The Effect of Non-Response on Population-Based HIV Prevalence Estimates: The Case of Rural Malawi
Francis Obare, University of Pennsylvania
I explore if non-response to HIV testing exerts a downward bias in HIV prevalence rates obtained from population-based surveys. I use biomarker data that was collected between March and August 2004 in three rural districts of Malawi (Balaka, Mchinji, and Rumphi) by the Malawi Diffusion and Ideational Change Project (MDICP). I employ a sensitivity analysis and a probit regression model with sample selection to determine the extent to which selective response to HIV testing biases the MDICP estimates of HIV prevalence. The results from sensitivity analysis show that non-response to HIV testing seems to exert a downward bias in HIV prevalence rate for one district, Rumphi, but not for Balaka and Mchinji districts. However, the selection model shows that overall, selective response to HIV testing does not significantly bias the MDICP HIV prevalence estimates. The case of Rumphi could therefore be genuine or an artifact of the assumptions made.