Spatial modelling of the relative risk of HIV/AIDS and associated risk factors in South Africa using integrated nested Laplace approximations

The coalescence of Human Immunodeficiency Virus (HIV), Tuberculosis (TB), and substance abuse in South Africa presents a multifaceted public health challenge. Individually catastrophic, these diseases create a synergistic relationship, intensifying morbidity and mortality rates. This study, therefore, aims to estimate HIV prevalence using TB risk factors, socioeconomic issues, and substance abuse in South Africa by leveraging Bayesian spatial modelling techniques. The study used Integrated Nested Laplace Approximation (INLA) to model the geographic distribution of HIV risk. Furthermore, a Local Indicator of Spatial Autocorrelation (LISA) technique was employed to identify the clustering pattern of the risk surface derived using INLA. Spatial predictors included condom use, the proportion of whites and blacks, tuberculosis cases, and needle sharing. We found a significant spatial disparity in HIV risk, with some districts exceeding an RR of 11. However, nearly 35% of districts had a reduced risk of RR <1. TB prevalence, being Black and sharing needles, drinking alcohol, and being White were not statistically significantly related to HIV risk, whereas having condomless sex statistically significantly increased HIV risk by 5.13%. The significant disparities in the spatial distribution of HIV risk suggest the need to develop geographically targetable interventions. Also, with condomless sex being the only important factor in HIV risk, we recommended a spatial intervention targeted at advocating and educating the public on safe sex practices.