An interpretable machine learning model for predicting stroke in patients with hypertension: insights from the SPRINT

Stroke is a leading cause of mortality worldwide, with hypertension being its most significant risk factor. However, few studies have specifically developed predictive models for stroke risk in hypertensive patients with elevated cardiovascular risk. This study aims to develop and validate an interpretable predictive model based on machine learning (ML) methods to assess stroke risk in hypertensive patients, enabling early prevention and improved prognosis. This study is a secondary analysis of the Systolic Blood Pressure Intervention Trial (SPRINT). Key clinical variables were identified by the least absolute shrinkage and selection operator (LASSO) regression and predictive models were constructed using six machine learning (ML) algorithms. The predictive performance of the models was compared using multiple evaluation metrics, including the area under the receiver operating characteristic curve (AUC), and the results were interpreted using SHapley Additive Explanations (SHAP). A total of 264 participants were selected from the SPRINT dataset using random under-sampling. LASSO regression identified 14 key variables, which were used to develop six ML models. The logistic regression (LR) model was the most effective, with an AUC of 0.810. SHAP analysis indicated that gender, age, systolic blood pressure (SBP), serum creatinine, and race were the most influential factors for stroke risk, with female and higher age, SBP, and serum creatinine levels associated with an increased risk of stroke. An interpretable ML model was developed and validated to predict stroke risk in patients with hypertension and elevated cardiovascular risk, identifying gender, age, and SBP as key contributors. This tool may facilitate individualized risk assessment and support targeted prevention strategies in hypertensive patients at elevated cardiovascular risk.