[Reliability of the declaration of underlying cause of infant deaths in the metropolitan region of southeastern Brazil].

The quality of official information on underlying causes of infant deaths was studied on the basis of data collected for a population-based study of the surveillance of infant mortality in the metropolitan region of Belo Horizonte, Brazil in 1989. The survey included the analysis of a sample of infant deaths carried out by comparing the underlying causes of death as coded on death certificates to those recorded by a group of doctors who abstracted information from hospital records. We verified that 11.7% of neonatal deaths did not have the underlying cause of death confirmed by the investigation (kappa = 0.61), and neither did 44.0% of post-neonatal deaths (kappa = 0.47). It is believed that this major disagreement among post-neonatal deaths is due to the close correlation observed among the major causes of death within this group (pneumonia, diarrhoea and malnutrition). For example, associated malnutrition was observed in 76.9% of those cases in which diarrhoea was coded as the underlying cause of death. It was concluded that the quality of the death certificates is not satisfactory. However, the composition of the main groups of causes presented no significant alteration after investigation and may be used in public health surveillance, especially if we regard pneumonia, diarrhoea and malnutrition as a group with the same determinants. Unfortunately, this group still accounts for a great number of otherwise avoidable deaths in Brazil.