Estimating mortality from external causes using data from retrospective surveys: a validation study in Niakhar (Senegal)

Gilles Pison, French National Museum of Natural History and Institut National d'Études Démographiques (INED)
Stéphane Helleringer, Columbia University
Bruno Masquelier, Université Catholique de Louvain
Malick Kante, Columbia University
Cheikh Tidiane Ndiaye, Agence Nationale de la Statistique et de la Démographie (ANSD)
Laetitia Douillot, Institut de Recherche pour le Développement (IRD)
Géraldine Duthé, Institut National d'Études Démographiques (INED)
Cheikh Sokhna, Institut de Recherche pour le Développement (IRD)
Valérie Delaunay, Institut de Recherche pour le Développement (IRD)

Mortality due to external causes (accidents, homicides or suicides) is poorly known in countries with limited registration of vital events and causes of death. Retrospective household-based surveys such as DHS increasingly collect siblings’ survival histories (SSH) in order to derive estimates of adult and pregnancy-related mortality. In some surveys, a few questions are also asked to identify violent deaths and this practice could be generalized to provide estimates of mortality from external causes. We conducted a validation study of SSH with such additional questions in Niakhar (Senegal), a locality where prospective data on adult mortality has been collected during demographic surveillance since 1962. We examine the sensitivity and specificity of SSH in recording adult deaths due to external causes. We then assess possible biases in SSH estimates of the proportion of adult deaths due to external causes.