INTRODUCTION: Intravenous thrombolysis (IVT) with tissue-type plasminogen activator (tPA) is a cornerstone of acute ischemic stroke treatment, yet its benefits are limited by the risk of symptomatic intracranial hemorrhage (sICH), a complication associated with substantial morbidity and mortality. Functional recovery is commonly evaluated using the 3-month modified Rankin Scale (mRS). This study aimed to develop predictive models for sICH and 3-month outcomes after tPA and to identify key prognostic variables that may support individualized treatment decisions. METHODS: Data from 434 ischemic stroke patients who received tPA at a tertiary medical center over 5.5 years were analyzed. Three supervised classification models were developed and evaluated using five-fold cross-validation. Model performance was assessed using area under the receiver operating characteristic curve (AUC), accuracy, recall, and precision, and was compared with six established scoring tools for predicting post-tPA sICH. RESULTS: The three machine learning models-Logistic Regression (AUC 0.87), Random Forest (AUC 0.82), and XGBoost (AUC 0.89)-outperformed all six conventional scoring tools in predicting post-tPA sICH. The 24-hour NIHSS score was the most influential predictor for both sICH and 3-month outcomes. A history of previous stroke and male sex were associated with higher sICH risk, while increasing age was strongly correlated with poorer 3-month outcomes. CONCLUSION: The proposed models demonstrated high predictive accuracy for both sICH and 3-month outcomes after tPA and highlighted variables with the greatest prognostic contribution. The 24-hour NIHSS score emerged as the strongest predictor in both tasks. Compared with established sICH scoring tools, these machine learning models provided superior performance, suggesting their potential value as decision-support tools to guide individualized management following thrombolytic therapy.
Lin et al. (Tue,) studied this question.