Abstract: Objective To construct and validate a prediction model for severe hyperbilirubinemia complicating ABO hemolytic disease of the fetus and newborn (ABO-HDNF), providing a basis for timely clinical intervention. Methods Weretrospectively analyzed clinical data from 306 ABO-HDNF infants diagnosed and treated between January 2022 and December 2024. Of these, 204 cases were regarded as the modeling set and 102 the validation set. Patients were stratified into severe and non-severe hyperbilirubinemia groups. Risk factors for severe hyperbilirubinemia were identified through univariate analysis and binary logistic regression. A nomogram model was constructed, with receiver operating characteristic (ROC) and calibration curves evaluating its performance. Clinical utility of the model was further evaluated using a clinical decision curve. Results Univariate and binary logistic regression analyses identified postnatal time at admission (OR=1.303), positive Direct Antiglobulin Test (OR=3.073), eluate agglutination strength (OR=3.189), and total bilirubin/albumin ratio (OR=1.863) as independent risk factors for severe hyperbilirubinemia (P<0.05). The nomogram demonstrated a C-index of 0.790 (95%CI: 0.713~0.867). Area under the curve (AUC) values were 0.790 (95%CI: 0.713~0.867) and 0.754 (95%CI: 0.638~0.871) for modeling and validation sets, respectively. Calibration curves indicated good consistency, and decision curve analysis showed high net clinical benefit. Conclusion The nomogram incorporating postnatal time, Direct Antiglobulin Test results, eluate agglutination strength, and total bilirubin/albumin ratio exhibits favorable predictive performance for severe hyperbilirubinemia complicating ABO-HDNF.

Key words: ABO hemolytic disease of the fetus and newborn, Severe hyperbilirubinemia, Predictive model Nomogram