemiological trends. Adolescent and emerging adult YMSM are as likely to engage in both unprotected insertive and receptive intercourse with both their main and casual partners as are older YMSM. Of note is the fact that black YMSM are more likely to engage in unprotected receptive anal intercourse with their causal partners than white YMSM, who are more likely to engage in oral intercourse with casual partners.
The findings regarding the episodic sexual behavior of the YMSM should be considered in light of the fact that, overwhelmingly, the participants in the study report casual partners who are of the same race/ethnicity and age. If YMSM of ethnic minority groups are engaging in higher risk sexual behaviors with their casual partners, and they are more likely to select partners of a concordant race/ethnicity and age where HIV prevalence is already high,2-4 then the likelihood of seroconversion increases substantially for YMSM of color. This is supported by the findings of our study, which indicate a much higher proportion of black and Latino men reporting a seropistive status than their white and API peers. Conversely, white YMSM tend to select white casual partners who have been shown to have a lower prevalence of HIV infection3 and are more likely to engage in oral sexual behavior with casual partners, which may be viewed a strategic behavior for harm reduction. For these men, the risk for HIV seroconversion may be less pronounced. Said differently, if black and Latino YMSM are choosing male sexual partners who are of the same age and of the same race/ethnicity, then they are engaging in sex with a subpopulation of men for whom the viremic pool may be more pronounced than for white YMSM who tend to choose from the subpopulation of white partners of the same age, where the viremic pool is less pronounced thus reducing their serconversion risk. This finding of racial matching aligns with a recent study of a broader age group of MSM19 and is complemented by the fact that white MSM demonstrate increased serconversion in the 30s and 40s,20 perhaps due to the fact that the viremic pool eventually becomes more concentrated in the white MSM population as these men age.
The findings of our investigation must be viewed with consideration of limitations. First, our sample of 13-year-old to 29-year-old YMSM may not be representative of other nonurban or exclusively gay-identified YMSM. Second, our study design actively recruited YMSM who were sexually active. Thus, the adolescents in our sample may not be representative of the larger population and findings regarding sexual behavior may be conflated with the fact that participation in the study required being sexually active with another man. Finally, this study was a cross-sectional investigation using self-report strategies of a nonprobability sample. To counter some of these design weaknesses, we gathered data using a computer assisted survey information (CASI) system, which has been shown to increase the accuracy of self-reported risk behaviors, and we gathered data from numerous venues across all geographic regions of New York City to ensure the diversity of our sample. Moreover, the use of episodic level data collection for sexual behaviors provides a level of confidence in terms of recall in comparison with aggregate level data.
The early sexual onset of YMSM, especially young men of color, suggest that HIV prevention strategies must be enacted at a younger age and with regard to the developmental needs of adolescents and emerging adults. Attention should also be paid to the cultural experiences of these men. Such strategies must be handled appropriately, as YMSM are not only developing sexually but they are also experiencing all of the demands that adolescents and emerging adults in the United States face. Innovative strategies for addressing HIV risk would seek to embed HIV prevention messages within the totality of these young men's experiences.
Finally, future directions for research should enact prospective cohort studies of YMSM as the develop through adolescence and young adulthood, and document the risk and factors that predispose YMSM to HIV serconversion and the protective factors that provide resilience and diminish HIV risk. Moreover, such exploration must more fully delineate the interactive effects of person and contextual factors and must not only consider patters across racial and/ethnic lines, but more importantly along socioeconomic lines, as poverty and health disparities may be more powerful in explaining HIV prevalence than race alone.21