不同机器学习模型预测严重创伤患者院内不良结局的效能比较  

Efficacy comparison of different machine learning models to predict adverse inhospital outcome in patients with severe trauma

作　　者：彭驰 杨帆 高翔 许硕贵 金志超 Peng Chi;Yang Fan;Gao Xiang;Xu Shuogui;Jin Zhichao(Department of Military Health Statistics,Faculty of Health Services,Naval Medical University,Shanghai 200433,China;Medical Company,Chinese PLA Unit 73676,Wuxi 214400,China;Department of Pathology,First Affiliated Hospital,Army Medical University,Chongqing 400038,China;Department of Emergency and Trauma,First Affiliated Hospital,Navy Medical University,Shanghai 200433,China)

机构地区：[1]海军军医大学卫生勤务学系卫生统计学教研室,上海200433 [2]中国人民解放军73676部队卫生连,无锡214400 [3]陆军军医大学第一附属医院病理科,重庆400038 [4]海军军医大学第一附属医院战创伤急救中心,上海200433

出　　处：《中华创伤杂志》2023年第6期545-550,共6页Chinese Journal of Trauma

基　　金：海军军医大学“三航”人才计划。

摘　　要：目的比较不同机器学习模型利用院前数据对严重创伤患者院内不良结局的预测效能。方法采用回顾性队列研究分析2017年1月至2018年12月美国国家创伤数据库(NTDB)中100135例严重创伤患者的临床资料,其中男69644例,女30480例(性别变量缺失11例);年龄16~89岁[(50.1±21.1)岁]。临床特征包括人口学信息(性别、年龄)、创伤类型(钝性伤或穿透伤)、院前时间[急救医疗服务(EMS)反应时间、EMS现场时间和EMS转运时间]、院前生命体征(收缩压、脉率、呼吸频率和血氧饱和度)、创伤评分[格拉斯哥昏迷评分(GCS)、损伤严重度评分(ISS)]。将原始数据按入院年份分为训练集(2017年)和测试集(2018年)。其中训练集50429例,测试集49706例,按有无发生不良结局将患者分为无不良事件发生组(94526例)和不良事件发生组(5609例)。训练集中不良事件组为2808例,测试集中不良事件组为2801例。所有模型均基于训练集构建,采用神经网络(NNET)、朴素贝叶斯(NB)、梯度提升树(GBM)、自适应增强机(Ada)、随机森林(RF)、袋装树(BT)、分类增强机(CatBoost)和极度梯度提升(XGB)8种机器学习算法根据患者临床特征构建严重创伤患者临床结局的预测模型。根据预测模型的灵敏度、特异度、受试者工作特征(ROC)的曲线下面积(AUC)和Hosmer-Lemeshow拟合优度检验来评价模型预测效能。结果NNET、NB、GBM、Ada、RF、BT、CatBoost和XGB模型在测试集中的灵敏度分别0.84,0.83,0.27,0.79,0.83,0.81,0.62,0.78;特异度分别为0.79,0.76,0.81,0.79,0.79,0.74,0.83,0.79;AUC分别为0.89(95%CI 0.88,0.90),0.86(95%CI 0.85,0.87),0.54(95%CI 0.53,0.55),0.86(95%CI 0.85,0.87),0.88(95%CI 0.88,0.90),0.83(95%CI 0.82,0.85),0.77(95%CI 0.76,0.79),0.86(95%CI 0.85,0.87),其中NNET模型的区分度最佳。NNET模型和NB模型的校准度也表现出良好的性能,Hosmer-Lemeshow拟合优度检验P值>0.05。结论NNET模型对严重创伤患者院内不良结局的预�Objective To compare the predictive performance of different machine learning models using pre-hospital data to predict adverse inhospital outcome in patients with severe trauma.Methods A retrospective cohort study was conducted to analyze the clinical data of 100135 patients with severe trauma from the National Trauma Data Bank(NTDB)from January 2017 to December 2018.There were 69644 males and 30480 females apart from 11 patients with missing gender information,with the range age of 16-89 years[(50.1±21.1)years].Clinical characteristics included demographic information(sex and age),trauma type(blunt or penetrating trauma),pre-hospital time[emergency medical services(EMS)response time,EMS scene time,and EMS transport time],pre-hospital vital signs(systolic blood pressure,pulse rate,respiratory rate,and oxygen saturation),trauma score[Glasgow coma score(GCS)and injury severity score(ISS)].The original data were divided into the training set(in the year 2017)and the testing set(in the year 2018)according to the year of admission,including 50429 patients in the training set and 49706 patients in the testing set.The patients were classified into non-adverse outcome group(n=94526)and adverse outcome group(n=5609),according to whether they had an adverse outcome or not.There were 2808 patients with adverse outcome in the training set and 2801 patients with adverse outcome in the testing set.All models were built based on the training set.Eight machine learning algorithms consisting of neural network(NNET),naive Bayes(NB),gradient boosting machine(GBM),adaptive boosting(Ada),random forest(RF),bagging tree(BT),categorical boosting(CatBoost)and extreme gradient boosting(XGB)were used to construct prediction models for clinical outcomes among patients with severe trauma based on their clinical features.Models were evaluated according to the sensitivity,specificity,area under the receiver operating characteristic(ROC)curve(AUC)and Hosmer-Lemeshow goodness-of-fit test.Results Of the NNET,NB,GBM,Ada,RF,BT,CatBoost and XGB mo

关 键 词：机器学习 急救医疗服务 创伤和损伤 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] R641[自动化与计算机技术—控制科学与工程]