OBJECTIVES. Race is a predictor of health outcomes and risk for some clinical conditions, for example, mother's race predicts risk for hyperbilirubinemia in newborns, with blacks at lowest risk. Little is known about the correlation of race as recorded in medical records with self-reported race. Also, use of maternal race to predict newborn risk for hyperbilirubinemia has not been tested for multiracial mothers and newborns. We sought to examine how maternal race documented in medical records correlates with self-reported race and to examine the correlation between mothers' and newborns' race in the context of risk for neonatal hyperbilirubinemia, focusing on multiracial mothers and newborns. DESIGN. A cohort study with 3021 newborns at >= 35 weeks gestation discharged from normal nursery between January 2001 and October 2002 with a telephone survey of their mothers within 6 months of birth. SETTING. The study was conducted in the Neonatology Department of Henry Ford Hospital. PATIENTS. There were 1773 mothers ( 58%) with incorrect telephone numbers. Of 1248 mothers contacted, 866 ( 69%) completed the interview. OUTCOME MEASURES. We measured mother's race in hospital database and mother's reported race for herself, her newborn, and the father, allowing <= 5 responses for each. RESULTS. Of mothers documented in the medical record as white, 64% self-reported as white. Among mothers recorded as black, 70% self-reported as black. Mothers identified 93 newborns as >= 2 races with primary race matching both parents for 41%, father for 25%, mother for 23%, and neither parent for 11%. Of 70 newborns whose parents were not the same race, mothers identified 45 ( 64%) as >= 2 races. CONCLUSIONS. There is incomplete overlap between racial identification in medical records versus self-report. Given 1 choice, mothers of multiracial infants overselect black in their newborns' ancestry. Because black race is the lowest risk category for neonatal hyperbilirubinemia, this may lead to underestimating their risk.