JMIR Publications

ABSTRACT

Background:

Mortality prediction in intensive care unit (ICU) patients with ischemic stroke and intracranial artery stenosis or occlusion remains challenging. Conventional scoring systems often lack accuracy and fail to provide individualized risk estimates.

Objective:

To develop and evaluate machine learning models for individualized mortality prediction in ICU patients with ischemic stroke and intracranial artery disease.

Methods:

We performed a retrospective cohort study including 5,280 ICU patients with ischemic stroke from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. After rigorous feature selection, we trained and tested multiple machine learning algorithms, including Light Gradient Boosting Machine (LightGBM) and Bagging classifiers. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), accuracy, recall, precision, and F1 score. SHapley Additive exPlanations (SHAP) were applied to interpret model predictions.

Results:

LightGBM and Bagging achieved the best discrimination, both with an AUC of 0.82 and accuracy exceeding 73%. The most influential predictors included Acute Physiology Score III, suspected infection, Charlson comorbidity index, and age. SHAP analysis demonstrated how these features contributed to individualized mortality risk estimates.

Conclusions:

Machine learning models can provide accurate and interpretable mortality predictions for ICU patients with ischemic stroke and intracranial artery stenosis or occlusion. These tools have the potential to enhance clinical decision-making through individualized risk assessment. External validation is warranted before clinical implementation. Clinical Trial: Not applicable