【摘要】 目的:了解缺血性脑卒中病人卒中后抑郁（PSD）发生现状，采用Logistic回归模型和决策树C5.0模型探讨其影响因素。方法:采用便利抽样法选取2022年1月—10月滨州市某三级甲等医院确诊为缺血性脑卒中的215例病人为研究对象进行调查，通过构建Logistic回归模型和决策树C5.0模型分析PSD发生的影响因素，比较两种模型的预测性能。结果:215例缺血性脑卒中病人完成调查，其PSD发生率为37.7%。Logistic回归模型显示，性别、脑卒中部位、日常生活活动能力量表（BI）评分、同型半胱氨酸和红细胞分布宽度是PSD发生的影响因素，模型受试者工作特征曲线下面积（AUC）为0.772。决策树C5.0模型显示，BI评分、脑卒中部位和是否合并糖尿病是PSD发生的影响因素，模型AUC为0.729。两种模型预测性能比较差异无统计学意义（Z=1.737,P=0.082）。结论:缺血性脑卒中病人PSD发生率较高，Logistic回归模型和决策树C5.0模型对其具有一定预测价值，建议两种模型联合使用以提高预测性能，实现PSD的早期识别和干预。更多还原

【Abstract】 Objective:To investigate the status quo of post-stroke depression(PSD) in patients with ischemic stroke, and to explore its influencing factors by Logistic regression model and decision tree model. Methods:A total of 215 patients who were diagnosed with ischemic stroke in a tertiary grade A hospital in Binzhou city were selected as subjects by convenience sampling method from January to October 2022.Logistic regression model and decision tree C5. 0 model were constructed for analyzing the influencing factors of PSD. The predictive performance of two models was compared. Results:A total of 215 patients with ischemic stroke in the study were investigated. The PSD incidence rate was 37. 7%. Logistic regression model showed that sex,lesion location,Barthel Index(BI) score,homocysteine and red blood cell distribution width(CV) were risk factors for PSD. The area under the curve(AUC) of receiver operator characteristic of the model was 0. 772. Decision tree C5. 0 model showed that BI score,lesion location and diabetes were influencing factors of PSD. The AUC of the model was 0. 729. There was no significant difference in the prediction efficiency between the two models(Z=1. 737,P=0. 082). Conclusions:The incidence of PSD in patients with ischemic stroke is relatively high,and Logistic regression model and decision tree C5. 0 model have certain predictive value for it. The two models should be combined to improve prediction performance and achieve early identification and intervention of PSD.更多还原