Development and Internal Validation of a Model Predicting the Risk of Recurrent Stroke for Middle-Aged and Elderly Patients: A Retrospective Cohort Study

Abstract

Objective: To develop and validate a model for predicting the risk of recurrent stroke among middle-aged and elderly stroke patients.

Methods: A total of 1,327 stroke patients from the China Health and Retirement Longitudinal Study (CHARLS) were included in the retrospective cohort study, and they were randomly divided into the training and test sets at a ratio of 7:3. Univariate and multivariate regression analyses were used to select the predictors in the training set, which were used to develop logistic regression model. The Delong test and area under the receiver operating characteristic curve were adopted to investigate the predicted performance of the model.

Results: The average follow-up time was 2.26 ± 0.52 years, and the incidence of recurrent stroke was 14.47%. The result indicated that duration of moderate exercise, duration of walking, social activities, and diastolic blood pressure were associated with the risk of recurrent stroke among the middle-aged and elderly stroke patients. A logistic regression model was constructed to predict the risk of recurrent stroke after 2 years: [Logit (P_R)=ln (P_R/(1-P_R) =-1.658-0.841 moderate exercise (<2 hours/day)-0.559∗moderate exercise (≥2 hours/day)-0.906∗walk (<2 hours/day)-1.131∗walk (≥2 hours/day)-0.474∗social activities 1-0.968∗social activities 2-1.248∗social activities 3 + 0.015∗diastolic blood pressure)]. The value of the area under the curve reached 0.75, showing that the logistic regression model performs well in the prediction of the risk of recurrent stroke.

Conclusions: A logistic regression model for predicting the risk of recurrent stroke was developed among middle-aged and elderly stroke patients after 2 years, and the model showed good discrimination and accuracy via internal validation.

Keywords: CHARLS; Elderly; Middle-aged; Prediction model; Recurrent stroke.