A simplified approach for estimating prevalence from mortality by HIV sero-status in rural South Africa

Makandwe Nyirenda, University of Kwazulu-Natal
Till Barnighausen, University of Kwazulu-Natal
Marie-Louise Newell, University of Kwazulu-Natal and African Population and Health Research Center (APHRC)

Of concern in voluntary surveys is the potential size of the non-response rate and its effect in biasing survey findings. Data from a longitudinal surveillance study in rural South Africa found mortality among HIV positive individuals was over ten times higher than among HIV negative individuals (Nyirenda et al. 2007). Mortality among those not tested for HIV was four-times higher than among the uninfected population. However, the high proportion of the population not tested was a source of concern. The aim of the present study is to explore a simplified mathematical approach for estimating the likely composition by HIV status of the not-tested group, using mortality levels and patterns of the HIV-positive, HIV-negative and the not-tested group. Results suggest prevalence may have been under-estimated by 3-10%. This approach is easily applicable in other similar studies and helpful in the assessment of the representativeness of the non-response category.