The Predictive Value of Retrospective Survey Data on Maternal Mortality in Sub-Saharan Populations: Results from a Validation Study in Eastern Senegal

Stephane Helleringer, Columbia University
Malick Kante, Columbia University (CU)
Géraldine Duthé, Institut National d'Études Démographiques (INED)
Armelle Andro, Université Paris I, Panthéon Sorbonne and Institut National d'Études Démographiques (INED)
Gilles Pison, Institut National d'Études Démographiques (INED)

In most developing countries, maternal mortality ratios (MMR) are derived from survey data on the survival of a respondent’s siblings. While siblings’ survival histories are convenient to collect, they may underestimate mortality levels because 1) the survival of siblings is often correlated, 2) the number and timing of deaths are frequently misreported, and 3) the deaths attributable to maternal causes may be grossly misclassified. The extent of bias in survey-based MMR estimates due to misclassifications of maternal causes is unknown. We use a unique dataset linking survey reports of a sibling’s death with prospective records of that sibling’s death obtained from demographic surveillance in Eastern Senegal. The data include genealogical information, precise dates of pregnancies, deliveries and deaths collected since the 1970's, as well as information on causes of deaths from verbal autopsies. They allow calculating the specificity and sensitivity of reports of pregnancy-related deaths obtained during retrospective surveys.