Data was analyzed on all infants born at Hlabisa Hospital and at seven village clinics in northern Zululand, South Africa, and those born before arrival at the hospital between May 1991 and April 1993 (total deliveries 8237) to determine the perinatal mortality rate (PNMR), to evaluate the quality of care, and to identify avoidable factors associated with the deaths. Findings were used to develop ways to enable effective monitoring of quality of care and to develop intervention strategies to prevent recurrence of avoidable deaths. There were 280 perinatal deaths (PNMR 34/1000). The PNMR increased 30% in the second year (29-39/1000; p = 0.01), mostly due to increased rates of attendance of high risk cases at the hospital and clinics. 27 of the perinatal deaths were considered avoidable. Had these deaths not happened, the overall PNMR would have stayed at 31/1000. The percentage of avoidable deaths decreased over time (19% in first 6 months vs. 3% in the last 6 months; p 0.01 and p 0.01 for linear trend) because of interventions implemented throughout the study period to improve quality of care. There were three groups of interventions: reorganization of the maternity services in the health ward, development of key standing orders and management guidelines, and creation of a monitoring system for perinatal mortality.