Second-generation HIV surveillance surveys collect biological and behavioral data from high-risk populations. Data from repeated surveillance surveys are important for monitoring trends in HIV and risk behaviors over time, demonstrating the effectiveness of HIV prevention efforts, and providing evidence of accountability to domestic and international HIV funders.1 This is particularly critical in light of “donor fatigue,”2 and the need to be able to lobby for scarce resources and use them effectively in the fight against HIV and AIDS.3,4 If countries most adversely affected by HIV and AIDS are to know how best to respond to the call to “know your epidemic,”5 community-level HIV biological and behavioral surveillance surveys (BBSS)1,6 are needed among subpopulations who engage in high-risk sexual behaviors within the general population. One such subpopulation is women who have multiple concurrent sexual partnerships—a high-risk sexual behavior thought to be a key driver of the HIV pandemic and which is receiving international attention and debate.7–12
Concurrency is generally defined as instances where an individual has 2 or more sexual relationships that overlap in time.13 The “concurrency theory” proposes that, particularly when an individual is newly infected with HIV compared with any other stage in disease progression, the probability of transmission of HIV to other members of the individual’s sexual network increases between 8- and 10-fold.14 At the population level, mathematical modeling has demonstrated that when any of the members of the sexual network are also engaging in multiple concurrent partnerships (MCP), they in turn place members of their other network at increased risk. The outcome is acceleration in the speed of HIV transmission facilitating the spread of HIV11,12,15 across interlinked sexual networks.16,17
Although women who have concurrent sexual partners might be easily accessed at antenatal clinics where regular HIV surveillance is routine, these surveys might be missing groups of women about which little is known, that is, young women and older women who are less likely to be sexually active, not of child-bearing age or who are not/have not been pregnant. Women who have MCP are diversely distributed within the general sexually active heterosexual population presenting challenges for recruiting them into community-level HIV BBSS. Respondent-driven sampling (RDS) has been used successfully among hard-to-reach, high-risk populations internationally18 and in South Africa19,20, and was chosen as a means to sample high-risk heterosexual women who have MCP for the study on which this article reports.
This article has several aims to describe the effectiveness of RDS to recruit women who have MCP, to report HIV prevalence and describe key characteristics among them, and to explore whether RDS-recruited women not accessed by routine sentinel surveillance and whether they were at differential risk for HIV infection.
From March to July 2011, we conducted a HIV BBSS using RDS among heterosexual women who reported having more than 1 sexual partner in the previous 3 months. The survey was conducted in a large urban settlement approximately 20 km south of Cape Town and characterized by high levels of unemployment and extreme poverty.21
Before the BBSS, we conducted formative research during January that comprised informal conversations with women from the target population. Information from these conversations was used to establish that women who had multiple sexual partners knew each other and were well networked. We were also able to establish an appropriate form and value for incentives and to identify initial recruits (seeds). Formative research also confirmed the existence of 3 types of male sexual partners: main partners (steady partners and husbands with whom women had ongoing regular sexual relationships), casual partners (partners with whom women had ongoing, but sporadic and secretive sexual relationships), and once-off partners (partners with whom women had sex once and never again).
Eligible women were 16–44 years of age, reported having 2 or more male sexual partners in the 3 months before the survey, and resided, worked, or socialized in the study community. The age restriction was applied because HIV prevalence rates rise rapidly and are at their highest among women in this age band.22
We implemented RDS using standard procedures.23 Seeds (initial participants) were purposively recruited by members of the research team who had established relationships with women while conducting formative research in the study community. Seeds and recruits were screened for eligibility, provided written informed consent, and completed an audio-computer–assisted survey interview (ACASI). Participants voluntarily provided dried blood spots (DBS) that were sent to a referral laboratory for HIV testing and were offered rapid HIV testing using 2 separate rapid tests simultaneously if they wished to know their HIV status. Women who tested HIV positive were referred by the HIV counselor to HIV-related services available in the study community. Seeds and recruits were provided in supermarket shopping vouchers to the value of R60 (±US$7.5) for completing ACASI and providing DBS and 3 recruitment coupons with which to recruit eligible women into the survey. Recruiters were instructed to recruit peers from their social networks who shared the same lifestyle as themselves. The recruitment process continued through a number of successive recruitment cycles or waves until the required sample size was reached. Unique recruitment coupon numbers linked recruiters to their recruits and DBS results to survey responses. Additional shopping vouchers to the value of R20 (±US$2.5) were given to participants for each recruit who successfully completed the survey.
