基于Lasso-Cox回归筛选影响心脏瓣膜病患者术后ICU时间的因素

Based on Lasso-Cox regression to analyze influence factors of the staying time in ICU after heart valvular surgery

目的采用Lasso-Cox回归法分析影响心脏瓣膜病患者术后ICU时间的因素。方法采用回顾性方法连续收集2014年1月至2015年12月930例我院行心脏瓣膜手术患者的临床资料,采用Lasso-Cox回归法筛选因素建立2个模型:Pre模型(纳入术前因素)和Pro模型(纳入术前和术中因素),通过受试者工作特征曲线(ROC)的曲线下面积(AUC)验证纳入不同因素对ICU时间预测的准确性。结果患者的ICU中位时间为70.5 h,Pre模型纳入影响心脏瓣膜病患者术后ICU时间的术前因素包括年龄、术前总胆红素、高血压病、术前肌酐清除率,Pro模型纳入影响心脏瓣膜病患者术后ICU时间的术前及术中因素包括年龄、术前总胆红素、高血压病、术前肌酐清除率、术前红细胞计数(RBC)、体外循环时间、主动脉阻断时间、三尖瓣置换术;Pre模型在24、48、72 h转出ICU端点的AUC值分别为0.6144(95%CI:0.5684~0.6611)、0.6543(95%CI:0.6251~0.7056)、0.6657(95%CI:0.6203~0.7056),Pro模型分别为0.7021(95%CI:0.6622~0.7403)、0.7123(95%CI:0.6690~0.7442)、0.7424(95%CI:0.7128~0.7833)。结论年龄、术前总胆红素、高血压病、术前肌酐清除率、术前RBC、体外循环时间、主动脉阻断时间、三尖瓣置换术均可影响心脏瓣膜病患者术后ICU时间;联合术前及术中因素能更准确地预测心脏瓣膜病患者术后ICU时间。

Objective To analyze the factors affecting the ICU staying time after heart valve surgery by Lasso-Cox regression.Methods The clinical data of 930 consecutive cases of patients with heart valve surgery between January 2014 and December 2015 in our hospital were collected.Influence actors were filtered by the Lasso-Cox regression and two models were established:the Pre model(using preoperative influence factors)and the Pro model(using preoperative and intraoperative influence factors).The accuracy of the prediction of ICU staying time for both of the Pre and Pro models was assessed and compared by area under curve(AUC)of the receiver operating characteristic curve(ROC).Results The median staying time in ICU was 70.5 h.Influence factors of the Pre model were age,preoperative total bilirubin,preoperative hypertension,preoperative creatinine clearance,influence factors of the Pro model were age,preoperative total bilirubin,preoperative hypertension,preoperative creatinine clearance,preoperative red blood cell count,cardiopulmonary bypass time,aortic cross clamp time,tricuspid valve replacement.The AUCs of the Pre model at 24 h,48 h and 72 h transferring from ICU were 0.6144(95%CI:0.5684-0.6611),0.6543(95%CI:0.6251-0.7056),0.6657(95%CI:0.6203-0.7056),the AUCs of the Pro model at 24 h,48 h and 72 h transferring from ICU were 0.7021(95%CI:0.6622-0.7403),0.7123(95%CI:0.6690-0.7442),0.7424(95%CI:0.7128-0.7833).Conclusions Influence factors of the ICU staying time after heart valve surgery are age,preoperative total bilirubin,preoperative hypertension,preoperative creatinine clearance,preoperative red blood cell count,cardiopulmonary bypass time,aortic cross clamp time,tricuspid valve replacement.Adding the intraoperative influence factors can increase the accuracy of the prediction of ICU staying time after heart valvular surgery.