儿童反复呼吸道感染的影响因素分析及其预测模型构建与验证

Factors influencing recurrent respiratory tract infection in children and construction and validation of predictive models

目的分析儿童反复呼吸道感染(recurrent respiratory tract infection,RRTI)的影响因素并构建预测模型,探究该模型的预测价值。方法采用回顾性研究,收集2021年2—12月在某院儿科就诊的589例呼吸道感染患儿临床资料及膳食炎症指数(dietary inflammatory index,DII)测评结果,按7∶3比例随机分配到建模集(412例)、验证集(177例)。对建模集资料进行单因素和多因素logistc回归分析筛选出儿童RRTI的影响因素,并构建列线图预测模型。模型的表现和临床实用性分别用受试者工作特征(receiver operating characteristic,ROC)曲线、校准曲线、决策曲线进行评估,同时引入验证集数据对模型进行外部验证。结果建模集412例呼吸道感染患儿首次确诊后1年随访中出现RRTI为64例(15.53%)。多因素logistic回归分析表明,年龄(OR=0.666,95%CI:0.485~0.915)、家长对RRTI防控认知(OR=0.689,95%CI:0.526~0.903)是儿童RRTI的保护因素(P<0.05);过敏史(OR=2.809,95%CI:1.459~5.407)、C反应蛋白高(OR=2.296,95%CI:1.275~4.101)、DII高(OR=2.125,95%CI:1.242~3.637)、家庭成员吸烟(OR=1.467,95%CI:1.080~1.991)是儿童RRTI的危险因素(P<0.05)。基于上述因素构建儿童RRTI列线图模型,模型的ROC曲线下面积为0.867(95%CI:0.815~0.935),最佳截断值(阈概率)为0.37,此时的灵敏度为0.860、特异度为0.824;校准曲线分析的Brier指数为0.097(P>0.05);引入验证集进行模型外部验证发现,模型的ROC曲线下面积为0.851(95%CI:0.793~0.924)、灵敏度为0.850、特异度为0.805;校准曲线分析的Brier指数为0.112(P>0.05)。当决策曲线中阈概率值设为37.0%,建模集和验证集的临床获益率分别为54%、59%,表明预测模型具有临床有效性。结论基于年龄、过敏史、C反应蛋白、DII指数、家庭成员吸烟、家长对RRTI防控认知度建立的预测模型对儿童RRTI风险具有一定的预测价值。

Objective To analyze the factors affecting recurrent respiratory tract infection(RRTI)in children,to construct a predictive model,and to explore the predictive value of the model.Methods Using retrospective research,we collected the clinical data and evaluation results of Dietary Inflammatory Index(DII)of 589 children with respiratory tract infections who were treated in Suzhou Municipal Hospital from February to December 2021.The cases were randomly allocated to the modeling set(n=412)and the validation set(n=177)in a ratio of 7:3.Single and multiple logistic regression analyses were performed on the modeling set data to screen the factors influencing RRTI in the children,and a column chart prediction model was constructed.The performance and clinical practicality of the model were evaluated using receiver operating characteristic(ROC)curve,calibration curve and decision curve respectively.Validation set data were simultaneously introduced for external validation of the model.Results In the modeling set of 412 children with respiratory tract infections,64(15.53%)had RRTI during the one-year follow-up after their first diagnosis.Multivariate logistic regression analysis revealed that age(OR=0.666,95%CI:0.485-0.915)and parental awareness of RRTI prevention and control(OR=0.689,95%CI:0.526-0.903)were protective factors for RRTI in the children(P<0.05).Allergic history(OR=2.809,95%CI:1.459-5.407),high C-reactive protein(OR=2.296,95%CI:1.275-4.101),high DII(OR=2.125,95%CI:1.242-3.637)and family smoking(OR=1.467,95%CI:1.080-1.991)were risk factors for RRTI in the children(P<0.05).A column chart model for prediction of RRTI in the children was constructed based on the above-mentioned factors.The area under the ROC curve of the model was 0.867(95%CI:0.815-0.935),and the optimal cutoff value(threshold probability)0.37,with the sensitivity and specificity at this point being 0.860 and 0.824 respectively.The Brier index for calibration curve analysis was 0.097(P>0.05).Introducing validation sets for external model validation found that the area under the ROC curve of the model was 0.851(95%CI:0.793-0.924),with the sensitivity and specificity being 0.850 and 0.805 respectively.The Brier index for calibration curve analysis was 0.112(P>0.05).When the threshold probability value in the decision curve was set to 37.0%,the clinical benefit rates for the modeling and validation sets were 54%and 59%respectively,which indicated that the prediction model had clinical validity.Conclusion The prediction model constructed based on age,allergy history,C-reactive protein,DII,smoking of family members and parents’awareness of RRTI prevention and control has a certain predictive value for RRTI risk in the children.