Socio-behavioural characteristics and HIV: findings from a causal modelling analysis of 29 sub-Saharan African countries

Background: Socio-behavioural factors may contribute to the wide variance in HIV prevalence between and within sub-Saharan African (SSA) countries. Methods: We used Bayesian network models to explore socio-behavioural variables that may be causally related to HIV. A Bayesian network consists of nodes and edges, where nodes represent variables, and edges represent the conditional dependencies between them. We populated our models with data from Demographic and Health Surveys conducted in 29 SSA counties in 2010 or later. We predefined and dichotomized twelve variables that could be related to HIV, including age, literacy, HIV knowledge and testing, domestic violence, sexual activity, and women9s empowerment. We first analysed data on men and women for each country separately and then aggregated the results. Findings: We analysed data from 190,273 men (range across countries 2,295-17,359) and 420,198 women (6,621-38,948). The two variables with the highest total number of edges were literacy and rural/urban location. Literacy was negatively associated with false beliefs about AIDS (women 24, men 27 countries) and, for women, early sexual initiation (21 countries). Literacy was also positively associated with ever being tested for HIV (women 8, men 22 countries) and the belief that women have the right to ask their husband to use condoms if he has a sexually transmitted infection (women 12, men 13 countries). Rural residence was positively associated with false beliefs about AIDS (women and men 14 countries) and the belief that beating one9s wife is justified (women 16, men 9 countries), and negatively associated with having been tested for HIV (women 12, men 11 countries). Interpretation: Literacy and urbanity were strongly associated with several factors that are important for HIV acquisition. Funding: This project was funded by the Swiss National Science Foundation (grant no. 163878)