| 白慧玲,宁可,赵小蝶,等.机械通气极早产儿早期持续肺动脉高压肺血管扩张剂使用的预测模型建立[J].安徽医药,待发表. |
| 机械通气极早产儿早期持续肺动脉高压肺血管扩张剂使用的预测模型建立 |
|
| 投稿时间：2026-04-12 录用日期：2026-05-11 |
| DOI： |
| 中文关键词: 肺动脉高压 极早产儿 预测模型 机器学习 |
| 英文关键词: |
| 基金项目: |
|
| 摘要点击次数: 108 |
| 全文下载次数: 0 |
| 中文摘要: |
| 目的 筛选机械通气极早产儿(胎龄<32周)早期(生后72h内)新生儿持续性肺动脉高压(PPHN)肺血管扩张剂使用的关键影响因素并构建预测模型?方法 采用单中心回顾性队列研究设计,收集2021年1月至2025年12月徐州医科大学附属医院新生儿重症监护室收治的326例胎龄<32周?需有创机械通气且入院72h内确诊为PPHN的极早产儿临床资料?以7:3比例将患儿随机分为训练集(n=229)和验证集(n=97),训练集按是否使用肺血管扩张剂分为治疗组与对照组?采用极端梯度提升(XGBoost)?随机森林(RF)?梯度提升机(GBM)三种机器学习算法平行筛选重要影响因素,分析3种机器学习方法 重叠覆盖 的风险因素,结合多因素Logistic回归验证独立危险因素并构建预测模型;通过ROC曲线(AUC值)?校准曲线?Hosmer-Lemeshow检验及决策曲线分析(DCA)评估模型性能?结果 三种机器学习模型交集筛选出7项关键因素,Logistic回归证实早发败血症?PS使用次数?OI?APTT为独立危险因素?模型训练集AUC=0.911,验证集AUC=0.916,校准良好?校准曲线及Hosmer-Lemeshow检验提示模型预测概率与实际发生频率高度一致,DCA显示净获益显著,灵敏度?特异度均超80%?结论 早期败血症?PS使用次数?OI?APTT是机械通气极早产儿早期PPHN干预需求的核心独立危险因素,基于此构建的预测模型具有优异的区分度?校准度及临床实用性,可用于早期识别高危患儿,优化干预时机? |
| 英文摘要: |
| Objective To screen the key influencing factors for the use of pulmonary vasodilators in very preterm infants (gestational age < 32 weeks) with early neonatal persistent pulmonary hypertension (PPHN, within 72 hours after birth) under mechanical ventilation and to construct a predictive model.Methods A single-center retrospective cohort study was conducted. Clinical data were collected from 326 very preterm infants with gestational age < 32 weeks who required invasive mechanical ventilation and were diagnosed with PPHN within 72 hours of admission, admitted to the Neonatal Intensive Care Unit of the Affiliated Hospital of Xuzhou Medical University from January 2021 to December 2025. The infants were randomly divided into a training set (n = 229) and a validation set (n = 97) at a ratio of 7:3. In the training set, patients were divided into a treatment group and a control group according to whether pulmonary vasodilators were used. Three machine learning algorithms, including extreme gradient boosting (XGBoost), random forest (RF), and gradient boosting machine (GBM), were applied in parallel to screen important influencing factors. Overlapping risk factors identified by the three machine learning methods were analyzed, and multivariate Logistic regression was used to verify independent risk factors and construct a predictive model. The model performance was evaluated using receiver operating characteristic (ROC) curve (AUC), calibration curve, Hosmer–Lemeshow test, and decision curve analysis (DCA).Results Seven key factors were screened from the intersection of the three machine learning models. Logistic regression confirmed that early-onset sepsis, number of pulmonary surfactant (PS) administrations, oxygenation index (OI), and activated partial thromboplastin time (APTT) were independent risk factors. The model achieved an AUC of 0.911 in the training set and 0.916 in the validation set, with favorable calibration. The calibration curve and Hosmer–Lemeshow test indicated high consistency between the predicted probability and the actual incidence. DCA showed significant net benefit, and both sensitivity and specificity exceeded 80%.Conclusions Early-onset sepsis, number of PS administrations, OI, and APTT are core independent risk factors for intervention requirements in very preterm infants with early PPHN under mechanical ventilation. The predictive model constructed based on these factors demonstrates excellent discrimination, calibration, and clinical practicability. It can be used for early identification of high-risk infants and optimization of intervention timing. |
|
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
