Empirical Bayes Estimation of Small Area Adult Mortality Risk in Addis Ababa, Ethiopia

Brodie D. Ferguson, Stanford University
Georges Reniers, University of Pennsylvania
Tekebash Araya, Addis Ababa University
James H. Jones, Stanford University
Eduard Sanders, Ethiopian Health and Nutrition Research Institute

In Sub-Saharan Africa, small area analyses of health differentials have been inhibited by the paucity of data as well as quality concerns. This paper takes advantage of near-complete mortality data collected in Addis Ababa via ongoing burial surveillance in order to explore the relationship between AIDS mortality and selected district attributes. We make use of Bayes' Theorem to generate smoothed age- and sex-specific probabilities of death by district between the ages of 15 and 60, distinguishing between AIDS and non-AIDS mortality using estimates based on lay diagnosis. Results show the degree of unexplained variance in AIDS-attributable mortality to be significantly higher than for mortality due to other causes after taking certain socioeconomic determinants into account. Since death due to AIDS in Addis Ababa has been highly correlated with HIV illness due to the lack of available treatment, the results provide useful insight into the spatial characteristics of urban HIV transmission.