摘要
目的探讨儿童肺炎合并塑形性支气管炎(PB)的危险因素,并构建决策树预测模型,评估其临床预测效能,为PB的早期识别提供参考。方法回顾性收集肺炎患儿533例作为训练集,另收集103例患儿作为同中心时间外验证集。若支气管镜下发现塑形物,则患儿确诊为PB。收集患儿一般资料、临床表现、临床体征、影像学表现及实验室指标等多维度临床指标共47项。通过单因素分析、Spearman秩相关分析及多因素Logistic回归模型筛选儿童肺炎合并PB的独立危险因素,并在此基础上构建决策树预测模型。采用受试者工作特征(ROC)曲线评估预测模型的预测价值,采用Bootstrap法重复抽样1000次进行模型的内部验证,采用Hosmer-Lemeshow检验评估模型的准确度,进行决策曲线分析(DCA)评估模型的临床实用性。在中心时间外验证集对模型的区分度、校准度及临床净获益进行外部验证。结果533例肺炎患儿中共有126例(23.64%)发生PB。发热热峰高、中性粒细胞百分比高、D-二聚体高、白蛋白低、免疫球蛋白A高、免疫球蛋白E高、铁蛋白高、≥2/3肺叶实变面积、有胸腔积液是儿童肺炎合并PB的独立危险因素(P均<0.05)。基于上述独立危险因素构建的决策树预测模型为5层结构,内部验证中,模型ROC曲线下面积(AUC)为0.860,Hosmer-Lemeshow检验提示模型校准度良好(P>0.05),DCA结果显示模型在较大阈值范围内(8%~100%)具有临床净获益。外部验证中,模型也有良好的区分度(AUC=0.926)与校准度(P>0.05),且临床净获益较高(9%~100%)。结论基于多维度临床指标筛选出的独立危险因素构建的PB决策树预测模型具有较好的预测效能,能为临床早期识别肺炎合并PB提供一定参考。
Objective To investigate the risk factors for pneumonia complicated with plastic bronchitis(PB)in chil⁃dren,to construct a decision tree prediction model,and to evaluate its clinical predictive performance,providing a refer⁃ence for early identification of PB.Methods A total of 533 children diagnosed with pneumonia were retrospectively en⁃rolled as the training cohort,and an additional 103 children were collected as a same-center temporal external validation cohort.PB was diagnosed when bronchial casts were observed by bronchoscopy.Forty-seven multidimensional clinical variables were collected,including general characteristics,clinical manifestations,physical signs,imaging findings,and laboratory indicators.Univariate analysis,Spearman rank correlation analysis,and multivariate Logistic regression were used to identify independent risk factors for PB.Based on these factors,a decision tree prediction model was constructed.The predictive performance of the model was evaluated using receiver operating characteristic(ROC)curves.Internal vali⁃dation was performed using the Bootstrap method with 1,000 resamples.Model calibration was assessed by the Hosmer-Lemeshow test,and clinical utility was evaluated by decision curve analysis(DCA).Results Among the 533 children with pneumonia,126 cases(23.64%)developed PB.Multivariate Logistic regression analysis showed that higher maximum body temperature,higher neutrophil percentage,elevated D-dimer,decreased albumin,increased immunoglobulin A,increased immunoglobulin E,elevated serum ferritin,consolidation involving≥2/3 of a lung lobe,and pleural effusion were independent risk factors for PB(all P<0.05).A five-layer decision tree model was constructed based on these inde⁃pendent risk factors.In internal validation,the area under the ROC curve(AUC)was 0.860,indicating good discrimina⁃tive ability;the Hosmer-Lemeshow test demonstrated good calibration(P>0.05),and DCA showed a positive net clinical benefit across a wide range of threshold probabilities(8%-100%).In external validation,the model also exhibited good discrimination(AUC=0.926)and calibration(both P>0.05),with favorable net clinical benefit over threshold probabili⁃ties of 9%-100%.Conclusion The decision tree prediction model for plastic bronchitis constructed based on indepen⁃dent risk factors derived from multidimensional clinical indicators demonstrates good predictive performance and may pro⁃vide a useful reference for early identification of PB in children with pneumonia.
作者
谢毓
许示沂
张俊艳
崔光焕
刘爱红
XIE Yu;XU Shiyi;ZHANG Junyan;CUI Guanghuan;LIU Aihong(Department of Pediatric Medicine,Shanxi Medical University,Taiyuan 030012,China;不详)
出处
《山东医药》
2026年第1期38-43,共6页
Shandong Medical Journal
基金
山西省中医药管理局科研课题项目(2024ZYYB033)。
