Frail status of elderly patients with mild ischemic stroke and construction of a prediction risk model based on random forest algorithm

Ma Yuting; Wang Xiaoyan; Dou Li

doi:10.3969/j.issn.1009-0126.2025.02.014

Abstract：Objective To investigate the influencing factors for frailty in elderly patients with mild ischemic stroke(MIS),construct a risk prediction model based on random forest algorithm and evaluate its application performance.Methods A total of 322 elderly MIS patients subjected by cluster sampling in the Neurodiagnosis and Treatment Center of our hospital during January and June 2022 were enrolled and divided into non-frailty group(n=261)and frailty group(n=61)according to the results of frailty screening scale.The general clinical data,and the scores of Daily Living Ability Scale,Simplified Geriatric Depression Scale,Social Support Rating Scale,36-Item Concise Health Questionnaire,and Simplified Intelligent State Examination Scale were compared between two groups.Random forest algorithm was used to construct a risk prediction model for frailty in the elderly MIS patients,and ROC curve was plotted to analyze the predictive value of the model.Results The incidence of frailty in elderly MIS patients was 18.94％.There were sta-tistical differences in age,marital status,monthly income,medical expense payment,sleep condi-tion and comorbidities in the frail group(P＜0.01).The frail group had significantly lower scores of support utilization,concise health status,physiological function,body pain,general health,social function,and emotional function,and higher scores of Activities of Daily Living Scale and the Simplified Geriatric Depression Scale than the non-frail group(P＜0.05).The random forest model obtained an accuracy of 88.28％,a sensitivity of 88.24％,a specificity of 88.89％,an F1 score of 0.933,and an AUC value of 0.816.Random forest algorithm ranked the important varia-bles influencing frailty in elderly MIS patients through variable importance scores,and the top five predictors were types of basic diseases(≥2 types),sleep condition,quality of life score,self-care ability score,and personal monthly income(＜2000 Yuan).Conclusion A risk prediction model for frailty in elderly MIS patients is constructed by using random forest algorithm.The model is helpful to identify key factors of frailty as early as possible,provide empirical evidence for clinical medical staff to implement early intervention,improve the health and quality of elderly MIS patients,and prevent or delay the occurrence of frailty.