2026 Volume 53 Issue 4 Pages 238-246
Objective: This study aimed to predict the discharge functional independence measure (FIM) score in patients with stroke in a convalescent rehabilitation ward using random forest based on baseline information and FIM score at admission and validate its accuracy.
Methods: This study included 304 patients with stroke. The explanatory variables were sex, age, body mass index, number of days from onset to admission, and FIM score at admission. The objective variable was the discharge FIM score. Prediction accuracy was evaluated in 213 patients for training and 91 patients for testing.
Results: The prediction accuracy was 0.86 and 0.85 for the training and test data, respectively.
Conclusion: Discharge FIM score can be predicted with high accuracy using baseline information and FIM score at admission, while preventing overfitting.
