摘要
【目的】探索在全血细胞计数(CBC)基础上结合临床因素建立分类模型,以预测严重的呼吸道感染人类腺病毒(HAdV)儿童患者。【方法】收集2023年11月至2024年9月期间确诊为HAdV感染的儿童患者的CBC参数及其相关临床资料。运用主成分分析和随机森林模型分析来选择严重病例的潜在的预测因素。【结果】共纳入了668例儿童患者,其中564例作为训练队列,104例作为验证队列。定义肺炎或/及热程≥5 d者(PorPF)为严重病例,通过主成分分析及特征重要性检验(Mean Decrease Gini值),单核细胞比例(PMono)、红细胞计数(RBC)和血小板计数(PLT)是为纳入病例CBC中最为重要的参数;经过Logistic回归分析显示,氧疗(OR=4.367,95%CI:1.568~12.161)及呼吸做功增加(OR=3.904,95%CI:2.146~7.101)是PorPF的相对危险因素,而较高的PMono(R=0.696,95%CI:0.640~0.757)、RBC(OR=0.201,95%CI:0.124~0.325)、PLT(OR=0.990,95%CI:0.987~0.994)为保护因素;PMono作为预测PorPF的指标时,其受试者工作特征曲线下面积(AUC)分别为:0.648、0.705;运用随机森林模型构建了包含4个风险因素[PMono、RBC、PLT、红细胞压积(HCT)]的分类模型以区分PorPF病例和普通病例(AUC分别为:0.688、0.768)。【结论】PMono、RBC、PLT可以作为儿童HAdV感染后出现肺炎或热程延长病例的CBC特征性指标。利用PMono、RBC、PLT及HCT构建风险因素模型可以较为简单准确预测儿童HAdV感染严重病例。
【Objective】To develop a classification model based on complete blood count(CBC)parameters combined with clinical factors to predict severe respiratory infections caused by Human adenovirus(HAdV)in pediatric patients.【Methods】From September 2023 to September 2024,the CBC parameters and related clinical data from pediatric patients diagnosed with HAdV infection were collected.Principal component analysis and random forest models were used to identify potential predictors of severe cases.【Results】A total of 668 pediatric patients were included,with 564 cases assigned to the training cohort and 104 cases to the validation cohort.Severe cases were defined as pneumonia and/or fever lasting≥5 days(pneumonia or prolonged fever,PorPF).Principal component analysis and feature importance analysis(Mean Decrease Gini value)identified the monocytosis ratio(PMono),red blood cell count(RBC),and platelet count(PLT)as the most critical CBC parameters.Logistic regression analysis revealed that oxygen therapy(OR=4.367,95%CI:1.568–12.161)and increased work of breathing(OR=3.904,95%CI:2.146–7.101)were relative risk factors for PorPF.Meanwhile,higher PMono(OR=0.696,95%CI:0.640–0.757),RBC(OR=0.201,95%CI:0.124–0.325),and PLT(OR=0.990,95%CI:0.987–0.994)were protective factors.When PMono was used as a predictive marker for PorPF,the area under the receiver operating characteristic curve(AUC)was 0.648 and 0.705,respectively.A random forest model incorporating four risk factors[PMono,RBC,PLT,and hematocrit(HCT)]was constructed to classify PorPF and general cases,achieving AUCs of 0.688 and 0.768,respectively.【Conclusions】PMono,RBC,and PLT may serve as characteristic CBC indicators for predicting pneumonia or prolonged fever in children with HAdV infection.A risk factor model built using PMono,RBC,PLT,and HCT offers a relatively simple and accurate approach to predicting severe cases in pediatric HAdV infections.
作者
钟俊炎
李俊祥
黄美
汪月娟
刘罗慧
陈晓辉
曹敏
ZHONG Junyan;LI Junxiang;HUANG Mei;WANG Yuejuan;LIU Luohui;CHEN Xiaohui;CAO Min(Department of Emergency,Shenzhen Maternity and Child Healthcare Hospital,Southern Medical University,Shenzhen 518028,China;Department of Laboratory Medicine,Shenzhen Maternity and Child Healthcare Hospital,Southern Medical University,Shenzhen 518028,China;Department of Pediatrics,Shenzhen Maternity and Child Healthcare Hospital,Southern Medical University,Shenzhen 518028,China)
出处
《中山大学学报(医学科学版)》
北大核心
2025年第5期889-898,共10页
Journal of Sun Yat-Sen University:Medical Sciences
关键词
人类腺病毒
全血细胞计数
儿童
肺炎
发热
Human adenovirus
complete blood count
pediatrics
pneumonia
fever
