The objectives of the investigation were (i) to study infants registered in a statewide audit of tertiary neonatal intensive care units in New South Wales, Australia in 1992 and who died, and (ii) to examine postmortem rates, quality of postmortem reports and compare clinical cause of death with postmortem report. Death rates, data on clinical cause of death and postmortem status were collected prospectively as part of the routine audit. Postmortem reports were examined by LS. Fifteen percent of the cohort died and 43% had a postmortem examination. The postmortem rate was highest in the 28-36 week gestation group and in babies dying of pulmonary haemorrhage, intracranial haemorrhage or sudden infant death syndrome. Fewer than 50% of babies with a major congenital anomaly had a postmortem. The postmortem changed the major diagnosis in 10% of cases and added useful information in 17%. We conclude that postmortem examination should be an essential part of any audit of neonatal intensive care unit outcomes.