Model Uncertainty in World Wide Estimates of Maternal Mortality

Robert G. White, University of Wisconsin at Madison
Alberto Palloni, University of Wisconsin at Madison

This paper assesses two important sources of uncertainty in estimates of levels and trends in worldwide maternal mortality. A method for assessing the sensitivity of methods for handling measurement error in country reports of maternal mortality is developed to assess the impact of selective non-response and model uncertainty. A set of simulations are undertaken to generate new uncertainty bounds for worldwide estimates of maternal mortality that account for these additional sources of uncertainty. The corrected estimates demonstrate the large impact that accounting for these two additional sources of uncertainty may incur for estimates of maternal mortality. The large magnitude of these corrected bounds further demonstrates how accounting for these two common sources of uncertainty undermines even modest estimates of trends in maternal mortality.