The survey instrument comprised 125 questions. After a brief ACASI tutorial, the survey asked for demographic information, sexual history (sexual debut, attending antenatal, or family planning clinics), sexual health (symptoms of sexually transmitted infection [STI] in the previous 3 months such as pain when urinating, a smelly discharge from the vagina, and sores on the “private parts”), characteristics of and condom use with 3 types of partners, concurrency (having sex with another man/men while still being in a sexual relationship with a current partner), number of partners in the past 3 months, perceptions of faithfulness among male sexual partners, and alcohol and drug use (quantity of alcohol consumed on most occasions and illegal drug use in the past 3 months).
HIV prevalence among women attending antenatal clinics in 2007 in the study setting was 30.2%.24 Thus, estimating we would find a HIV prevalence of 30%, allowing for an error margin of 5%, and a design effect of 2.5, we calculated the required minimum sample size for HIV prevalence to be 756.
Evaluation of RDS
The number of recruitment waves required to reach equilibrium and network homophily indices was obtained from the freely available Respondent-Driven Sampling Analysis Tool 6.0 (RDSAT) (www.respondentdrivensampling.org). According to RDS theory, equilibrium is reached at the point where the hypothetical population proportions on key variables should not change by more than 2%, no matter how much the sample proportions change as a consequence of more individuals entering into the sample.25,26 Network homophily ranges between +1 (indicating exclusive preference for and recruitment from one’s own group) and −1 (indicating exclusive preference for and recruitment from outside of one’s own group)10,11. Network homophily indices close to 0 suggest that social ties among recruits and recruiters cross networks, thereby overcoming biases that solely in- or outgroup recruitment may have introduced.
HIV Prevalence and Characteristics of Women With MCP
Estimates of population proportions and 95% confidence intervals were calculated using RDSAT for HIV status, demographics, and risk behaviors. This software package generates sample weights to take into account the variations in participants’ network sizes (degree weight) and differential recruitment and homophily (recruitment weight). In the bivariate analyses, we estimated crude risk ratios of HIV status by all covariates separately (Table 1).
TABLE 1-a Sociodemog...Image Tools
Risk Among Women Who Had/Had Not Attended Public Health Clinic
TABLE 1-b Sociodemog...Image Tools
TABLE 1-c Sociodemog...Image Tools
In a bivariate analysis, we estimated crude risk ratios of attendance at either an antenatal or a family planning public health care clinic by all covariates separately and controlling for age (Table 2). The dependent variables were weighted with population weights generated by RDSAT 6.0. All odds ratios and corresponding P values were calculated using STATA, version 10.0.
Dried Blood Spots
DBS samples were sent to a referral laboratory for anonymous HIV testing where serum was eluted from samples and tested with a fourth-generation HIV enzyme-linked immunosorbent assay (Vironostika Uniform II plus 0). Initially, reactive samples were retested with a third-generation (antibody only) HIV enzyme-linked immunosorbent assay (SD Bioline). Samples that were reactive in both assays were reported as positive. Discordant samples were tested by Western blot (HIV1/2 Biorad).
Ethical clearance was obtained from the Research Ethics Committee, Faculty of Health Sciences, at the University of Cape Town.
Fifteen seeds were recruited during the 17 weekend days (ie, Saturdays and Sundays) of operation. Four seeds did not recruit any peers and 7 generated 50 or more recruits over a maximum of 19 waves (range 7–19). Four of 11 recruitment chains had a length of more than 10 waves and produced 62.2% (526 of 845) of recruits. Equilibrium on 17 key variables was reached between 2 and 5 waves of recruitment. Indices of homophily had a maximum of 0.193 and minimum of −0.057 indicating that there was little preference for either in- or outgroup recruiting.
We recruited 1092 women. Two hundred forty-five women were ineligible (190 had <2 sexual partners in the past 3 months, 42 did not fit the age criterion, and 13 had already participated in the survey). Our final sample comprised 845 women whose mean age was 23.9 years. Few women were married (6.4%); 49.6% lived in informal dwellings (corrugated iron shacks); 77% relied on family members, sexual partners, or government social grants for income; and 31.8% reported not have enough money for food. Almost a quarter (22.3%) reported sexual debut between 10 and 14 years. Close to 40% (39.6%) of women accepted our offer of HIV counseling and testing and received their HIV results on site. Women reported a range of 2–20 male sexual partners in the previous 3 months: 19% reported 4 or more partners in this time, 86% reported having MCP in the past 3 months, and 78.3% thought any of their most recent partners had other sexual partners. Condom use with most recent main partner at last sex was low (23.5%) compared with condom use at last sex with casual (51.8%) and once-off (42.1%) partners. Post hoc analysis revealed that rates of condom use at last sex among women who thought their main and casual partners were unfaithful and was not different compared with those who thought these partners were faithful. Almost half (49.1%) of women reported drinking 5 or more alcoholic drinks on the last occasion and 20.4% reported using any illegal drug in the past 3 months.
