新生儿溶血病患儿重度高胆红素血症列线图分析

Analysis of column chart model for severe hyperbilirubinemia in children with hemolytic disease of newborns

目的:分析新生儿溶血病(hemolytic disease of newborns, HDN)患儿重度高胆红素血症影响因素和预测模型,为临床预防、治疗、规范化管理提供依据。方法:选取2021年1月—2023年10月联勤保障部队第九〇〇医院诊断和治疗的HDN患儿147例作为测试集,同期中南大学湘雅医学院附属株洲医院HDN患儿86例作为验证集。根据是否发生重度高胆红素血症将患儿分为重度高胆红素血症组和非重度高胆红素血症组,单因素和多因素分析HDN患者重度高胆红素血症的影响因素,受试者工作曲线(ROC)、校正曲线评估模型预测效能。结果:单因素和多因素分析显示,黄疸异常开始时间、放散液凝集强度、直接抗人球蛋白试验(DAT)阳性、脐血胆红素、HDN类别均为HDN患儿重度高胆红素血症的独立影响因素,OR值分别为0.387、2.549、3.015、1.043、2.559。基于二元logistic回归影响因素建立列线图模型,经验证模型在测试集预测曲线下面积(AUC)为0.842,灵敏度为78.43%,特异度为73.96%,在验证集预测AUC为0.813,灵敏度为84.61%,特异度为61.67%。决策曲线分析显示,当列线图预测概率为0~0.34时,使用该模型可获得更高的临床收益。结论:HDN重度高胆红素血症列线图模型具有较好的准确性、校准度,可为HDN发生重度高胆红素血症的预防、诊断、管理提供参考依据。

Objective: To construct a predictive model for severe hyperbilirubinemia in children with hemolytic disease of newborns(HDN), and provide a basis for clinical prevention, treatment, and standardized management. Methods: A total of 147 children with HDN diagnosed and treated at the 900th Hospital between January 2021 and October 2023 were selected as the test set, and 86 HDN children at the Zhuzhou Hospital affiliated to Xiangya Medical College of Central South University during the same period were selected as the validation set. According to the occurrence of severe hyperbilirubinemia, the children were divided into severe hyperbilirubinemia group and non severe hyperbilirubinemia group. The clinical data differences between the two groups were compared. Based on the principle of binary logistic regression analysis, a HDN severe hyperbilirubinemia column chart model was constructed, and the predictive performance of the model was evaluated through receiver operating curve(ROC) and correction curve. Results: Univariate and multivariate analyses showed that the onset time of jaundice abnormalities, coagulation intensity of dispersed fluid, DAT positive, bilirubin in cord blood, and HDN category were all independent influencing factors for severe hyperbilirubinemia in the children with HDN, the OR values were 0.387, 2.549, 3.015, 1.043 and 2.559, respectively. Based on the above influencing factors, a risk prediction model was constructed. After verification, the model predicted an area under the curve(AUC) of 0.842, sensitivity of 78.43%, and specificity of 73.96% in the test set. The AUC predicted in the validation set was 0.813, sensitivity of 84.61%, and specificity of 61.67%. Decision curve analysis showed that when the predicted probability of the column chart model was between 0 and 0.34, using this model to predict the occurrence of severe hyperbilirubinemia in HDN could achieve higher clinical benefits. Conclusion: The HDN severe hyperbilirubinemia column chart model may have good accuracy and calibration, which can provide reference for the prevention, diagnosis, and management of severe hyperbilirubinemia in HDN.