[Reliability of the certification of the basic causes of neonatal deaths: implications for the study of preventable mortality].

The main causes of neonatal mortality, the reliability of the underlying cause of death registered in the death certificate, and the impact of problems of reliability on the analysis of preventable death were studied. The information on death certificates from a 15% sample of neonatal deaths between May 1986 and April 1987 in the Greater Metropolitan Region of Rio de Janeiro was compared to the information in the hospital records of the 452 deceased infants. A "modified underlying cause" considered most correct according to disease classification rules was identified from the records. The great majority of deaths (87%) were due to perinatal causes. Agreement between the originally declared and modified underlying causes of death was poor: 38% for 3 digits of the International Classification of Diseases Codes (CID-9) and 33% for 4 digits. The modified underlying causes are more weighted towards maternal conditions and complications, which increased by a factor of 12.8, and towards complications of the placenta, umbilical cord, labour and delivery, which rose by a factor of 6.2 in relation to the original causes. The utilization of the "modified" underlying cause elevated considerably (58%) the proportion of deaths considered reducible by the classification of neonatal death proposed by the SEADE Foundation. Seventy-five percent (75%) of deaths were considered reducible or partially reducible. One hundred and seven (24%) of the deaths of them being in infants of normal birthweight, of which 60% considered preventable. Four (4) deaths from congenital syphilis, 3 from perinatal hemolytic diseases, and 21 unattended home deaths of infants were also identified. In summary, important problems were identified in the reliability of the declaration of the underlying causes of neonatal death, whose correction tends to elevate the proportion considered reducible or preventable. The potential for the use of death certificate data for the monitoring of quality is evident, nonetheless improvements are needed in the quality of these data.