Prediction of poststroke independent walking using machine learning: a retrospective study

Area of Science:

• Rehabilitation Medicine
• Artificial Intelligence in Healthcare
• Clinical Prediction Modeling

Background:

• Predicting walking independence post-stroke is crucial for patient outcomes and resource allocation.
• Current prediction methods require enhancement for improved accuracy and clinical utility.

Purpose of the Study:

• To compare the predictive performance of logistic regression (LR) and machine learning (ML) models, including eXtreme Gradient Boosting (XGBoost), Support Vector Machines (SVM), and Random Forest (RF).
• To identify key prognostic factors for independent walking in stroke survivors.

Main Methods:

• Retrospective analysis of 778 stroke patients admitted between February 2020 and January 2023.
• Model training and validation using 80% and 20% of the data, respectively.
• Performance evaluation based on Area Under the Curve (AUC), accuracy, sensitivity, specificity, Positive Predictive Value (PPV), Negative Predictive Value (NPV), and F1 score.

Main Results:

• XGBoost demonstrated superior predictive performance with a significantly higher AUC compared to SVM and RF.
• XGBoost achieved higher accuracy, sensitivity, PPV, and F1 score than LR, with comparable AUC.
• Key predictors identified include age, Fugl-Meyer Assessment of the Lower Extremity (FMA-LE) at admission, Functional Ambulation Category (FAC) at admission, and lower limb spasticity.

Conclusions:

• The XGBoost model is the most effective tool for predicting independent walking in stroke patients at hospital discharge.
• Age, FMA-LE, FAC, and lower extremity spasticity are confirmed as critical factors influencing walking prognosis after stroke.