Spatial clustering of "measured" and "unmeasured" risk factors for HIV infections in hyper-endemic communities in KwaZulu-Natal, South Africa: results from geoadditive models

Abstract

Sub-Saharan Africa contains more than 60% of all HIV infections worldwide. HIV prevalence was currently estimated to be at least 15% in KwaZulu-Natal and the epidemic is described as hyper-endemic. Knowledge of spatial clustering of risk factors which are linked to new HIV infections is important for prioritizing areas to change the trajectory of the epidemic. Geoadditive models were used to investigate spatial characteristics of the risk factors from two clinical trial units (Umkomaas and Botha's Hill) in the province of KwaZulu-Natal, South Africa. Study population was a cohort of women who screened and enrolled in an HIV prevention biomedical intervention trial. The results suggest high HIV incidence rates (5.8 and 8 per 100 person-year). Considerable spatial variations in behavioural factors within a relatively small geographical region, low level of education, early age at sexual debut, higher number of sexual partners, not being married/cohabitating with a sexual partner and sexual activity in exchange for money, gift and drugs were all determined to be clustered in certain regions; they were accounted for 25% (Umkomaas) and 65% (Botha's Hill) of the excess new HIV infections in two clinical trial units. Results from our study highlighted existence of significant spatial heterogeneity in "measured" and "unmeasured" risk factors in a relatively small region. As the HIV funding has been declining, identifying, targeting and reaching the most-at-risk individuals will likely play a significant role in developing the most efficient and cost-effective prevention programmes and subsequently will change the trajectory of the epidemic.

Keywords: HIV infection; South Africa; geoadditive models; hyper-endemic.