Evaluation des estimations indirectes de mortalité dans trois observatoires de population au sénégal

In Sub-Saharan Africa, censuses are a key source of data for estimating mortality because death registration is incomplete. To assess the reliability of census-based mortality estimates, we apply different methods to individual-level data extracted from the 2002 and 2013 Senegalese censuses covering three health and demographic surveillance systems (HDSS) in Bandafassi Mlomp and Niakhar. The under-five mortality rates inferred from reports on children ever born and surviving are lower than expected based on longitudinal demographic surveillance. Estimates derived from reports on parental survival are also much lower than the adult mortality levels observed in HDSS field sites. By contrast, age-specific death rates based on recent deaths reported in households are consistent with HDSS data, except for infant mortality, which is significantly under-reported in 2002. This evaluation confirms that indirect estimates of mortality obtained from census data should not be considered in isolation but must instead be systematically compared with each other. Direct evaluation studies conducted at the individual level using record linkages are needed to better identify the various sources of bias.