SEXUAL BEHAVIOR AND SEXUAL NETWORKS IN SOUTH AFRICA: IMPLICATIONS FOR HIV TRANSMISSION
Embargo until
2020-08-01
Date
2017-06-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Johns Hopkins University
Abstract
HIV is largely spread through sexual transmission in sub-Saharan Africa. Despite major biomedical innovations in HIV prevention, South Africa continues to bear a disproportionate burden of HIV. This dissertation aims to assess sexual behavior among people living with HIV comparing those on antiretroviral therapy (ART) to those not, describe sexual mixing patterns and number of sexual partners, and characterize sexual networks consistent with limited network data and assess the impact of network structure on disease potential. Throughout this dissertation, we present analyses of the Human Sciences Research Council’s 2012 South African National HIV Prevalence, Incidence and Behaviour Survey (SABSSM IV), a nationally representative household based cross- sectional survey.
We first use logistic regression to assess the relationship between ART and sexual behavior among those living with HIV. We find that ART is associated with increased odds of condom use among those living with HIV (aOR>2), but not associated with reporting multiple sexual partners. This aim suggests that people living with HIV not yet on ART in South Africa likely contribute the greatest number of transmissions (both due to sexual behavior and ART reducing infectivity), and reinforces the importance of engaging individuals living with HIV in care.
Next, we use mixing matrices and Newman’s assortativity coefficients to describe sexual mixing patterns, and fit a number of count distributions to degree distribution (number of sexual partners in the past year) data. Sexual mixing patterns are strongly assortative in South Africa, with assortativity coefficients for age, race, education, HIV- status, number of sexual partners and ARV status indicating strong assortativity in household partnerships (>0.6). Number of sexual partners was low (mean in past year = 1.34) but men were 5 times more likely to report 2+ partners in the past year. Our findings suggest that the strongly assortative nature of sexual networks in South Africa could have implications for HIV combination prevention intervention efficacy.
Finally, we use a nonparametric Markov chain Monte Carlo approach to simulate complete sexual networks consistent with mixing patterns and degree distribution data. We assess network characteristics on these consistent networks, and assess the impact of assumptions to balance male and female degree on these networks. We then estimate the impact of network structure on disease transmission. Simulated sexual networks consistent with our limited sexual network data varied little, but were highly dependent on assumptions made to balance male and female degree distributions. Networks with FSW populations had the greatest potential for HIV spread. Network characteristics were associated with potential HIV spread. Our results suggest the importance of capturing highly connected individuals in survey data, as these individuals will play a major role in disease transmission.
Sexual networks have the potential to dramatically influence the impact of HIV combination prevention interventions. This dissertation builds upon a body of work to provide an improved understanding of sexual behavior, sexual mixing and sexual networks within South Africa. These results can be utilized in the development of interventions to predict the potential effect an intervention could have in order to efficiently target interventions to have the greatest impact on HIV burden in South Africa.
Description
Keywords
sexual networks, sexual behavior, HIV, HIV transmission, sexual assortativity, sexual mixing