Machine learning-based models to predict 90-day non-excellent functional outcomes among patients with mild acute ischemic stroke after intravenous thrombolysis.

About one quarter of patients with mild acute ischemic stroke treated with intravenous thrombolysis had a non-excellent functional outcome at 90 days.

• 507 patients with mild acute ischemic stroke who received intravenous thrombolysis were included.
• 136 patients (26.8%) experienced a non-excellent functional outcome at 90 days.
• Data were retrospectively collected at Tianjin Fourth Central Hospital between January 2024 and June 2025.

LASSO regression selected six predictors of 90-day non-excellent functional outcome.

• The selected predictors were age, baseline NIHSS score, platelet count, monocyte count, blood glucose level, and neutrophil-to-lymphocyte ratio.
• Candidate variables included demographic, clinical, laboratory, and neuroimaging data extracted from electronic medical records.

XGBoost achieved the highest numerical discrimination among seven machine learning models in the validation cohort.

• Validation AUC was 0.753 (95% CI, 0.660-0.840).
• Seven machine learning models were developed and compared.
• The model was described as showing adequate calibration and clinical usefulness.

Model performance was evaluated across multiple dimensions, and SHAP analysis was used to interpret predictor contributions.

• Assessment covered discrimination, calibration, clinical utility, and classification metrics.
• SHAP analysis was used to show the relative importance and direction of association of each predictor with model predictions.

The study concludes that the XGBoost model is promising for early risk stratification, though it was validated only internally.

• The model was developed and internally validated in a single-center retrospective cohort.
• The authors describe the AUC as the highest 'numerical' value, suggesting differences among models may not be definitive.