成人心脏瓣膜置换术后异体红细胞输注智能模型构建  

Construction of intelligent model for allogeneic red blood cell infusion after adult cardiac valve replacement

作　　者：邓波 杨眉[2] 谢弢[2] 王敏 厉忠逵[2] 夏英 黄海[1] DENG Bo;YANG Mei;XIE Tao;WANG Min;LI Zhongkui;XIA Ying;HUANG Hai(Guizhou Medical University,Guiyang,Guizhou 550001,China;Department of Blood Transfusion,Guizhou Provincial People’s Hospital,Guiyang,Guizhou 550002,China)

机构地区：[1]贵州医科大学,贵阳550001 [2]贵州省人民医院输血科,贵阳550002

出　　处：《重庆医学》2022年第16期2750-2754,2761,共6页Chongqing medicine

基　　金：贵州省创新人才团队资助项目(2019-5610)。

摘　　要：目的探索成人心脏瓣膜置换术后异体红细胞输注的相关因素,运用数学模型,量化术后异体红细胞的输注,优化输血治疗方案。方法回顾性分析2018年1月至2020年1月贵州省人民医院收治的1104例行心脏瓣膜置换术成人患者的临床资料,通过SPSS26.0分析结果,筛出建模的指标,运用Lasso方法建立智能模型。最后采用受试者工作特征曲线下面积(AUC)等指标对模型进行评价。结果术后输注红细胞患者与术后未输红细胞患者年龄、体重指数(BMI)、纽约心脏病协会分级、术中失血量、术中异体输注红细胞量、术后健康评分、术后心包引流量、术后纵隔引流量、术后第1次血红蛋白(Hb)、术后血细胞比容、术后血小板计数、术后输注血浆量比较,差异均有统计学意义(P<0.05),经Lasso方法筛选后最终决定7个指标建模(术后输注异体红细胞量=0.0251×BMI-0.0274×术中输注异体红细胞量+2.3339×10^(-4)×术中失血量+8.024×10^(-4)×心包引流量+4.0253×10^(-4)×纵隔引流量+0.0019×术后输注血浆量-0.0147×术后第1次Hb+2.2935)。模型在训练集和测试集的AUC均大于0.8。结论建立的术后输注异体红细胞量的预测模型具有较好的拟合精度和预测效果。该模型可促进心脏瓣膜置换术后患者实施输血方案的规范化,提供便捷的临床输血路径,避免经验性输血。Objective To investigate the related factors of allogeneic red blood cell infusion after adult heart valve replacement,and to quantize postoperative allogeneic red blood cell infusion by using the mathematic model and optimize the blood transfusion treatment scheme.Methods The clinical data of 1104 adult patients with heart valve replacement in Guizhou Provincial People’s Hospital from January 2018 to January 2020 were retrospectively analyzed.The modeling indexes were screened out by the SPSS26.0 analysis results,and the intelligent model was established by the Lasso method.Finally,the area under the receiver operating characteristic(ROC)curve and other indicators were used to evaluate the model.Results The differences in the age,BMI,NYHA stage,intraoperative blood loss amount,intraoperative red blood cell transfusion amount,postoperative health score,postoperative pericardium drainage amount,postoperative mediastinal drainage amount,postoperative Hb at the frst time,postoperative Hct,postoperative Plt count and postoperative plasma transfusion amount between the patients with red cells infusion and the patients without red cells infusion had statistical significance(P<0.01).After screening with the Lasso method,7 indicators were finally determined for modeling(postoperative allogeneic red blood cell infusion amount=0.0251×BMI-0.0274×total amount of intraoperative allogeneic red blood cell infusion+2.3339×10^(-4)×intraoperative blood loss amount+8.024×10^(-4)×pericardial drainage amount+4.0253×10^(-4)×mediastinal drainage amount+0.0019×postoperative plasma infusion amount-0.0147×postoperative Hb value+2.2935).The AUC values of the model in the training and the tests sets were 0.815 and 0.820,respectively.Conclusion The constructed predictive model of postoperative allogeneic red blood cell infusion has good fitting accuracy and predictive effect.This model can help to standardize the postoperative implement of blood transfusion protocol for the patients with heart valve replacement,provide a convenie

关 键 词：心脏瓣膜置换术 异体红细胞输注 Lasso方法 数学模型 

分 类 号：R446.61[医药卫生—诊断学]