Machine learning-based prognostication of mortality in stroke patients.

Area of Science:

• Neurology
• Data Science
• Public Health

Background:

• Predicting stroke mortality is essential for tailoring patient care and improving survival rates.
• Machine learning (ML) offers a promising approach for developing accurate predictive models in healthcare.

Purpose of the Study:

• To design and evaluate a machine learning model for predicting one-year mortality after stroke.
• To identify key predictors of stroke mortality using ML and explainability techniques.

Main Methods:

• Utilized data from the National Multiethnic Stroke Registry, including 9840 patients.
• Trained and evaluated eight ML models, with XGBoost showing optimal performance.
• Employed SHapley Additive exPlanations (SHAP) to identify influential predictors.

Main Results:

• The XGBoost model achieved high accuracy (94.5%) and AUC (87.3%) in predicting one-year stroke mortality.
• Key predictors included National Institutes of Health Stroke Scale (NIHSS) score, age, hospital length of stay, mode of arrival, heart rate, and blood pressure.
• Higher NIHSS, age, and longer hospital stay were associated with increased mortality, while ambulance arrival and lower BMI predicted poorer outcomes.

Conclusions:

• The developed ML model effectively predicts stroke mortality, highlighting the importance of clinical and demographic factors.
• Findings support the use of ML for enhanced stroke risk assessment and personalized care strategies.
• Prospective validation is recommended to confirm clinical effectiveness and facilitate broader adoption.