传染性单核细胞增多症患儿肝损害的风险预测评分模型构建研究

Study on the Construction of a Scoring Model for the Risk Prediction of Liver Damage in Children with Infectious Mononucleosis

目的:构建传染性单核细胞增多症(IM)患儿肝损害的风险预测评分模型。方法:回顾分析2015年1月—2019年12月收治的381例IM患儿的临床资料,按照2∶1比例将患儿随机分为训练集和验证集,采用多因素Logistic回归分析法探寻IM患儿肝损害的危险因素,并应用R 3.62软件构建列线图风险预测评分模型,绘制受试者工作特征(ROC)曲线、校准曲线、决策曲线验证模型的区分度、校准度以及临床有效性。结果:训练集和验证集临床资料比较差异无统计学意义(P>0.05);多因素Logistic回归分析结果显示,年龄>6岁、肝肿大、热程>7d、CD8^(+)升高、全血EBV DNA≥3.38 lg copies/mL均是IM患儿肝损害的危险因素(P<0.05),CD4^(+)/CD8^(+)升高是IM患儿肝损害的保护因素(P<0.05);基于以上6个影响因素构建列线图风险预测评分模型,ROC曲线显示,该模型在训练集中的敏感度、特异度和曲线下面积分别为85.25%、89.39%、0.857(0.772~0.943),在验证集中的敏感度、特异度和曲线下面积分别为80.00%、87.10%、0.837(0.756~0.919);训练集和验证集中校准前后预测曲线走向一致且贴近;决策曲线显示,在训练集阈概率9%~90%范围内、验证集阈概率2%~99%范围内该模型具有良好的临床应用效果。结论:基于年龄、肝肿大、热程、CD8^(+)、CD4^(+)/CD8^(+)和全血EBV DNA构建IM患儿肝损害风险的列线图预测评分模型具有良好的区分度、校准度以及临床应用效果。

Objective:To construct a risk prediction scoring model for liver damage in children with infectious mononucleosis(IM).Methods:The clinical data of 381 IM patients admitted from January 2015 to December 2019 were retrospectively analyzed,and the children were randomly divided into the training set and the validation set in a 2∶1 ratio.Multivariate Logistic regression analysis was used to identify the risk factors for liver damage in IM children,and R 3.62 software was used to construct a risk prediction scoring model for nomogram.The receiver operating characteristic(ROC)curve,calibration curve,and decision curve were plotted to validate the model's discrimination,calibration and clinical effectiveness.Results:There was no statistically significant difference in clinical data between the training set and the validation set(P>0.05).The results of multivariate Logistic regression analysis showed that age>6 years old,liver enlargement,fever duration>7 days,elevated CD8^(+),and whole blood EBV DNA≥3.38 lg copies/mL were all risk factors for liver damage in IM children(P<0.05),while elevated CD4^(+)/CD8^(+)was a protective factor for liver damage in IM children(P<0.05).Based on the above six influencing factors,a risk prediction scoring model for nomogram was constructed,and the ROC curve showed that the sensitivity,specificity,and area under the curve of the model in the training set were 85.25%,89.39%and 0.857(0.772~0.943),respectively.The sensitivity,specificity,and area under the curve in the validation set were 80.00%,87.10%and 0.837(0.756~0.919),respectively.The predicted curve trend before and after calibration in the training and validation sets was consistent and close.The decision curve showed that the model had good clinical application results in the range of training set threshold probability of 9%~90%and validation set threshold probability of 2%~99%.Conclusion:A nomogram scoring model for predicting the risk of liver damage in children with IM based on age,liver enlargement,fever,CD8^(+),CD4^(+)/CD8^(+),and whole blood EBV DNA has good discrimination,calibration,and clinical application effectiveness.