心脏术后患者机械通气时间延长风险预测模型的构建及验证

Construction of Risk Prediction Model for Prolonged Mechanical Ventilation in Patients after Cardiac Surgery

目的探讨心脏术后患者发生机械通气时间延长的危险因素,并构建风险预测模型。方法2020年7月—2021年5月,选取武汉某三级甲等医院心脏大血管外科行心脏手术的471例住院患者作为研究对象,根据机械通气时间是否≥48 h分为延长组357例和非延长组114例,将2组临床指标进行比较,筛选出独立危险因素,并利用Logistic回归进行风险预测模型的构建。2021年6-11月,选取行心脏手术的81例患者作为验证组研究对象。应用R软件构建心脏术后机械通气时间延长列线图模型,并进行验证。结果最终纳入年龄(OR=1.035)、BMI(OR=1.143)、术前中性粒细胞计数(OR=1.979)、术中使用抗菌药(OR=0.520)、手术时长(OR=1.548)、体外循环时长(OR=1.291)6个因素构建心脏术后机械通气时间延长风险预测模型。内部验证:Calibration图示校准曲线与理想曲线几乎重合,ROC曲线下面积为0.848,灵敏度和特异度分别为0.772和0.790。外部验证:Calibration图示校准曲线在理想曲线附近,ROC曲线下面积为0.822,灵敏度和特异度分别为0.766和0.824。结论本研究基于年龄、BMI、术前中性粒细胞计数、术中使用抗菌药、手术时长、体外循环时长6项独立危险因素,构建的心脏术后机械通气时间延长风险预测模型具有良好的一致性和预测效能,可为临床个性化预测机械通气时间延长风险提供参考。

Objective To explore the risk factors of prolonged mechanical ventilation(PMV)in patients after cardiac surgery,and to construct a risk prediction model.Methods A total of 471 patients were selected from July,2020 to May,2021 in department of cardiovascular surgery in a tertiary grade-A hospital in Wuhan,and divided into PMV group(n=357)and non-PMV group(n=114)according to the mechanical ventilation time.The risk factors of PMV were screened out,and the risk prediction model was constructed through logistic regression.Eight-one patients from June to November,2021 were selected to verify the predictive value of the model.A nomogram was developed by R software and validated to predict the risk factors of PMV.Results Age(OR=1.035),BMI(OR=1.143),neutrophil count before surgery(OR=1.979),intraoperative use of antibiotic(OR=0.520),length of operation(OR=1.548),and duration of cardiopulmonary bypass(OR=1.291)were the independent risk factors of PMV.A prediction model was constructed by these 6 factors,which was internally verified by the modeling group.The calibration curve of the calibration chart was close to the ideal curve.The area under the ROC curve was 0.848.The sensitivity and specificity were 0.772,and 0.790.respectively.The external verification showed that the calibration curve of the calibration chart was near the ideal curve,and the area under the ROC curve was 0.822.The sensitivity and specificity were 0.766 and 0.824,respectively.Conclusion The risk prediction model for PMV in patients after cardiac surgery has good discrimination and accuracy.It can provide scientific guidance for individualized prevention of PMV.