Model-based estimates of adult mortality in sub-Saharan Africa: comparisons with data on parental and sibling survival

Due to the lack of civil registration data in sub-Saharan Africa, adult mortality rates are often estimated in two stages; first, measures of child survival are combined with model schedules of mortality, then AIDS-related deaths are factored in. Model-based estimates are mostly sensitive to the choice of age pattern used to infer the background mortality and to violations of assumptions made to predict the number of AIDS-related deaths. They need to be regularly evaluated against survey or census data. In this paper, United Nations estimates of adult mortality are compared with survival probabilities calculated from sibling histories collected in DHS surveys. UNAIDS orphan prevalences are compared with proportions of orphaned children observed in censuses and household surveys. If parental and sibling survival provide plausible lower bounds, then adult mortality rates are underestimated in countries severely affected by the HIV/AIDS epidemic, especially in Southern Africa and among males. In addition, as the HIV epidemic unfolds, the age patterns of mortality derived from DHS surveys differ markedly from model outputs.