首发脑卒中患者出院后1个月内康复锻炼低依从性风险预测Logistic回归模型与决策树模型构建及其对比研究  

Construction and Comparative Study of Logistic Regression Model and Decision Tree Model for Predicting Low Compliance of Rehabilitation Exercise of Patients with First-Episode Stroke within One Year after Discharge

作　　者：柏晓燕 陈丹[1] 李艳[1] 徐博[1] BAI Xiaoyan;CHEN Dan;LI Yan;XU Bo(Department of Neurosurgery,Nanjing Drum Tower Hospital,the Affiliated Hospital of Nanjing University Medical School,Nanjing 210008,China)

机构地区：[1]南京大学医学院附属鼓楼医院神经外科,江苏省南京市210008

出　　处：《实用心脑肺血管病杂志》2025年第3期57-62,共6页Practical Journal of Cardiac Cerebral Pneumal and Vascular Disease

基　　金：江苏省健康科研项目(F20210665);南京市卫生科技发展专项资金项目(YKK23092)。

摘　　要：目的构建、对比首发脑卒中患者出院后1个月内康复锻炼低依从性风险预测Logistic回归模型及决策树模型。方法回顾性选取2022年4月—2024年3月南京大学医学院附属鼓楼医院收治的223例首发脑卒中患者作为研究对象,根据康复锻炼依从性分为中高依从性组182例和低依从性组41例。收集患者的临床资料,采用多因素Logistic回归分析探讨首发脑卒中患者出院后1个月内康复锻炼低依从性的影响因素,基于多因素Logistic回归分析结果构建Logistic回归模型及基于单因素分析结果构建决策树模型。采用Omnibus检验及5折交叉验证法验证Logistic回归模型的稳定性。绘制ROC曲线以分析Logistic回归模型及决策树模型对首发脑卒中患者出院后1个月内康复锻炼低依从性的预测价值,采用Hosmer-Lemeshow拟合优度检验评价Logistic回归模型及决策树模型的拟合程度。结果单因素分析结果显示,两组年龄、文化程度、脑卒中类型、美国国立卫生研究院卒中量表(NIHSS)评分、偏瘫发生率、焦虑/抑郁发生率比较,差异有统计学意义(P<0.05)。多因素Logistic回归分析结果显示,年龄、文化程度、脑卒中类型、NIHSS评分、偏瘫、焦虑/抑郁是首发脑卒中患者出院后1个月内康复锻炼低依从性的独立影响因素(P<0.05)。基于上述结果分别构建Logistic回归模型及决策树模型。Logistic回归模型与决策树模型预测首发脑卒中患者出院后1个月内康复锻炼低依从性的AUC分别为0.839[95%CI(0.785~0.885)]、0.827[95%CI(0.771~0.875)],两种预测模型相应AUC比较,差异无统计学意义(P>0.05)。Hosmer-Lemeshow拟合优度检验结果显示,Logistic回归模型与决策树模型的拟合程度均较好(χ^(2)值分别为9.498、9.142,P值分别为0.417、0.356)。结论本研究基于年龄、文化程度、脑卒中类型、NIHSS评分、偏瘫、焦虑/抑郁构建的首发脑卒中患者出院后1个月内康复锻炼低Objective To constuct and compare the Logistic regression model and decision tree model for predicting low compliance of rehabilitation exercise of patients with first-episode stroke within one year after discharge.Methods A total of 223 patients with first-episode stroke admitted to Nanjing Drum Tower Hospital,the Affiliated Hospital of Nanjing University Medical School from April 2022 to March 2024 were retrospectively selected as the study subjects.Patients were divided into the medium-high compliance group and the low compliance group according to the rehabilitation exercise compliance.Clinical data of the patients were collected.Multivariate Logistic regression analysis was used to explore the influencing factors of low compliance of rehabilitation exercise in patients with first-episode stroke within one year after discharge.The Logistic regression model was constructed based on the results of multivariate Logistic regression analysis,and decision tree model was constructed based on the results of univariate analysis.Omnibus test and five fold cross validation method were used to verify the stability of the Logistic regression model.ROC curve was drawn to analyze the predictive value of the Logistic regression model and decision tree model for low compliance of rehabilitation exercise of patients with first-episode stroke within one year after discharge.Hosmer-Lemeshow goodness of fit test was used to evaluate the fitting degree of the Logistic regression model and decision tree model.Results Univariate analysis results showed that,there were significant differences in age,educational level,stroke type,National Institute of Health Stroke Scale(NIHSS)score,incidence of hemiplegia and incidence of anxiety/depression between the two groups(P<0.05).Multivariate Logistic regression analysis results showed that,age,educational level,stroke type,NIHSS score,hemiplegia and anxiety/depression were independent influencing factors of low compliance of rehabilitation exercise in patients with first-episode stroke withi

关 键 词：卒中 康复锻炼依从性 LOGISTIC回归模型 决策树模型 

分 类 号：R743[医药卫生—神经病学与精神病学]