We sought to quantify neonatal mortality (< 28 days) in a 10-hospital system, determine what proportion was associated with suboptimal neonatal care and make recommendations on how neonatal mortality rates (NMRs) could be used in quality improvement efforts. Deaths were identified using electronic linkage to the State of California Death Certificate Tapes. Individual fatalities were reviewed by a minimum of two physicians who did not care for the infant. Deaths were classified as either being associated with suboptimal care or not. For deaths where suboptimal care was an issue, emphasis was on delineating the process involved in the death. Subjects were all neonatal deaths among 64,469 babies born in 1990-91 in the 10 birth facilities of the Kaiser Permanente Medical Care Program, Northern California Region. A total of 241 neonatal deaths were identified. Adjusting for prematurity by increasing the follow-up period in preterm babies (included as neonatal deaths if they died up to 40 weeks corrected gestational age + 27.9 days) increased overall mortality rates by 5%. Birthweight-specific NMRs in Kaiser Permanente are similar to those of other published reports. Among the 198 deaths in babies weighing > or = 500 g at birth, only 14 (7%) were possibly associated with suboptimal care. In populations with access to health insurance, reporting only aggregate NMRs is of limited use. The number of deaths that could be ascribed to suboptimal neonatal care is very small and measuring variations in rates of such deaths is difficult. Future measurements of quality of care will require more sophisticated measures, database systems, review strategies and dissemination methods.