Studying multiple causes of death in the absence of death certificates: taking advantage of probabilistic cause-of-death estimation methods (InterVA)

Ariane Sessego, Ecole Normale Supérieure
Géraldine Duthé, Institut National d'Études Démographiques (INED)
Yempabou Bruno Lankoande, Institut National d'Etudes Démographiques (INED)/Institut Supérieur des Sciences de la Population (ISSP)
Kassoum Dianou, Catholic University of Louvain (UCLouvain) and Joseph Ki-Zerbo University (UJKZ)

While interest in multimorbidity has been rising to study more precisely the complex morbid processes that adults experience, health data in LMICs are scarce and rarely allow such investigations. Focusing on multimorbidity leading to death, we aim to develop an approach to estimate multiple causes of death using available data. In settings where certification of death by physicians is not available, verbal autopsies (VA) have been developed to diagnose likely causes of death from detailed information related to the death collected via the final caregivers. With an increasing use of probabilistic models to interpret VAs, we investigate their potential for identifying multiple causes using a database of 72,330 adult deaths from 22 Health and Demographic Surveillance System (HDSS) sites located in Asia and Africa, and detailed data from the Ouagadougou HDSS in Burkina Faso (1,700 deaths). The Bayesian model InterVA-4 attributes multiple likely causes to 11% of deaths. However, some combinations result more from uncertain diagnosis than multimorbidity. Elaborating an index of similarity between causes based on the InterVA’s probability matrix, we aim to differentiate competing causes (uncertainty) from co-occurring causes (multimorbidity).

Keywords: Civil Registration and Vital Statistics (CRVS), Health and morbidity, Mortality, Sustainable Development Goals (SDGs)

See extended abstract.

  Presented in Session 183. Measurement of Mortality and Causes of Death