Logistic回归和决策树模型在手术患者住院超30天影响因素分析中的应用  

Application of Logistic Regression and Decision Tree Models in the Analysis of Influencing Factors of Surgical Patients Hospitalized for More Than 30 Days

作　　者：刘萍 施辉[1] 吴梦平 张骞峰[1] Liu Ping;Shi Hui;Wu Mengping;Zhang Qianfeng(The First People’s Hospital of Lianyungang,Lianyungang 222000,Jiangsu Province,China;不详)

机构地区：[1]连云港市第一人民医院,江苏省连云港市222000

出　　处：《中国病案》2024年第3期61-65,共5页Chinese Medical Record

基　　金：江苏省医院协会2021年度医院管理创新研究立项课题(JSYGY-3-2021-435);连云港市科协2022年度软课题研究项目(Lkxyb22160)。

摘　　要：目的分析影响手术患者住院超30天的主要因素,促进医疗质量持续改进。方法调取某医院胃肠外科2019年1月1日-2021年12月31日手术患者病案,筛选三级和四级手术患者,年龄≥14岁,采用系统抽样加整群抽样选择一组病例,通过回顾性调查356名患者病案信息,应用SPSS 20.0软件进行分析,采用单因素分析、二元logistic回归和决策树回归分析影响住院超30天主要因素。结果二元Logistic回归结果显示术后并发症(OR=5.565)、危急值(OR=3.285)、术前住院天数≥7天(OR=1.264)、手术时间长≥180分钟(OR=2.678)、输血(OR=2.903)是患者住院超30天独立影响因素。决策树分析结果显示影响住院超30天的因素术前为住院天数≥7天、输血、术后并发症、危急值、低蛋白血症。2种模型AUC(ROC曲线下面积)为89%和80.4%(P<0.001),模型预测具有一定准确性。结论2种模型结果具有一定互补性,作为预测住院患者住院超30天具有一定准确性和实际意义。医院应重点关注恶性肿瘤与有腹部手术史患者,通过提升医疗技术水平,优化服务流程,强化围手术期管理,降低并发症的发生率,从而降低患者住院日及住院费用。Objectives This study aims to analyze the main factors affecting surgical patients'hospitalization for more than 30 days to promote continuous improvement of medical quality.Methods The medical records of surgical patients from the gastrointestinal surgery department of a certain hospital from January 1,2019 to December 31,2021 were retrieved.Level III and IV surgery patients aged≥14 years were screened.Systematic sampling plus cluster sampling were used to select a group of cases,and the medical record information of 356 patients was retrospectively investigated.SPSS20.0 software was used for analysis,and single factor analysis,binary logistic regression,and decision tree regression were used to analyze the main factors affecting hospitalization for more than 30 days.Results The results of binary Logistic regression showed that postoperative complications(OR=5.565),critical value(OR=3.285),preoperative hospital stay≥7 days(OR=1.264),Operation Time≥180 minutes(OR=2.678),and blood transfusion(OR=2.903)were the independent influencing factors.The results of decision tree analysis showed that the influencing factors of hospitalization over 30 days were hospitalization≥7 days,blood transfusion,postoperative complications,critical value,and hypoalbuminemia.The AUC(area under the ROC curve)of the two models was 89%and 80.4%,respectively(P<0.001),and the model predictions had a certain accuracy.Conclusions The results of the two models are complementary to each other and have certain accuracy and practical significance as the prediction of over 30 days of hospitalization.Hospitals should focus on patients with malignant tumors and patients with a history of abdominal surgery.By improving the level of medical technology,optimizing service processes,strengthening perioperative management,and reducing the incidence of complications,patients’hospitalization days and hospitalization expenses can be reduced.

关 键 词：LOGISTIC回归 决策树 住院超30天 影响因素 

分 类 号：R197.323[医药卫生—卫生事业管理]