Improving survey data on pregnancy-related deaths in low and middle-income countries: a validation study in Senegal
Résumé
ObjectiveIn low- and middle-income countries (LMICs), siblings' survival histories (SSH) are often used to estimate maternal mortality, but SSH data on causes of death at reproductive ages have seldom been validated. We compared the accuracy of two SSH instruments: the standard questionnaire used during the demographic and health surveys (DHS) and the siblings' survival calendar (SSC), a new questionnaire designed to improve survey reports of deaths among women of reproductive ages.
MethodsWe recruited 1189 respondents in a SSH survey in Niakhar, Senegal. Mortality records from a health and demographic surveillance system (HDSS) constituted the reference data set. Respondents were randomly assigned to an interview with the DHS or SSC questionnaires. A total of 164 respondents had a sister who died at reproductive ages over the past 15years before the survey according to the HDSS.
ResultsThe DHS questionnaire led to selective omissions of deaths: DHS respondents were significantly more likely to report their sister's death if she had died of pregnancy-related causes than if she had died of other causes (96.4% vs. 70.9%, P<0.007). Among reported deaths, both questionnaires had high sensitivity (>90%) in recording pregnancy-related deaths. But the DHS questionnaire had significantly lower specificity than the SSC (79.5% vs. 95.0%, P=0.015). The DHS questionnaire overestimated the proportion of deaths due to pregnancy-related causes, whereas the SSC yielded unbiased estimates of this parameter.
ConclusionStatistical models informed by SSH data collected using the DHS questionnaire might exaggerate maternal mortality in Senegal and similar settings. A new questionnaire, the SSC, could permit better tracking progress towards the reduction in maternal mortality.