基于CHAID决策树和Logistic回归的肺癌患者术后肺部并发症预测效果的研究  

Prediction of postoperative pulmonary complications in patients with lung cancer based on CHAID decision tree and logistic regression

作　　者：王新[1] 王炯杰[1] 王雷[1] 韩亮[2] 潘雪峰[1] 张淼[1] WANG Xin;WANG Jiongjie;WANG Lei;HAN Liang;PAN Xuefeng;ZHANG Miao(Department of Thoracic Surgery, Xuzhou Central Hospital, Xuzhou 221009, China)

机构地区：[1]江苏徐州徐州市中心医院胸外科,221009 [2]徐州市中心医院肿瘤内科,221009

出　　处：《临床肿瘤学杂志》2021年第10期898-902,共5页Chinese Clinical Oncology

基　　金：国家自然科学基金面上资助项目(81772479)。

摘　　要：目的比较CHAID决策树和二分类Logistic回归两种模型对肺癌术后肺部并发症的预测能力,为找出重点人群提供建议。方法回顾性分析徐州市中心医院胸外科2017年1月至2020年5月接受胸腔镜手术的肺癌患者352例。分别建立决策树模型和Logistic回归模型,比较两者的受试者工作特征曲线(ROC)的曲线下面积(AUC)、灵敏度、特异度和约登指数。结果术后共54例患者发生肺部并发症。决策树模型中,决策点所涉及到的因素是年龄、糖尿病、COPD;决策树模型的AUC为0.868(95%CI:0.828~0.901),灵敏度为79.63%,特异度为78.52%。Logistic回归模型中,年龄、糖尿病、COPD和术中出血量是术后肺部并发症发生的危险因素(OR>1,P<0.05);Logistic回归模型的AUC为0.886(95%CI:0.849~0.918),灵敏度为87.04%,特异度为81.54%。两模型的差异无统计学意义(Z=1.576,P=0.1150)。结论结合运用CHAID决策树和Logistic回归模型,可以有效筛选出肺癌术后肺部并发症的危险因素,有助于制定针对性防治策略,降低并发症的发生率。Objective To compare the prediction ability of CHAID decision tree model and binary Logistic regression model for postoperative pulmonary complications of lung cancer,and to provide suggestions for identifying key population.Methods 352 patients with lung cancer who received thoracoscopic surgery from January 2017 to May 2020 in thoracic Surgery department of Xuzhou Central Hospital were retrospectively analyzed.The decision tree model and Logistic regression model were established respectively,and the area under ROC curve(AUC),sensitivity,specificity and Yuden index were compared between them.Results A total of 54 patients had pulmonary complications.In the decision tree model,the factors involved in decision points are age,diabetes and COPD.The AUC of decision tree model was 0.868(95%CI:0.828-0.901),the sensitivity was 79.63%,and the specificity was 78.52%.In Logistic regression model,age,diabetes,COPD and intraoperative blood loss were risk factors for postoperative pulmonary complications(OR>1,P<0.05).The AUC of Logistic regression model was 0.886(95%CI:0.849-0.918),the sensitivity was 87.04%,and the specificity was 81.54%.There was no significant difference between the two models(Z=1.576,P=0.1150).Conclusion The combination of CHAID decision tree and Logistic regression model can effectively screen out the risk factors of postoperative pulmonary complications of lung cancer,which is conducive to the development of targeted prevention and treatment strategies and reduce the incidence of complications.

关 键 词：肺癌 肺部并发症 决策树模型 LOGISTIC回归模型 

分 类 号：R734.2[医药卫生—肿瘤]