采用决策树和logistic回归模型分析全髋关节置换术后手术部位感染的影响因素

Influencing factors of surgical site infection after total hip arthroplasty using decision tree and logistic regression model

目的 采用决策树与logistic回归模型分析髋关节置换术后手术部位感染发生的影响因素,为制订手术部位感染控制策略提供依据。方法 回顾性收集2012年1月至2018年12月在温州市人民医院行全髋关节置换术患者的临床资料,以全髋关节置换术后发生手术部位感染的28例患者为病例组,以相同时间段内收治的未发生髋关节置换术后手术部位感染的576例患者为对照组,采用logistic回归和决策树模型分析手术部位感染发生相关影响因素。结果 logistic回归模型和决策树模型均筛选出4个危险因素,即手术时间≥180 min、术中出血量≥200 mL、抗生素使用种类≥2种以及术前血清白蛋白<35 g/L。决策树CHAID法分析结果显示,手术时间是影响手术部位感染发生最主要的因素。结论 CHAID决策树模型与logistic回归可互为补充,从各个方面描述影响手术部位感染发生的因素及作用,为进一步制定预防与控制手术部位感染的措施提供依据和参考。

Objective To investigate risk factors for surgical site infection after total hip arthroplasty based on exhaustive CHAID decision tree and logistic regression model,so as to provide the basis for the development of control strategy of surgical site infection(SSI).Methods A retrospective analysis was performed on the patients undergoing total hip arthroplasty from January 2012 to December 2018 in Wenzhou People's Hospital.Twenty four patients with surgical site infection after total hip arthroplasty were selected as the case group,and 576 patients who did not have surgical site infection after total hip arthroplasty in the same period were selected as the control group.Logistic regression and decision tree model were used to determine influencing factors for surgical site infection.Results Four risk factors were found by logistic regression model and decision tree model,including the operation time no less than 180 min,the amount of intraoperative bleeding more than 200 mL,using at least 2 kinds of antibiotics,and serum albumin less than 35`g/L.The results of CHAID decision tree analysis showed that the operation time was the main factor affecting the SSI.Conclusion In order to provide evidence and reference for prevention and control measure of surgical site infection,it is essential to combine these two different methods to describe influencing factors of surgical site infection from different aspects.