Strengthening vital statistics in Brazil: investigation of ill-defi ned causes of death and implications on mortality statistics

Ministry of Health of Brazil. Universidade Federal de Minas Gerais. Grupo de Epidemiologia e Avaliação em Saúde. Belo Horizonte, MG, Brasil. Pan American Health Organization. Washington, DC, USA. Universidade Federal de Minas Gerais. Grupo de Epidemiologia e Avaliação em Saúde. Belo Horizonte, MG, Brasil. Universidade Federal de Minas Gerais. Grupo de Epidemiologia e Avaliação em Saúde. Belo Horizonte, MG, Brasil. Fundação Oswaldo Cruz. Instituto de Comunicação e Informação Científica e Tecnológica em Saúde. Rio de Janeiro, RJ, Brasil. Background The number of ill-defi ned causes of death (IDCD) has been a signifi cant problem among registered causes of death in Brazil. The IDCD proportion was 14·3% in 2000, with huge regional diff erences. In 2005, the Brazilian government implemented a project in order to decrease the IDCD in states and municipalities in the poorest regions. This study aims to compare the distribution of causes of death between IDCD investigated and observed data. Methods For each death certifi cate with IDCD all attempts were made by health-service professionals to trace existing information about the fi nal disease and cause of death. The sources of information were hospital records, municipality health departments, autopsies, family health teams, and civil registry records. For deaths that occurred at home, verbal autopsy questionnaires were applied. The underlying cause of death was assigned using the documented evidence available. Findings During 2006–10, 27·6% (132 056 of 478 674) of IDCD were investigated, and 64·2% (84 751 of 132 056) of these deaths were reclassifi ed into a defi ned group of causes. Deaths from diabetes, neuropsychiatric, and maternal conditions occurred with a higher frequency among ill-defi ned causes investigated and were under-diagnosed in observed data. Injuries were also under-reported, responsible for 7·7% (6551 of 84 751) of the diagnoses among IDCD in that period. Cancer and respiratory diseases occurred with a higher frequency among registered deaths than among IDCD (16·3% vs 11·1% for cancer and 11·0 vs 7·8% for respiratory diseases) when compared with circulatory or endocrine diseases (30·6% vs 43·1% for circulatory and 6·2% vs 10·2% for endocrine diseases). Interpretation These results show that the distribution of the cause of deaths after investigation of IDCD was diff erent from those in observed data. Therefore, these diff erences must be taken into consideration when making redistribution of ill-defi ned causes based on observed data to avoid bias. The investigation of IDCD is critical to the creation of a reference for applying corrections to the observed data as well as to strategically improve the quality of mortality data.