Small Area Estimation of HIV Prevalence using National Survey data: South Africa case study
Lwando O. Kondlo, Samuel Manda
Biostatistics, Medical Research Council, Pretoria, Gauteng, South Africa

National surveys for human immunodeficiency virus (HIV) provide the best available source of information at national level. These surveys do not, however, contain sufficient sample sizes to give reliable estimates by themselves of small areas. It is imperative to report the results of these surveys in small areas for more effective monitoring of local HIV prevalence. But, the reporting of HIV prevalence in South Africa has been reportedly been either at a national or at the level of province, but not at the local level/municipality.

This study deals with the determination of the HIV predictive model at the local level. The HIV predictive model is utilised to estimate HIV prevalence of young adult age 15-49 years, in the 257 municipalities using South African National HIV Prevalence, HIV incidence, Behaviour & Communication survey data. A number of spatially relevant covariate variables associations with HIV/AIDS were identified and used in the predictive model for HIV prevalence. Cross validation was used to determine the accuracy of the HIV predictive model. The refined HIV model (model fit diagnostics) was then applied in mapping of HIV prevalence to provide a good picture of the geographical variation of the epidemic that overall national HIV prevalence estimates are incapable to do.

The prevalence of HIV is high in more black African dominant areas. Urban or rural community has no effect on HIV. Population density has no effect either. Urban/rural and population density are being confounded with black African. There was a large local municipality variation in the prevalence of HIV in South Africa.

Understanding the variation of HIV infection within a country is also essential for determining where prevention and treatment programs need to focus resources; and to effectively target projects and monitor progress. Using small area techniques to predict local level HIV prevalence could be a useful tool for informing policies to achieve Millennium Development Goal 6.

Keywords: Small area estimation; Survey data; HIV prevelance

Biography: Lwando Kondlo has a Masters degree in Statistics. Obtained from the University of the Western Cape. His is currently a statistician at MRC-South Africa. His research interest includes: Measurement error models, small area estimation- sample size allocation & indirect method of estimating prevalence, HIV/AIDS studies