Objective To identify predictors of perinatal and infant mortality variations between primary care trusts (PCTs) and identify outlier trusts where outcomes were worse than expected. Design Prognostic multivariable mixed models attempting to explain observed variability between PCTs in perinatal and infant mortality. We used these predictive models to identify PCTs with higher than expected rates of either outcome. Setting All primary care trusts in England. Population For each PCT, data on the number of infant and perinatal deaths, ethnicity, deprivation, maternal age, PCT spending on maternal services, and “Spearhead” status. Main outcome measures Rates of perinatal and infant mortality across PCTs. Results The final models for infant mortality and perinatal mortality included measures of deprivation, ethnicity, and maternal age. The final model for infant mortality explained 70% of the observed heterogeneity in outcome between PCTs. The final model for perinatal mortality explained 80.5% of the between-PCT heterogeneity. PCT spending on maternal services did not explain differences in observed events. Two PCTs had higher than expected rates of perinatal mortality. Conclusions Social deprivation, ethnicity, and maternal age are important predictors of infant and perinatal mortality. Spearhead PCTs are performing in line with expectations given their levels of deprivation, ethnicity, and maternal age. Higher spending on maternity services using the current configuration of services may not reduce rates of infant and perinatal mortality.