Estimating cause-specific mortality in Madagascar: a comparison of the Global Burden of Disease Study 2013 with registration data from the capital city

Objective: Due to incomplete death registration in Madagascar, trends in cause-specific mortality are known from epidemiological models such as the Global Burden of Disease Study 2013 (GBD 2013). The external validity of such model-based estimates is difficult to establish, however, since all possible data are usually included. In Madagascar, the GBD 2013 study used death registration data from Antananarivo but only for the period 1984-1995. Using death records up to 2014, we aim to describe trends in mortality by age and sex and compare the distribution of causes of death observed in death records with GBD 2013. Methods: Information on the sex of the deceased, date of birth and death and underlying cause of death were transcribed from death registers. Causes of death were converted in GBD cause categories. The completeness of the registration of deaths was assessed using indirect demographic techniques. Findings: The completeness of the registration of deaths is higher than 90% in Antananarivo. The proportion of deaths attributed to "garbage codes" amounts to 19.4% when considering codes for symptoms, signs and ill-defined conditions, deaths from injuries whose intent is not determined, and ill-defined cardiovascular diseases and neoplasms. Significant differences are observed between the distributions of deaths by cause in the GBD 2013 estimates and death registers referring to the capital city. Conclusions: Death registration in Antananarivo, and possibly other African cities, is a low-cost and untapped data source which can help to refine disease burden estimates.