Ahmed, Saifuddin, Qingfeng Li, Carolyn Scrafford, and Thomas W. Pullum. 2014. An Assessment of DHS Maternal Mortality Data and Estimates. DHS Methodological Reports No. 13. Rockville, Maryland, USA: ICF International.
This report evaluates the data quality of sibling survival history in DHS surveys and assesses its impact on maternal mortality estimates. The data are from 112 surveys conducted between 1990 and 2013; the surveys all included a sibling history module. We examined three aspects of data quality: completeness in sibling death reporting, completeness in reporting of all siblings and their age; and completeness in reporting of time of death during a calendar year and pregnancy status at the time of death. The overall quality of sibling history data is similar to the quality of data of individual women respondents in terms of age-misreporting. Reporting deaths in relation to pregnancy status and placing the deaths to calendar time remain major challenges for the respondents. The classification of maternal death relies on the reported timing of a sister’s death in relation to her pregnancy, delivery, and postpartum period. Many countries had very high non-response rates (10%-40%) for pregnancy status at the time of death, a situation that introduces substantial misclassification bias into identification of maternal deaths. DHS surveys currently report maternal mortality estimates without any adjustment for missing responses to timing of death. This may substantially underestimate the maternal mortality ratio (MMR) and other maternal mortality indicators in countries with high non-response rates. Trend analysis, however, suggests that the missing responses have declined significantly in the recent rounds of DHS surveys. We recommend developing a standard protocol for handling missing data on maternal status at the time of death and adjusting maternal mortality estimates with an appropriate statistical method for handling missing data.