决策树法和Logistic回归模型在ICU老年患者压疮影响因素中的应用

Application of decision tree method and Logistic regression model in influencing factors of pressure sore in elderly patients in ICU

目的:分析ICU老年患者压疮发生的危险因素,应用Logistic回归分析和决策树算法构建压疮发生风险预测模型。方法:选取2017-01-01~2019-12-31在河北省某三级甲等医院ICU住院的848例老年患者,采用回顾性调查方法收集患者资料,将筛选出的压疮发生的危险因素构建Logistic回归模型和决策树模型,分析比较两种模型的预测结果。结果:决策树法和Logistic回归分析法均显示自主运动障碍、水肿、吸烟和使用血管加压剂为ICU老年住院患者发生压疮的危险因素;Logistic回归模型AUC为0.815,大于决策树的0.753,差异有统计学意义(Z=2.473,P<0.05);Logistic回归模型的约登指数为0.526,灵敏度为75.3%,均高于决策树模型的0.395和61.6%。结论:Logistic回归模型对ICU老年患者压疮风险预测效果良好,可以为临床压疮预防提供参考依据。

Objective:To analyze the risk factors of pressure sore in elderly patients in ICU,and to construct a pressure ulcer risk prediction model using Logistic regression analysis and decision tree algorithm.Methods:A total of 848 elderly patients hospitalized in the ICU of a tertiary A hospital in Hebei Province from January 2017 to December 2019 were included in this study.Retrospective investigation method was used to collect patient′s data,and the risk of pressure sore selected by multi-factor analysis factors,Logistic regression model and decision tree model were built,theprediction results of the two models were compared.Results:Both decision tree model and Logistic regression analysis showed that voluntary dyskinesia,edema,smoking and the use of vasopressors were risk factors for pressure sore in elderly hospitalized ICU patients,Logistic regression the model AUC was 0.815,which was greater than 0.753 of the decision tree,and the difference was statistically significant(Z=2.473,P<0.05),the Youden index of the Logistic regression model was 0.526,and the sensitivity was 75.3%.Both were higher than the 0.395 and 61.6%of the decision tree model.Conclusion:Logistic regression model has a good effect on predicting the risk of pressure sore in elderly patients in ICU,and can provide a reference for clinical pressure ulcer prevention.