Abstract： Objective To evaluate the predictive accuracy of several risk-assessment strategies to predict the risk of significant neonatal hyperbilirubinemia, and to establish the best prediction model.Methods The transcutancous bilirubin (TcB) levels of 4907 term and near-team infants were measured.Trace blood bilirubin levels of the infants whose TcB levels ≥250 μmol/L were detected. Clinical data of newborns and their mothers were collected and were analyzed with Logistic regression model to investigate its correlation with signifrcant hyperbilirubinemia. Clinical high risk factors of significant neonatal hyperbilirubinemia were determined. Accuracy of three prediction methods for significant hyperbilirubinemia was compared by receiver operating characteristic (ROC) curve. The three methods included: whether predischarge bilirubin level (within 72 hours after birth) expressed in risk zone on an hour-specific bilirubin nomogram; clinical risk factors other than predischarge bilirubin level; and combination of the predischarge bilirubin risk zone and other clinical risk factors. Results Two hundred and eighty-six newborns (5.8%) were found with significant hyperbilirubinemia. The risk factors of significant neonatal hyperbilirubinemia were divided into three groups according to OR: (1) Major risk factors:predischarge (within 72 hours after birth) bilirubin level in the high risk-zone (OR=96. 39, 95% CI:53.32-174.27, P = 0. 000), large cephalohematoma (OR = 36.45, 95% CI: 10. 02-132.56,P=0. 0076), gestational age 35-36+6 weeks (OR= 30. 72, 95% CI 14.47-65.23, P=0. 0001) and exclusive breast feeding and weight loss was ＞9% of birth-weight (OR=22.44, 95% CI: 4.42-114. 03, P=0. 0016). (2) Minor risk factors: gestational age 37-37+6 weeks (OR=3.26, 95% CI:1.92-5. 55, P=0. 0232), predischarge bilirubin level in P76-P95(OR=13. 64, 95% CI: 8. 10-22.97,P=0. 0001) and bruising (OR = 2.32, 95% CI: 1.14-4.71, P = 0. 0497). (3)Protective factors (those factors associated with decreased risk of hyperbilirubinemia): predischarge bilirubin level in low-risk zone (≤P40) (OR=0. 00), gestational age ≥40 weeks (OR=0.21, 95% CI: 0.09-0.44,P=0. 0402) and mixed breeding (OR=0. 75, 95% CI: 0. 58-0.95, P=0.0059). The area under the ROC curve of predischarge bilirubin level was 0. 8687 and 0. 7375 for clinical risk factors other than predischarge bilirubin level. The area under the ROC curve of a combination of the predischarge bilirubin risk zone and additional clinical risk factors was 0. 9367. Conclusions The risk of significant neonatal hyperbilirubinemia could be simply and accurately predicted by infant's predischarge bilirubin level and the combination of predischarge bilirubin level, and clinical risk factors might improve the accuracy of prediction significantly.