STUDY OBJECTIVE: To determine if there are significant differences in birth outcomes and survival for infants delivered by certified nurse midwives compared with those delivered by physicians, and whether these differences, if they exist, remain after controlling for sociodemographic and medical risk factors. DESIGN: Logistic regression models were used to examine differences between certified nurse midwife and physician delivered births in infant, neonatal, and postneonatal mortality, and risk of low birthweight after controlling for a variety of social and medical risk factors. Ordinary least squares regression models were used to examine differences in mean birthweight after controlling for the same risk factors. STUDY SETTING: United States. PATIENTS: The study included all singleton, vaginal births at 35-43 weeks gestation delivered either by physicians or certified nurse midwives in the United States in 1991. MAIN RESULTS: After controlling for social and medical risk factors, the risk of experiencing an infant death was 19% lower for certified nurse midwife attended than for physician attended births, the risk of neonatal mortality was 33% lower, and the risk of delivering a low birthweight infant 31% lower. Mean birthweight was 37 grams heavier for the certified nurse midwife attended than for physician attended births. CONCLUSIONS: National data support the findings of previous local studies that certified nurse midwives have excellent birth outcomes. These findings are discussed in light of differences between certified nurse midwives and physicians in prenatal care and labour and delivery care practices. Certified nurse midwives provide a safe and viable alternative to maternity care in the United States, particularly for low to moderate risk women.