Gender-based violence and HIV sexual risk behavior: alcohol use and mental health problems as mediators among women in drinking venues, Cape Town

Abstract

Gender-based violence is a key determinant of HIV infection among women in South Africa as elsewhere. However, research has not examined potential mediating processes to explain the link between experiencing abuse and engaging in HIV sexual risk behavior. Previous studies suggest that alcohol use and mental health problems may explain how gender-based violence predicts sexual risk. In a prospective study, we examined whether lifetime history of gender-based violence indirectly affects future sexual risk behavior through alcohol use, depression and post-traumatic stress disorder (PTSD) in a high-risk socio-environmental context. We recruited a cohort of 560 women from alcohol drinking venues in a Cape Town, South African township. Participants completed computerized interviews at baseline and 4 months later. We tested prospective mediating associations between gender-based violence, alcohol use, depression, PTSD, and sexual risk behavior. There was a significant indirect effect of gender-based violence on sexual risk behavior through alcohol use, but not mental health problems. Women who were physically and sexually abused drank more, which in turn predicted more unprotected sex. We did not find a mediated relationship between alcohol use and sexual risk behavior through the experience of recent abuse or mental health problems. Alcohol use explains the link between gender-based violence and sexual risk behavior among women attending drinking venues in Cape Town, South Africa. Efforts to reduce HIV risk in South Africa by addressing gender-based violence must also address alcohol use.

Figure 1

Mediational model of gender-based violence predicting sexual risk behavior through alcohol use, CES-D, and PTSD. Note: ***p < .001; Indirect Effect = .02, 95% CI: .009 to .041, p <.01; Not displayed in this model are refrigerator in the home (Yes/No) and Time 1 % unprotected sex acts in the past month that were entered as control variables each directly predicting the outcome.

Figure 2

Alternative model where alcohol use also predicts sexual risk behavior through gender-based violence. Note: ***p < .001; **p < .01; *p < .05; Indirect effect of lifetime gender-based violence on % unprotected sex= .04, 95% CI: .02 to .06, p <.001; Indirect effect of alcohol use on % unprotected sex=0. Not displayed in this model are refrigerator in the home (Yes/No) and Time 1 % unprotected sex acts in the past month that were entered as control variables each directly predicting the outcome.