儿童重症腺病毒肺炎危险因素分析及预测模型构建

Risk Factor Analysis and Prediction Model Construction of Severe Adenovirus Pneumonia in Children

目的 探讨影响儿童重症腺病毒肺炎的危险因素,构建风险预测模型,绘制临床决策曲线,为临床诊疗和预后提供指导。方法 对2017年1月~2021年12月因腺病毒肺炎住院的65例儿童进行回顾性分析,探讨重症腺病毒肺炎患儿的临床特征及危险因素,构建风险预测模型并绘制列线图。模型内部验证采用Bootstrap重抽样法。模型性能的量化使用受试者工作特征(receiver operating characteristic, ROC)曲线下面积(area under the curve, AUC)和校准曲线图进行拟合度测试和校准。绘制临床决策曲线分析模型的临床效用。结果 单因素分析结果显示,重症与非重症肺炎患儿的住院天数、咳嗽天数、喘息、三凹征、乳酸脱氢酶(lactate dehydrogenase, LDH)比较,差异均有统计学意义(P<0.05)。多因素Logistic回归分析结果显示,住院天数、喘息、LDH与儿童重症腺病毒肺炎显著相关。列线图显示,发生重度腺病毒肺炎概率为80.3%。预测模型结果显示,影响重症腺病毒肺炎患儿感染的AUC为0.899,其敏感度为80.0%,特异性为80.0%,预测效果较好。校准曲线图较好地显示了实际概率和预测概率之间的良好一致性。临床决策曲线显示了该模型对几乎所有阈值概率的明显净收益。结论 腺病毒肺炎好发于婴幼儿期,住院天数、LDH水平和有无喘息可作为预测儿童腺病毒肺炎严重程度的指标,为临床诊疗和预后提供指导性依据。

Objective To explore the risk factors for severe adenovirus pneumonia in children,construct a risk prediction model,draw a clinical decision curve,and provide guidance for clinical diagnosis,treatment and prognosis.Methods A retrospective analysis was conducted for 65 children hospitalized with adenovirus pneumonia from January 2017 to December 2021,to explore the clinical features and risk factors of children with severe adenovirus pneumonia,construct a risk prediction model and draw a nomogram.Bootstrap resampling was used for model internal validation.The quantification of model performance used area under the curve(AUC)of receiver operating characteristic(ROC)curve and calibration curve for fit testing and calibration.Draw the clinical decision curve to analyze the clinical utility of the model.Results The results of univariate analysis showed that there were statistically significant differences in hospital stays,cough days,wheezing,trident sign and lactate dehydrogenase(LDH,P<0.05).The results of multivariate Logistic regression analysis showed that the hospital days,wheezing and LDH were significantly associated with adenovirus pneumonia.The nomogram showed that the probability of severe adenovirus pneumonia was 80.3%.The results of the prediction model showed that the AUC affecting the infection of children with severe adenovirus pneumonia was 0.899,and its sensitivity was 80.0%,and its specificity was 80.0%,and the prediction effect was good.The calibration curve showed a good agreement between the actual probability and predicted probability.The clinical decision curve showed a clear net benefit of the model for almost all threshold probabilities.Conclusion Adenovirus pneumonia tends to occur in infants.Hospital stays,LDH level and wheezing can be used as indicators to predict the severity of adenovirus pneumonia in children,and provide a guiding basis for clinical diagnosis and treatment and prognosis.