Estimer la mortalité aux âges élevés dans les pays à données incomplètes : les estimations basées sur la survie des parents sont-elles affectées par des biais de sélection ?

Not enough is known about old-age mortality in countries with limited civil registration and vital statistics systems. Surveys using parental survival histories (PSH) could help bridge this gap, providing estimates on deaths over age 50. We assess the potential for sample selection biases in these mortality estimates with data from three Health and Demographic Surveillance Systems in Senegal, covering around 9,600 women and 8,500 men. We use Cox models to estimate variation in mortality of men and women aged 50–89 according to the number of their adult daughters (typical survey respondents). Results indicate that older adults without surviving children are associated with a higher risk of mortality. PSH-based mortality estimates might therefore be lower than the ‘true’ levels of mortality over age 50, though we find no systematic bias. PSH is thus a promising instrument for measuring mortality in old ages, with sample selection bias likely minimal.