Errors in reported ages and dates in surveys of adult mortality: A record linkage study in Niakhar (Senegal)

Sibling survival histories are a major source of adult mortality estimates in countries where death registration is incomplete. We evaluate age and date reporting errors in sibling histories collected during a validation study in the Niakhar Health and Demographic Surveillance System (Senegal). Participants were randomly assigned to the questionnaire used in Demographic and Health Surveys or to a questionnaire incorporating an event history calendar, recall cues and increased probing strategies. We linked 60–62% of survey reports of siblings to the reference database using manual and probabilistic approaches. Both questionnaires had high sensitivity (>96%) and specificity (>97%) in recording the vital status of siblings. Respondents underestimated the ages of living siblings, ages at death, and the time since deaths. These reporting errors introduced downward biases in mortality estimates. The revised questionnaire improved the reporting of ages of living siblings but not the reporting of ages at death or the timing of deaths.