Evaluation of the modified HIV proximate determinant framework in Zimbabwe using statistical and geo-statistical methods
The HIV proximate determinant framework is a conceptual framework for the study of distribution and determinant of infection in population. A modification to this framework was developed by Till and Tanser taking into account the role of the community. This study aimed to evaluate the modified proximate determinant framework in Zimbabwe using statistical and geo-statistical methods.
METHODS:
This study used the Zimbabwe demographic and health survey of 2010/2011 which included a representative sample of 9171 women and 7104 men- aged 15 to 49. Following the proximate determinant framework, mixed effect binary logistic regression models were fitted. Spatial clustering was assessed using Kuldorrff spatial scan technique. Spatial and non-spatial random geoadditive Bayesian models were then fitted. Model fit was assessed using the deviance information criterion and Bayesian information criterion.
RESULTS:
Prominent significant underlying risk factors of HIV infection were age group and widowhood adjusting for proximate determinants. Total lifetime partners and symptoms of sexually transmitted infections emerged as significant proximate determinants of infection. Evidence of global and local spatial autocorrelation was found. Kulldorff spatial scan technique identified three hotspots including Matabeleland South province. No evidence of significant spatial heterogeneity was found with respect to underlying risk factors.
CONCLUSIONS:
A combined application of statistical and geo-statistical modeling procedures using the modified proximate determinant framework provides a robust approach to investigate HIV spread in generalized epidemic settings.