目的 探讨基于临床、影像及电生理学特征的联合模型对新生儿急性胆红素脑病（ABE）的风险分层能力及诊断价值。 方法 回顾性分析275例高胆红素血症新生儿的临床资料、听性脑干反应（auditory brainstem response， ABR）、振幅整合脑电图（amplitude-integrated electroencephalogram， aEEG）及头颅MRI图像；按 7∶3随机分为训练集（n = 193）和验证集（n = 82）。通过单因素及多因素logistic回归分析筛选独立危险因素构建列线图联合模型。采用受试者工作特征（receiver operating characteristic， ROC）曲线、Hosmer-Lemeshow检验、校准曲线及决策曲线分析（decision curve analysis， DCA）评估模型的区分度、拟合度、校准度及临床净获益。应用Delong检验比较各单因数与联合模型的曲线下面积（area under the curve， AUC），若AUC差异无统计学意义，则进一步采用净重分类改善指数（net reclassification improvement， NRI）评估模型的增量价值。 结果 葡萄糖-6-磷酸脱氢酶（G-6-PD）缺乏、听觉异常及苍白球/壳核（globus pallidus/putamen， G/P）信号强度比值被确定为ABE的独立危险因素并纳入建模。在训练组中，联合模型诊断ABE的AUC为0.968，显著高于各单项因素（均P < 0.05）。在验证组中，联合模型AUC为0.945，NRI分析显示联合模型显著改善了风险重分类（NRI = 1.25，P < 0.001）。在训练组及验证组中Hosmer-Lemeshow检验（均P > 0.05）、校准曲线及DCA均证实该模型具有良好的拟合度、校准度及广泛的临床实用性。 结论 临床-影像-电生理多模态联合模型显著提升了ABE的诊断效能与风险分层能力，为早期识别高危患儿及制定精准诊疗策略提供了可靠的客观依据。

Objective To investigate the risk stratification capability and diagnostic performance of a combined clinical， radiological， and electrophysiological model in neonatal acute bilirubin encephalopathy （ABE）. Methods We retrospectively analyzed clinical data， auditory brainstem response （ABR） parameters， amplitude-integrated electroencephalography （aEEG） findings， and cranial magnetic resonance imaging （MRI） features from 275 neonates with hyperbilirubinemia. The neonates were randomly allocated to a training cohort （n = 193） and a validation cohort （n = 82） in a 7∶3 ratio. Independent risk factors identified through univariate and multivariable logistic regression analyses were incorporated to construct a combined predictive nomogram. Model performance was evaluated using the area under the receiver operating characteristic （ROC） curve （AUC）， the Hosmer-Lemeshow （HL） goodness-of-fit test， calibration curves， and decision curve analysis （DCA） to assess discrimination， goodness-of-fit， calibration， and clinical utility， respectively. DeLong’s test was used to compare the AUC of the combined model with those of the individual predictors. Furthermore， the net reclassification improvement （NRI） was calculated to evaluate the incremental predictive value of the combined model. Results Glucose-6-phosphate dehydrogenase （G-6-PD） deficiency， abnormal ABR findings， and the globus pallidus/putamen （G/P） MRI signal intensity ratio were identified as independent risk factors and incorporated into the model. In the training cohort， the combined model achieved an AUC of 0.968， significantly outperforming any individual predictor （all P < 0.05）. In the validation cohort， the model maintained a high AUC of 0.945. Additionally， NRI analysis demonstrated that the combined model significantly improved risk reclassification （NRI = 1.25， P < 0.001）. The HL test （P > 0.05）， calibration curves， and DCA confirmed that the model exhibited satisfactory goodness-of-fit， excellent calibration， and favorable clinical utility in both cohorts. Conclusions The multimodal clinical-radiological-electrophysiological model significantly improves both diagnostic performance and risk stratification capabilities in ABE. It provides a reliable and objective tool for the early identification of high-risk neonates and the formulation of personalized clinical management strategies.