The role of the autopsy in verifying clinical diagnosis and as a quality assurance tool in pediatric emergency medicine has not been studied. We reviewed the charts of all children who died soon after arriving at to the pediatric emergency department between October 1985 and December 1989. Opinions as to clinical diagnoses and cause of death were obtained by presenting a summary of patient data, in a blinded fashion, to three emergency pediatricians. Clinical diagnoses were then compared with autopsy diagnoses using the Class Error System. Major diagnostic errors (Class I, Class II) were examined to determine if the autopsy was more useful in any particular patient group. There were 69 children: 36 (52%) were female, and 30 (43%) were infants. Autopsies were performed on 52 (75%) patients. Autopsy diagnoses were categorized as follows: sudden infant death syndrome (SIDS) 14 (27%); underlying disease 15 (29%); trauma 13 (25%); sepsis 8 (16%); and aspiration 2 (4%). No errors were made in 67% of cases. The autopsy confirmed the major clinical diagnosis in 85% of patients. There were no Class I errors. Class II error rate was 15%. Most Class II errors occurred in patients between one and five years of ate (57%), and in patients who had sepsis or underlying disease (95%). Our data show that autopsy is useful in determining the cause of death and may be useful for education and quality assurance in pediatric emergency medicine.