Racial differences in the accuracy of perceived partner HIV status among men who have sex with men (MSM) in Atlanta, Georgia
We compared perceptions of partner HIV status to HIV test results in a cross-sectional study of sexual networks of men who have sex with men (MSM) in Atlanta. We then examined differences between black and white MSM in the predictive value of perceived partner status. We recruited men (“seeds”) using time-space venue sampling. These seeds then referred up to three partners, who could also refer partners. All participants reported sexual behavior and HIV status for recent partners and received HIV tests. For partners who enrolled, we compared laboratory diagnoses to their partner’s perception of their status. Black MSM who perceived themselves to be HIV negative were more likely than perceived-negative white MSM to have a positive partner among those they perceived to be HIV negative or whose status was unknown to them (OR=6.6). Furthermore, although frequency of unprotected anal intercourse (UAI) was similar by race, black men were more likely to have had UAI with an unknown-positive partner (OR=9.3).
Keywords HIV, men who have sex with men, health disparities, sexual networks, serosorting
In the United States, HIV continues to disproportionately affect men who have sex with men (MSM). Despite being less than 5% of the population,1 MSM accounted for 63% of new HIV infections and 78% of new infections among men, from 2007 to 2010.2 In addition, racial disparities among this group are profound; young black MSM, aged 13 to 24 years, represented a majority of new HIV infections among MSM in the United States during the same period.2 The disparity in Atlanta, Georgia, is particularly stark. HIV prevalence among black MSM in a recent cohort study was 43.4% compared with only 13.2% of white MSM.3 Among MSM aged 18 to 24 years, prevalence was 30.1% and 5.6%, respectively.3
Despite disparities in HIV incidence, researchers have found few behavioral differences between black and white MSM.4–7 However, HIV-positive black MSM may be less likely to know their HIV status compared to white men,4,5,8 and serodisclosure is less common among those who have been diagnosed HIV-positive.4,9Among men who previously tested HIV negative but who have not had a recent test, incorrect serodisclosure may also result in a greater proportion of partners of black MSM misperceiving them to be HIV negative.8
Testing and disclosure have an impact on HIV risk. Specifically, the effectiveness of serosorting and seroadaptation, or selecting sexual partners and sexual activities or roles based on the HIV status of one’s partner, depends on accurate knowledge of each partner’s serostatus.10 Although some researchers have cautioned against serosorting,11,12 studies and models have indicated that it can be an effective method of reducing HIV transmission compared to having unprotected anal intercourse (UAI) with men of unknown or discordant statuses.13,14 To make it a viable option for risk reduction, individuals must know their own HIV status, accurately disclose their status, and learn their partner’s status before making a decision about which sexual activities to engage in (if any), which role to take (if engaging in anal sex), and whether to use condoms.10
Although serosorting behavior appears to be equally common among both white and black MSM,9 findings from some studies indicate racial differences in its effectiveness in preventing HIV, with serosorting being less protective among black MSM than among white MSM.4,15 Reasons for the difference in the protective effect of serosorting may be due to different rates of HIV disclosure among white and black MSM4 or inaccurate knowledge of HIV status.8,15 Although researchers have examined the differences in rates of testing,15–20 serodisclosure,9,21–23 and inaccurate serodisclosure,8 little is known about the proportion of partners for whom perceived HIV status is incorrect, or the predictive value of perceived partner HIV status, among black and white MSM. In this article, we examine the accuracy of perceived partner HIV status at last sex among a sample of black and white MSM in Atlanta, Georgia, using HIV test results from a sexual network study.
