基于决策树算法的老年脓毒症ICU转出病人短期内重返ICU预测模型构建  

Construction of short-term return to ICU prediction model for elderly patients with sepsis who were transferred out of ICU based on decision tree algorithm

作　　者：胡瑶 李梦云 刘洁琼 李宁 HU Yao;LI Mengyun;LIU Jieqiong;LI Ning(The First Hospital of Nanchang,Jiangxi 330008 China)

机构地区：[1]南昌市第一医院,330008

出　　处：《全科护理》2025年第4期610-615,共6页Chinese General Practice Nursing

基　　金：江西省中医药管理局科技计划项目,编号:2020B0197。

摘　　要：目的:调查老年脓毒症重症监护室(ICU)转出病人短期内重返ICU的危险因素,构建决策树预测模型并进行验证。方法:选取2022年4月—2023年8月由医院ICU转出的140例老年脓毒症病人作为建模组,按照病人短期内是否重返ICU分为重返组23例和非重返组117例,采用单因素分析和二元Logistic回归分析筛选老年脓毒症病人短期内重返ICU的影响因素,并构建决策树预测模型。另按7∶3分配比选取2023年9月—2024年4月由ICU转出的60例老年脓毒症病人作为验证组,采用受试者工作特征(ROC)曲线评估模型预测效能。结果:建模组140例老年脓毒症病人短期内重返ICU的发生率为16.42%(23/140);多因素分析结果显示,年龄≥75岁、合并糖尿病、ICU住院时间≥5 d、机械通气、高水平SOPA评分、高水平APACHEⅡ评分是老年脓毒症病人短期内重返ICU的独立影响因素(均P<0.05);根据重返ICU病人所占节点的构成比代表老年脓毒症病人重返ICU的风险,筛选出4类高危人群,即高水平SOPA评分病人占该节点构成的83.3%;低水平的SOPA评分、高水平的APACHⅡ评分病人占该节点构成的83.3%;低水平的SOPA评分、低水平的APACHⅡ评分、ICU住院时间≥5 d、年龄≥75岁病人占该节点构成的38.5%;低水平的SOPA评分、低水平的APACHⅡ评分、ICU住院时间<5 d病人占该节点构成的13.3%。ROC曲线中,建模组的曲线下面积(AUC)为0.980[95%CI(0.956,1.000)],灵敏度为91.30%,特异度为97.40%;验证组的AUC为0.980[95%CI(0.951,1.000)],灵敏度为90.00%,特异度为92.00%。结论:根据老年脓毒症病人短期内重返ICU的相关影响因素构建决策树预测模型,预测效果良好,临床医护工作者或可根据此模型筛选出短期内易重返ICU的高风险老年脓毒症病人,从而及时开展干预措施。Objective:To investigate the risk factors of short-term return to ICU in elderly patients with sepsis who had been transferred out of ICU,and construct a decision tree prediction model and verify it.Methods:A total of 140 elderly patients with sepsis who were transferred out of ICU from Hospital from April 2022 to August 2023 were selected as the modeling group.According to whether the patients returned to ICU in the short term,they were divided into the re-entry group(23 cases) and the non-re-entry group(117 cases).Univariate analysis and binary Logistic regression analysis were used to screen the influencing factors of elderly sepsis patients returning to ICU in the short term,and a decision tree prediction model was constructed.In addition,60 elderly patients with sepsis who were transferred out of ICU from September 2023 to April 2024 were selected as the validation group according to 7:3 allocation ratio,and the predictive efficacy of the model was evaluated by ROC curve.Results:In the modeling group,the incidence of 140 elderly patients with sepsis returning to ICU within a short period was 16.42%(23/140).The results of multivariate analysis showed that age≥75 years old,diabetes mellitus,ICU stay≥5 days,mechanical ventilation,high SOPA score and high APACHEⅡ score were independent influencing factors for elderly sepsis patients returning to ICU within a short period(all P<0.05).Based on the proportion of patients returning to the ICU within each node,representing the risk of elderly sepsis patients returning to the ICU,four high-risk groups were identified.Patients with high SOPA scores accounted for 83.3% of the total number in this node.Patients with low SOPA scores and high APACHEⅡ scores accounted for 83.3% of the total number in this node.Patients with low SOPA scores,low APACHEⅡ scores,ICU stay≥5 days,and age≥75 years accounted for 38.5% of the total number in this node.Patients with low SOPA scores,low APACHE Ⅱ scores,and ICU stay <5 days accounted for 13.3% of the total number in thi

关 键 词：老年脓毒症 短期 重返ICU 决策树 预测模型 

分 类 号：R47[医药卫生—护理学]