HIV prevalence was 28.8% (confidence interval 24.3 to 33.4). Being between 20 and 29 years was significantly related to HIV infection. Just over half (50.8%) of women reported a symptom of a STI in the previous 3 months.
Women who had not attended either an antenatal or a family planning public health clinic (10.1%) compared with those who had were more likely to be younger than 20 years, report sexual debut at 10–14 years, report a symptom of a STI, and had taken illegal drugs in the past 3 months. These associations remained statistically significant after controlling for age where appropriate.
If RDS is executed properly, and provided that certain analytical assumptions are met, it is able to generate valid probability-based population estimates and standard errors from the data. Results from weighted RDS-generated data are thus theoretically representative of the population of interest allowing conclusions to be drawn and recommendations to be made with confidence.
RDS was effective and feasible for recruitment of women who reported having 2 or more male sexual partners in the past 3 months. Women had little difficulty recruiting others from their social networks with whom they were well-enough acquainted to know they were part of the same target population. The fact that referral chains ranged between 7 and 19 waves indicate that our BBSS reached sociometric depth within the networks we sampled. All variables reached equilibrium within 2–6 recruitment waves and bias from the nonrandom selection of seeds was therefore theoretically eliminated. All variables’ homophily indices showed neither a tendency to ingroup nor outgroup preferences. Supermarket vouchers proved to be an acceptable form of incentive.
HIV Prevalence and Characteristics of Women With MCP
HIV prevalence among our sample of women was considerably higher than the national prevalence rate among females (13.6%).22 Whereas peak HIV prevalence among females in the national study occurred among women aged 25–29 years (32.7%)22, we found that peak prevalence occurred at age 20–24 years (31.9%) among our sample. Furthermore, HIV prevalence was almost 4 times higher among our sample of women aged 16–19 years (28.5%) compared with women in the national study (6.7%).22 Clearly, women in our study setting are exposed to HIV infection at an earlier age compared with women in South Africa. Given the high prevalence of HIV among our sample, it is not unexpected that very high levels of HIV-related risk behavior were reported. Yet inexplicably, we found no significant associations between HIV status and a range of demographic characteristics (besides age) and HIV-related risk behaviors. Furthermore, a post hoc multivariate regression analysis revealed no significant predictors of HIV infection when controlling for age. Because of the cross-sectional study design, we have no knowledge of when HIV infection occurred and we cannot assume that HIV infection was a result of risk behavior/s. However, this finding suggests that HIV infection is more likely a result of engaging in a complex amalgam of risk behaviors than any one or a few specific behavior/s.
Considerably more women in our study reported early sexual debut compared with the national study (27% vs 5.9%).22 Early sexual debut among women is related to greater risk for HIV because of a lesser likelihood of condom use27, greater risk of sexual abuse, and a greater likelihood of engaging in other HIV-related risk behaviors.28 The high rate of early sexual debut among our study’s women might also account for the earlier HIV peak prevalence among them and the significantly higher HIV prevalence among 16- to 19-year-olds compared with women in the national study.
A very high percentage (86%) of women in our sample reported MCP in the past 3 months. This is significantly higher than that reported by another study among women over the age of 18 years in our study setting (13.3%)10 and among women between 14 and 25 years who reported last partner concurrency in Cape Town (12.1%).29 Although these discrepant findings might be partly a result of the studies’ different eligibility criteria and/or the time period for reporting of MCP, of note is the high rate of suspected partner infidelity reported by our study participants (78%). Coupled with no difference in condom use among those women who suspected main or casual partner infidelity compared with those who did not, these findings suggest that there were many interlinked high-risk sexual networks among our participants increasing their vulnerability and exposure to HIV and AIDS.16 Furthermore, it has been suggested that the riskiest concurrent partnerships are those that occur when both partners engage in the behavior.8
Condom use at last sex with main partners was low. This has been reported in several studies where (mis)perceptions of faithfulness and trust, and thus, safety that inform noncondom use in main partner relationships is proposed as an explanation for low condom use with these partners.30,31 Although condom use with casual and once-off partners was higher than with main partners, there were still a substantial number of women who had not used condoms at last sex with these partners. Because little is known about once-off partners’ past and current HIV risk profile, low condom use with these, perhaps “riskiest” of all partners, places those who engage in them at increased risk of exposure to HIV and AIDS.
