Machine Learning-based Development of a Prediction Model for Activities of Daily Living Dysfunction in Stroke Survivors: A Cross-sectional Study (Preprint)

BACKGROUND Stroke is a leading cause of disability worldwide, imposing a significant burden on patients, families, and society. Despite advancements in acute phase stroke treatment, a large number of stroke survivors experience limitations in activities of daily living, which severely impacts their quality of life. Understanding the factors associated with ADL dysfunction in stroke survivors and predicting its occurrence at an early stage is crucial for developing targeted interventions. OBJECTIVE To create and verify a prediction model for activities of daily living (ADL) dysfunction in stroke survivors, pinpoint key predictors, and analyze the traits of those at risk. METHODS Data from the China Health and Retirement Longitudinal Study wave 5 was used in this cross-sectional study. 1,131 stroke survivors were included and split into training and testing sets. The least absolute shrinkage and selection operator regression and multivariate logistic regression were applied for model development. Model performance was evaluated using the area under the receiver operating characteristic curve(AUC), calibration plots, and decision curve analysis. SHapley Additive exPlanations values were calculated to understand predictor importance. RESULTS Six variables (age, the 10-item Center for Epidemiologic Studies Depression Scale score, memory disorder, self-rated health, pain count, and heavy physical activity) were identified as significant predictors. The model showed good discriminatory power (training set AUC = 0.804, testing set AUC = 0.779), accurate calibration, and clinical utility. CONCLUSIONS A prediction model for ADL dysfunction in stroke survivors was successfully developed and validated. It can help in formulating personalized medical plans, potentially enhancing stroke survivors' ADL ability and quality of life.