构建多模态机器学习驱动的老年脑卒中患者认知衰弱风险预测模型OA

Objective:To identify key influencing factors and construct a machine learning predictive model for cognitive frailty in elderly stroke patients,providing a basis for early clinical identification and intervention.Methods:This cross-sectional study(from Sep 2023 to Apr 2024)enrolled 420 elderly stroke patients from a tertiary hospital in Liaoning Province using convenience sampling.Cognitive frailty was assessed using the FRAIL Scale and Montreal Cognitive Assessment(MoCA),and participants were grouped accordingly.Data were collected via general information questionnaires,Barthel Index(BI),Geriatric Depression Scale(GDS-15),Mini Nutritional Assessment Short Form(MNA-SF),and Perceived Social Support Scale(PSSS).The subjects were divided into a training set and a validation set in a 7∶3 ratio.After variable selection via LASSO regression,five machine learning algorithms,including Logistic Regression and Random Forest,were employed to construct predictive models.Model performance was evaluated using receiver operating characteristic(ROC)curve and confusion matrices,with the optimal model interpreted using SHAP.Results:The prevalence of cognitive frailty among elderly stroke patients was 17.6%.The Support Vector Machine(SVM)model demonstrated the best performance,with an area under the ROC curve(AUC)of 0.978(95%CI:0.931-1.000)in the training cohort and 0.901(95%CI:0.807-0.995)in the validation cohort.SHAP analysis identified coronary heart disease,weekly exercise frequency,and age as the primary influencing factors.Conclusion:Cognitive frailty in elderly stroke patients is influenced by multiple factors,and the SVM prediction model can assist to identify high-risk individuals early and implement interventions.