High levels of alcohol and drug use were reported by almost half of our sample of women. Whether in conjunction with sex or as a regular social activity, high levels of alcohol use have been associated with inconsistent and noncondom use,32,33 transactional sex where money and/or alcohol is exchanged for sex,34 and intimate partner violence and sexual coercion.35 Although there is a paucity of research on the association between noninjecting drug use and HIV in South Africa, drug use in sexual contexts was related to a number of HIV risks such as inconsistent condom use, a history of STI, and intimate partner violence.36
Risk Among Women Who Had/Had Not Attended Public Health Clinics
Ten percent of the sample comprised women who had not accessed an antenatal or family planning public health clinic, which suggests that through RDS, we were able to recruit a group of women who might be missed in other surveys among women. These women were more likely to be younger than 20 years, have sexual debut at an early age, and report symptoms of a STI and drug use. As these risk behaviors increase vulnerability to HIV (see below), these findings highlight the importance of health-seeking behaviors and efforts to increase attendance at health facilities where women, particularly those who are not in the prime child-bearing years between 20 and 29 years, can receive treatment and advice about sexual and reproductive health should be a priority in our study setting.
As with any survey that enquires about sexual behaviors and other sensitive issues, there is the potential for participants to under- or overreport these behaviors. We believe that accuracy of self-report for sexual and other sensitive behaviors was maximized though the use of ACASI.37
Having incentives available to participants in a context of extreme poverty might have caused women to misrepresent themselves to enroll in the study. We employed 2 strategies to minimize this possibility: we believe that we set the correct form and amount of incentives through intensive enquiries during formative research and we constantly changed our method of questioning women about having MCP, so that questions (and responses) could not be anticipated and communicated to recruits. Although we believe that this last strategy largely prevented people who did not have MCP enrolling, it did increase the number of ineligible participants.
It is evident from this study’s findings that the prevalence of HIV-related risk behaviors among women who have MCP is high. HIV prevention interventions would be well advised to focus on the spectrum of behaviors that put women at risk.
1. Diaz T, Garcia-Calleja JM, Ghys P, et al.. Advances and future directions in HIV surveillance in low- and middle income countries. Curr Opin HIV AIDS. 2009;4:253–259.
2. Médecins Sans Frontières. No Time to Quit: HIV/AIDS Treatment Gap Widening in Africa. Brussels: Médecins Sans Frontières; 2010.
3. Izazola-Licea JA, Wiegelmann J, Arán C, et al.. Financing the response to HIV in low-income and middle-income countries. J Acquir Immune Defic Syndr. 2009;52:S119–S126.
4. Rehle T, Lazzari S, Dallabetta G, et al.. (2004). Second-generation HIV surveillance: better data for decision-making. Bull World Health Organ. 2004;82:121–127.
5. UNAIDS. UNAIDS Annual Report. Knowing Your Epidemic. Geneva, Switzerland: UNAIDS; 2008
6. Magnani R, Sabin K, Saidel T, et al.. Review of sampling hard-to-reach and hidden populations for HIV surveillance. AIDS. 2005;19(suppl 2):S67–S72.
7. Epstein H. The mathematics of concurrent partnerships and HIV: a commentary on Lurie and Rosenthal, 2009. AIDS Behav. 2010;14:29–30.
8. Green EC, Mah TL, Ruark A, et al.. A framework of sexual partnerships: risks and implications for HIV prevention in Africa. Stud Fam Plann. 2009;40:63–70.
9. Lurie MN, Rosenthal S. Concurrent partnerships as a driver of the HIV epidemic in sub-Saharan Africa? The evidence is limited. AIDS Behav. 2010;14:17–24.
10. Mah T. Concurrent Sexual Partnerships and HIV/AIDS Among Youths in the Cape Metropolitan Area. Cape Town, South Africa: Centre for Social Science Research, University of Cape Town; 2008. Working Paper No. 226.
11. Mah T, Halperin DT. The evidence for the role of concurrent partnerships in Africa’s HIV epidemics: a response to Lurie and Rosenthal. AIDS Behav. 2010;14:25–28.
12. Morris M, Kretzschmar M. Concurrent partnerships and the spread of HIV. AIDS. 1997;11:641–648.
13. UNAIDS. Addressing Multiple and Concurrent Partnerships in Southern Africa: Developing Guidance for Bold Action. Geneva, Switzerland: UNAIDS; 2009.
14. Pilcher CD, Tien HS, Eron JJ, et al.. Brief but efficient: acute HIV infection and the sexual transmission of HIV. J Infect Dis. 2004;189:1785–1792.