The Men’s Atlanta Networks (MAN) Project was a cross-sectional study of sexual networks among MSM in Atlanta, Georgia, that was conducted from April 2011 to March 2013. We recruited participants using a chain referral sampling method.24 “Seed” participants referred their recent male partners, who could then enroll and refer partners, as well. This method yielded chains of up to 5 recent sex partners. Seeds were recruited using time–space venue sampling.25 The venue sampling frame was based upon that used for the second wave of the National HIV Behavioral Surveillance System (NHBS) and another cohort study of MSM conducted by the Emory University team.3,25
Seeds were eligible if they resided in the Atlanta metropolitan area, had sex with a man in the 3 months prior to the survey, were non-Hispanic black or white, and were aged between 18 and 40 years. Each seed could refer up to 3 recent male sexual partners, defined as men with whom they had sex in the previous 12 months; preference was given to partners within the previous 6 months, if available. Referred partners were eligible if they were aged ≥18 years, lived in the Atlanta metropolitan area, and had sex with the participant who referred them; partners could be of any race and ethnicity to accurately reflect the networks of black and white MSM. This study was approved by the Emory University Institutional Review Board (00047855).
Following informed, written consent, participants received tests for HIV, gonorrhea, chlamydia, and syphilis, and they completed a self-administered computer questionnaire. The survey questions included demographic information and dyadic data collection that assessed sexual behavior and partner characteristics for up to 10 sexual partners from the 12 months prior to enrollment. Individuals also reported their perceived HIV status prior to receiving their rapid HIV test result, and HIV tests were administered regardless of self-reported HIV status.
Participants provided the age, race/ethnicity, and perceived HIV status of each partner, if known. They also reported details about their last sexual encounter with the partner, including whether they had anal sex, and, if so, their anal sex roles and whether condoms were used. Questions about last sex also included whether it was in the context of group sex and if alcohol or drugs were used. The computer-assisted survey was implemented in SurveyGizmo v.2.6.
Participants also provided names and other descriptors of partners. During a referral session with the study counselor in which select survey responses were reviewed, participants referred eligible partners to the study from the pool of sexual partners described in the survey. Referred partners were contacted by the participant himself, and by study staff if desired, to assess interest in participation. Thus, participants’ dyadic survey data were able to be directly linked to the study data of referred partners who later enrolled. For enrolled partners, we were able to link a subset of their dyadic survey data back to the participants who referred them, using probabilistic name matching with LinkPlus26 and manual review. Using detailed partner and participant data (e.g. name, phone number, age, and e-mail address), LinkPlus produced a list of potential pairs and the probability that they were the same person. We reviewed each match and made our final determination based on how well the characteristics and information for each person in the pair agreed. These linked dyadic responses formed the basis of our analysis of perceived and actual partner HIV statuses.
We compared demographic characteristics of white and black participants using frequencies and chi-square tests or, in cases where the expected cell counts were <5, Fisher exact tests. We then focused on the set of reported partnerships where both members were enrolled as part of the chain referral, for which linked dyadic data were thus available.
We computed the joint distribution of participant-perceived and actual partner HIV statuses separately for black and white participants. We depicted this distribution using a tree diagram that first shows perceived HIV status and within each perceived status, the actual HIV statuses per HIV screening-depicted differences in perceived and laboratory-diagnosed partner HIV status.
We compared the odds of having an HIV-positive partner among black and white MSM, using odds ratios (ORs) and 95% confidence intervals (CIs), further stratifying by perceptions of partner’s HIV status (HIV negative, HIV positive, and HIV unknown) and whether they engaged in UAI at last sex.
Finally, we summarized the characteristics of participants, partners, and dyads, in which the partner’s HIV-positive serostatus was unknown to the participant, either because he was perceived to be HIV negative or his status was not disclosed. To focus this analysis on partnership with HIV transmission risk, we restricted the participant sample to those who perceived themselves to be HIV negative at the start of the study. Characteristics included age and screened HIV status of participant; age, perceived HIV status, and race and ethnicity of the partners; and whether they discuss HIV status, whether the partner was main or casual, whether they lived together, whether they planned to have sex again, whether they met online or off-line, and whether they used a condom at last sex. We then compared the proportion of participant-unknown partner seropositivity, participant characteristics, partner characteristics, and characteristics of the dyad using chi-square analyses or Fisher exact tests in the same manner described above. Data were analyzed using SAS v9.3.27