基于logistic回归模型和CHAID决策树模型的单肺通气期间低氧血症影响因素分析  

Analysis of the influencing factors of hypoxemia during one-lung ventilation based on logistic regression model and Chi-squared automatic interaction detector decision tree model

作　　者：朱修锦 岳维[2] 王静[1] 李晶[1] Zhu Xiujin;Yue Wei;Wang Jing;Li Jing(School of Anesthesiology,Shanxi Medical University,Taiyuan 030001,China;Department of Anesthesiology,the Second Hospital of Shanxi Medical University,Taiyuan 030001,China)

机构地区：[1]山西医科大学麻醉学院,太原030001 [2]山西医科大学第二医院麻醉科,太原030001

出　　处：《国际麻醉学与复苏杂志》2023年第12期1301-1308,共8页International Journal of Anesthesiology and Resuscitation

摘　　要：目的借助logistic回归模型和X^(2)自动交互检测法(Chi-squared automatic interaction detector,CHAID)决策树模型探讨单肺通气(one-lung ventilation,OLV)期间发生低氧血症的影响因素及其相互关系.方法选取2021年10月-2022年6月在山西医科大学第二医院行OLV胸科手术的患者142例,收集患者性别、年龄、BMI、Hb水平、吸烟史、手术侧(左/右)、第一秒用力呼气量(forced expiratory volume in first second,FEV1)、第一秒用力呼气量占用力肺活量百分率(percentage of forced expi-ratory volume in first second to forced vital capacity,FEV1/FVC)、自主通气时PaO_(2)、侧卧位10 min时PaO_(2)、侧卧位10 min时依赖肺和非依赖肺呼气末二氧化碳分压差值(d-PETCO_(2))等资料.根据OLV期间是否发生过低氧血症将患者分为低氧血症组(23例)和未发生低氧血症组(119例).对两组患者的变量资料进行单因素关联性分析,其中P<0.05的变量用于构建logistic回归模型和CHAID决策树模型,分析胸科手术患者OLV期间发生低氧血症的影响因素及其相互关系.结果142例行OLV胸科手术的患者中,低氧血症的发生率为16.20%(23/142),单因素关联性分析显示BMI、手术侧(左/右)、FEV1/FVC、侧卧位10 min时PaO_(2)和d-PETCO_(2)与低氧血症的发生有一定的相关性.进一步logistic建模结果提示:侧卧位10 min时d-PETCO_(2)[比值比(odds ra-tio,OR)0.52,95%CI 0.36~0.75,P<0.001]、FEV1/FVC(OR 0.82,95%CI 0.70~0.97,P=0.016),侧卧位10 min时PaO_(2)(OR=0.98,95%CI0.96~1.00,P=0.018)对低氧血症发生的影响有统计学意义.CHAID决策树结果提示:侧卧位10 min时d-PETCO_(2)、FEV1/FVC和BMI是低氧血症结局发生的影响因素,树的首层是根据患者侧卧位10 min时d-PETCO_(2)的水平划分,树的第二层对于侧卧位10 min时d-PETCO_(2)在(1,4]之间或者>4的患者分别根据BMI和FEV1/FVC水平进行划分,侧卧位10 min时d-PETCO_(2)和FEV1/FVC、侧卧位10 min时d-PETCO_(2)和BMI之间分别存在相互作�Objective To discuss the influencing factors of hypoxemia during one-lung ventilation(OLV)by logistic regression model and Chi-squared automatic interaction detector(CHAID)decision tree model and their relationship.Methods A total of 142 patients who underwent OLV thoracic surgery in the Second Hospital of Shanxi Medical University from October 2021 to June 2022 were selected.Their gender,age,body mass index(BMI),hemoglobin(Hb)level,smoking history,operation side(left/right),forced expiratory volume in first second(FEV1),percentage of forced expiratory volume in first second to forced vital capacity ratio(FEV1/FVC),arterial partial pressure of oxygen(PaO_(2))during spontaneous ventilation,PaO_(2) in the lateral position for 10 min,and the difference in end-expiratory partial pressure of carbon dioxide(d-PETCO_(2))between the dependent lung and the non-dependent lung in the lateral position for 10 min were collected.According to the occurrence of hypoxemia during OLV,the patients were divided into two groups:a hypoxemia group(n=23)and a non-hypoxemia group(n=119).Univariate correlation analysis was performed using the variable data of the two groups,where those with P<0.05 were used to construct a logistic regression model and a CHAID decision tree model,in order to explore the influencing factors of hypoxemia during OLV in patients undergoing thoracic surgery.Results Among 142 patients undergoing OLV thoracic surgery,the incidence of hypoxemia was 16.20%(23/142).Univariate correlation analysis indicated that BMI,operation side(left/right),FEV1/FVC,and PaO_(2) and d-PETCO_(2) values in the lateral position for 10 min were related to hypoxemia.Furthermore,logistic regression analysis showed that d-PETCO_(2) in the lateral position for 10 min[odds ratio(OR)0.52,(95%confidence interval(CI)0.36,0.75),P<0.001],FEV1/FVC[OR 0.82,(95%CI 0.70,0.97),P=0.016],and PaO_(2) in the lateral position for 10 min[OR 0.98,(95%CI 0.96,1.00),P=0.018]had statistical effect on the occurrence of hypoxemia.The CHAID decision tree suggested

关 键 词：单肺通气 低氧血症 临床预测模型 LOGISTIC回归 决策树 

分 类 号：R614[医药卫生—麻醉学]