Development of a new model for estimating maternal mortality ratio at national and sub-national levels and its application for describing sub-national variations of maternal death in Ethiopia

Background In developing countries lacking functional vital registration system, statistical models are being increasingly used for estimating maternal mortality ratio (MMR). Yet, most of the models have limited applicability at sub-country level. This paper introduces a new model for estimating MMR at national and sub-national levels based on maternal health-related indicators. Further, it applies the model for explaining sub-national variations of MMR in Ethiopia. Methods Country level data on MMR and other nine potential predictors of maternal death were extracted from 248 national Demographic and Health Surveys and other related surveys conducted in 80 low-and middle-income countries since 1990. Additional data were obtained from the World Bank and the World Health Organization databases. The potential model predictors were: contraceptive prevalence rate (CPR); utilizations of antenatal care (ANC), health institution delivery (HID), Caesarian section (CS) and postnatal care (PNC); and prevalence of maternal anemia, thinness (Body Mass Index (BMI < 18.5 kg/m(2)), short stature (height less than 145 cms) and HIV. Stepwise Generalized Estimating Equation (GEE) with Negative Log-binomial link was employed to model MMR as a function of the covariates. Results The ultimate model comprised six significant predictors and the equation is provided as: Ln (MMR) = 6.464-0.013(CPR)-0.006(HID)-0.003(PNC)-0.027(CS rate) + 0.060(HIV prevalence) + 0.011 (thinness prevalence). The variation explained by the model was 54.3%) and the mean (+/- SD) relative standard error (8.6 +/- 2.6%) suggested the model has a reasonable precision. Application of the model to describe sub-national variation of maternal mortality in Ethiopia indicated, in 2016 the highest MMRs per 100,000 live births were in Somali (805) and Afar (795) regions. Between 2000 and 2016, all the regions of Ethiopia have significantly reduced MMRs; yet the rate of decline was lower in the aforementioned two regions. Oromiya region contributed for 46% of all maternal deaths in Ethiopia. Conclusion In developing countries lacking dependable maternal mortality data; the model can be used to estimate national and sub-national MMR with reasonable accuracy and precision.