15. Halperin D, Epstein H. Why is HIV prevalence so severe in southern Africa? The role of multiple concurrent partnerships and lack of male circumcision: implications for AIDS prevention. S Afr J HIV Med. 2007;26:19–25.
16. Halperin D, Epstein H. Concurrent sexual partnerships help to explain Africa’s high HIV prevalence: implications for prevention. Lancet. 2004;364:4–6.
17. AIDS Support and Technical Assistance Resources Project. Addressing Multiple and Concurrent Sexual Partnerships in Generalized Epidemics. Washington, DC: USAID; 2009.
18. Malekinejad M, Johnston LG, Kendall C, et al.. Using respondent-driven sampling methodology for HIV biological and behavioral surveillance in international settings: a systematic review. AIDS Behav. 2008; 12: S105–S130.
19. Chopra M, Townsend L, Johnston L, et al.. Estimating HIV prevalence and risk behaviors among high-risk heterosexual men with multiple sex partners: use of respondent-driven sampling. J Acquir Immune Defic Syndr. 2009;51:70–75.
20. Townsend L, Johnston LG, Flisher AJ, et al.. Effectiveness of respondent-driven sampling to recruit high risk heterosexual men who have multiple female sexual partners: differences in HIV prevalence and sexual risk behaviors measured at two time points. AIDS Behav. 2010;14:1330–1339.
21. Department of Social Services and Poverty Alleviation. The Population Register Update: Khayelitsha. Pretoria, South Africa: Department of Social Services and Poverty Alleviation; 2006.
22. Shisana O, Rehle T, Simbayi LC, et al.. South African National HIV Prevalence, Incidence, Behaviour and Communication Survey, 2008: A Turning Tide Among Teenagers? Cape Town, South Africa: HSRC Press; 2009.
25. Heckathorn D. Respondent driven sampling: a new approach to the study of hidden populations. Soc Probl. 1997;44:174–199.
26. Heckathorn D. Respondent driven sampling II: deriving valid population estimates from Chain-Referral samples of hidden populations. Soc Probl. 2002;49:11–34.
27. Baumgartner JL, Geary Waszak C, Tucker H, et al.. The influence of early sexual debut and sexual violence on adolescent pregnancy: a matched case-control study in Jamaica. Int Perspect Sex Reprod Health. 2009;35:21–28.
28. Geary Waszak C, Baumgartner JL, Tucker H, et al.. Early sexual debut, sexual violence, and sexual risk-taking among pregnant adolescents and their peers in Jamaica and Uganda. Durham, North Carolina: Family Health International Working Paper #8; 2008.
29. Kenyon C, Boulle A, Badri M, et al.. “I don’t use a condom (with my regular partner) because I know that I’m faithful, but with everyone else I do”: the cultural and socioeconomic determinants of sexual partner concurrency in young South Africans. J des Aspects Sociaux du VIH/SIDA. 2010;7:34–43.
30. McPhail C, Campbell C. “I think condoms are good, but, aai I hate those things”: condom use among adolescents and young people in a southern African township, South Africa. Soc Sci Med. 2001;52:1613–1627.
31. Richards JE, Risser JM, Padgett PM, et al.. Condom use among high-risk heterosexual women with concurrent sexual partnerships, Houston, Texas, USA. Int J STD AIDS. 2008;19:768–771.
32. Kalichman SC, Simbayi LC, Kaufman M, et al.. Alcohol use and sexual risks for HIV/AIDS in sub-Saharan Africa: a systematic review of empirical findings. Prev Sci. 2007;8:141–151.
33. Simbayi LC, Kalichman SC, Jooste S, et al.. HIV/AIDS risks among South African men who report sexually assaulting men. Am J Health Behav. 2006;30:158–166.
34. Townsend L, Ragnarsson A, Mathews C, et al.. “Taking care of business”: alcohol as currency in transactional sexual relationships among players in Cape Town, South Africa. Qual Health Res. 2011;21:41–50.
35. Zablotska IB, Gray RH, Serwadda D, et al.. Alcohol use before sex and HIV acquisition: a longitudinal study in Rakai, Uganda. AIDS. 2006;20:1191–1196.
36. Peltzer K, Simbayi L, Kalichman S, et al.. Drug use and HIV risk behaviour in three urban South African communities. J Soc Sci. 2009;18:143–149.
37. Gribble JN, Miller HG, Cooley PC, et al.. The impact of T-ACASI interviewing on reported drug use among men who have sex with men. Subst Use Misuse. 2000;35:869–890.
© 2013 Lippincott Williams & Wilkins, Inc.