An outcome-based index of the effectiveness of perinatal care was empirically tested using 3,441,448 linked birth-death records. In order to construct the index, an "expected" perinatal mortality rate was computed for 504 California hospitals, based on each newborn's birth weight, sex, race, and plurality. The index was then defined as the ratio of the observed mortality rate to that expected. From the sixteen-fold range observed in the unadjusted rates, a two-fold variation could be attributed to differentials in the efficacy of medical care. The remainder resulted from differences in patient risk and from binomial variations. The observed--expected ratio was observed to be significantly correlated with a number of traditional indicators of medical care quality. Multiple regression techniques revealed that the mortality ratio was significantly lower in larger delivery services, in urban hospitals, in hospitals performing above-average numbers of cesarean sections in those recording Apgar scores, and in hospitals having higher specialist-to-generalist ratios. Conversely, the ratio was significantly higher in hospitals with larger percentages of Spanish-surnamed mothers and in private proprietary hospitals. By separating the index into its observed and expected components, the regression model accounted for 82 per cent of the variance in the observed perinatal mortality rates.