摘要
目的探讨早产儿肺功能异常的危险因素,构建预测模型并绘制列线图。方法回顾性纳入2023年3月至2024年2月在温州医科大学附属第一医院新生儿重症监护病房治愈出院的胎龄<34周的85例早产儿为研究对象,纠正胎龄至37~42周行肺功能检查,根据肺功能检查结果,分为肺功能正常或轻度异常组58例和肺功能中重度异常组27例。采用机器学习最小绝对收缩和选择算子(LASSO)回归和极致梯度提升(XGBoost)算法构建模型,通过ROC曲线、校准曲线、临床决策曲线和临床影响曲线评估模型效能,并绘制列线图。结果LASSO回归模型筛选出的变量为出生体重,验证集AUC为0.933。XGBoost模型筛选出的变量为出生体重、支气管肺发育不良(BPD)和新生儿呼吸窘迫综合征(RDS),验证集AUC为0.967。校准曲线显示两种模型预测概率与实际观察结果高度一致,提示预测与观察匹配度良好;临床决策曲线和临床影响曲线显示XGBoost模型在不同阈值下均能提供更高的临床净获益。XGBoost模型列线图将出生体重、BPD和新生儿RDS整合为风险评分,可直观预测早产儿肺功能异常发生率。结论出生体重、BPD及新生儿RDS是早产儿肺功能异常的独立预测因素,XGBoost模型具有较高的临床实用性。
Objective To identify risk factors for pulmonary function abnormalities in preterm infants,and to develop and validate a corresponding prediction model with a nomogram.Methods This retrospective study included 85 preterm infants with a gestational age<34 weeks who were discharged from the neonatal intensive care unit of the First Affiliated Hospital of Wenzhou Medical University between March 2023 and February 2024.Pulmonary function tests were performed at a corrected gestational age of 37-42 weeks.Based on the test results,infants were categorized into a normal-to-mild abnormality group(58 cases)and a moderate-to-severe abnormality group(27 cases).Machine learning algorithms,including least absolute shrinkage and selection operator(LASSO)regression and extreme gradient boosting(XGBoost),were employed to construct prediction models.Model performance was assessed using ROC curves,calibration curves,clinical decision curves,and clinical impact curves.A nomogram was subsequently developed.Results The LASSO regression model identified birth weight as the sole significant predictor,yielding an AUC of 0.933 in the validation set.The XGBoost model identified birth weight,bronchopulmonary dysplasia(BPD),and neonatal respiratory distress syndrome(RDS)as the sole significant predictors,achieving a higher AUC of 0.967 in the validation set.Calibration curves indicated good agreement between predicted probabilities and actual observations for both models.Clinical decision curves and clinical impact curves demonstrated that the XGBoost model provided greater clinical net benefit across various threshold probabilities.The nomogram derived from the XGBoost model integrated birth weight,BPD,and neonatal RDS into a risk score,enabling intuitive estimation of the probability of pulmonary function abnormalities in preterm infants.Conclusion Birth weight,BPD,and neonatal RDS are independent predictors of pulmonary function abnormalities in preterm infants.The XGBoost model demonstrates high clinical utility.
作者
张渊博
徐昌富
朱艳可
王丹
王楸
ZHANG Yuanbo;XU Changfu;ZHU Yanke;WANG Dan;WANG Qiu(Department of Pediatrics,the First Affiliated Hospital of Wenzhou Medical University,Wenzhou 325000,China;不详)
出处
《浙江医学》
2026年第4期373-379,I0002,共8页
Zhejiang Medical Journal
基金
温州市科技计划项目(Y20180255)。
关键词
早产儿
肺功能异常
新生儿支气管肺发育不良
Preterm infants
Pulmonary function abnormalities
Bronchopulmonary dysplasia
