基于logistic回归和决策树法预测肿瘤合并胆管占位患者ERCP术后并发胰腺炎的研究  

Prediction of Postoperative Pancreatitis in Patients with Tumor and Bile Duct Occupation Based on Logistic Regression and Decision Tree Method

作　　者：杨磊 YANG Lei(Endoscopic Diagnosis and Treatment Center,Henan Cancer Hospital,Zhengzhou 450000,China)

机构地区：[1]河南省肿瘤医院内窥镜诊断治疗中心,河南郑州450000

出　　处：《河南医学研究》2024年第9期1578-1582,共5页Henan Medical Research

摘　　要：目的基于logistic回归和决策树法,研究肿瘤合并胆管占位患者在内镜下逆行胰胆管造影术(ERCP)术后并发胰腺炎的影响因素。方法选择2020年3月至2023年3月河南省肿瘤医院收治的149例下段胆管癌合并胆管占位患者,均接受ERCP术治疗。入院后,收集患者的临床资料,根据患者术后是否并发胰腺炎,将其分为发生组与未发生组,并对比两组患者的临床资料。分别采用logistic回归和决策树法分析患者ERCP术后并发胰腺炎的影响因素,采用受试者工作特征(ROC)曲线分析logistic回归分析与决策树模型分析的结果。结果149例患者术后共有29例(19.46%)并发胰腺炎,据此将其分为发生组(n=29)与未发生组(n=120)。两组患者的性别、年龄、体重指数(BMI)、高血压、糖尿病、吸烟、饮酒、总胆固醇(TC)、甘油三酯(TG)、低密度脂蛋白胆固醇(LDL-C)、高密度脂蛋白胆固醇(HDL-C)、空腹血糖(FPG)、淀粉酶(AMY)等一般资料相比,差异无统计学意义(P>0.05);发生组患者的急性生理学与慢性健康状态评分系统Ⅱ评分、胰腺炎病史、胰腺显影、存在插管困难与未发生组相比,差异有统计学意义(P<0.05)。logistic多因素回归分析结果显示,胰腺炎病史、胰腺显影、存在插管困难是ERCP术后并发胰腺炎的危险因素(P<0.05)。决策树模型结果显示,胰腺炎病史、胰腺显影、存在插管困难均是影响ERCP术后并发胰腺炎的危险因素。logistic回归和决策树模型的分类准确率分别为85.37%和90.72%,准确度均较好。ROC曲线结果显示,logistic回归的曲线下面积为0.798,决策树模型的曲线下面积为0.843(P<0.05)。结论临床针对存在胰腺炎病史、胰腺显影、插管困难的下段胆管癌合并胆管占位患者,可给予有效治疗及干预;将logistic回归和决策树法联合应用可从不同层面发现下段胆管癌合并胆管占位患者在ERCP术后并发胰腺炎的影响因素,从而加强早期Objective To study the influencing factors of complicated pancreatitis after endoscopic retrograde cholangiopancreatography(ERCP)based on logistic regression and decision tree method.Methods A total of 149 patients with lower segment cholangiocarcinoma complicated with bile duct space admitted to Henan Cancer Hospital from March 2020 to March 2023 were selected to receive ERCP.After admission,the clinical data of the patients were collected.According to whether the patients had postoperative pancreatitis,they were divided into the occurrence group and the non-occurrence group,and the clinical data of the two groups were compared.Logistic regression and decision tree were used to analyze the influencing factors of postoperative pancreatitis in patients with ERCP,and receiver operating characteristic(ROC)curve was used to analyze the results of logistic regression analysis and decision tree model analysis.Results A total of 29(19.46%)of the 149 patients with postoperative pancreatitis were divided into the occurrence group(n=29)and non-occurrence group(n=120).Gender,age,body mass index(BMI),hypertension,diabetes,smoking,alcohol consumption,total cholesterol(TC),triglyceride(TG),low density lipoprotein cholesterol(LDL-C),high density lipoprotein cholesterol(HDL-C),fasting blood glucose(FPG),amylase(AMY)and other general data of the two groups were compared.There were no statistical difference(P>0.05).The acute physiology and chronic health evalutionⅡ(APACHEⅡ)score,history of pancreatitis,pancreatic imaging and intubation difficulties of patients in the occurrence group were statistically significant compared with those in the non-occurrence group(P<0.05).Logistic regression analysis showed that history of pancreatitis,pancreatic development and difficulty in intubation were risk factors for pancreatitis after ERCP surgery(P<0.05).The results of decision tree model showed that history of pancreatitis,pancreatic development and intubation difficulties were all risk factors for postoperative pancreatitis after ERCP

关 键 词：LOGISTIC回归 决策树法 胆管占位 内镜下逆行胰胆管造影术 胰腺炎 

分 类 号：R657[医药卫生—外科学]