To identify factors associated with the performance of a neonatal autopsy, we surveyed autopsy practice patterns in a tertiary care neonatal intensive care unit for 1 year. After 56 neonatal deaths, 33 (59%) autopsies were performed. We used multivariable analysis to compare the clinical and demographic characteristics of infants who had an autopsy with those who did not. We found two variables to be negatively correlated with having an autopsy performed: birth asphyxia (p < 0.05) and Medicaid coverage (p < 0.05). Early neonatal death (< or = 2 days of age) was not correlated with lack of an autopsy. Pulmonary hypoplasia occurred more often in the group having autopsy. However, only birth asphyxia was significantly correlated with lack of an autopsy after adjusting for other variables in the analysis. The primary cause of death was not associated with performance of an autopsy. We found no significant association between consent for autopsy and the characteristics of the physician requesting it. These data suggest that the infant's socioeconomic status and diagnosis may influence parental consent for autopsy. There may be other characteristics of physicians who request it, parents who consent, or other socioeconomic factors influencing consent for the neonatal autopsy.