全科护理2025，Vol.23Issue(4)：610-615,6.DOI:10.12104/j.issn.1674-4748.2025.04.004
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
摘要
Abstract
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 APACHE Ⅱ 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 APACHE Ⅱ scores accounted for 83.3%of the total number in this node.Patients with low SOPA scores,low APACHE Ⅱ 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 this node.In ROC curve,the AUC of the modeling group was 0.980[95%CI(0.956,1.000)],the sensitivity was 91.30%,and the specificity was 97.40%.The AUC of the validation group was 0.980[95%CI(0.951,1.000)],the sensitivity was 90.00%,and the specificity was 92.00%.Conclusion:A decision tree prediction model was constructed based on the relevant influencing factors of elderly sepsis patients returning to the ICU within a short period.The model demonstrated good predictive performance.Clinical healthcare workers may use this model to identify high-risk elderly sepsis patients who are prone to returning to the ICU in the short term,thereby enabling timely intervention measures.关键词
老年脓毒症/短期/重返ICU/决策树/预测模型Key words
senile sepsis/short term/return to ICU/decision tree/prediction model引用本文复制引用
胡瑶,李梦云,刘洁琼,李宁..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[J].全科护理,2025,23(4):610-615,6.基金项目
江西省中医药管理局科技计划项目,编号:2020B0197. （）
