OBJECTIVE: To identify the frequency, risks of fetal and early neonatal mortality and the determinants of accidental home deliveries. METHODS: A population-based case control study of fetal and early neonatal deaths was carried out in the southern area of São Paulo, Brazil. Data were collected through home interviews and hospital record reviews. The reasons reported by the mothers were obtained from interviews and risk factors for home delivery were obtained comparing home to hospital deliveries. Data were analyzed separately for fetal and early neonatal deaths and survivors. Odds ratios, 95% confidence intervals and Fisher's exact test were used in estimating risk factors and mortality risk. RESULTS: The 0.2% frequency of home deliveries was underestimated in the live births information system. After adjustment, it reached 0.4%, comparable to other urban areas in Europe. All home deliveries identified were accidental and were associated to an increased fetal and early neonatal mortality. Mothers' social conditions and pregnancy characteristics were associated to accidental home deliveries and these factors are different outcomes studied (fetal losses, early neonatal deaths and survivors). In 30%, mothers reported lack of available transportation to the hospital as a reason for home delivery. Failure of health services in identifying labor women and non-availability of emergency care contributed to accidental home deliveries. CONCLUSIONS: Though rare events in urban areas, accidental home deliveries should be of special concern to health services because they seem to be avoidable and imply in increased risk of death.