Maternal Mortality Estimation at Sub-National Level with Empirical Bayes Method: An Illustration of District-Level MMR Estimates in Bangladesh

Saifuddin Ahmed, Johns Hopkins University

National estimation of maternal mortality ratio (MMR) is notoriously difficult in developing countries, and few countries have attempted ever to estimate MMR at sub-national levels. Maternal mortality is more pronounced among poorer segments of population, in rural areas and in areas where maternal care service availability and utilization are lacking Exploiting such a priori knowledge about the determinants of maternal mortality, we use empirical Bayes prediction method to estimate maternal mortality at sub-national levels from the spatial distribution of such factors in the area. We illustrate an example of MMR estimation in 64 districts of Bangladesh with the data from Bangladesh Maternal Morbidity and Mortality Survey (BMMS), 2001. BMMS estimated MMR of 382 per 100,000 live births, and our empirical Bayes estimates show that MMR ranged from 160 to 1055 per 100,000 live births at district levels in Bangladesh. The results will help in prioritizing and targeting the high mortality areas.