Development and Comparison of Machine Learning Models Incorporating the Weight-Adjusted Waist Index for Predicting Ischemic Stroke in Older Adults With Hypertension

Abstract Objective Ischemic stroke is a major cause of disability and mortality among elderly patients with hypertension. Identifying reliable and easily obtainable predictive indicators is essential for early risk stratification. The Weight-Adjusted Waist Index (WWI), a novel marker of central obesity, has not been fully evaluated in stroke prediction models. This study aimed to develop and compare multiple machine learning models for predicting ischemic stroke in elderly hypertensive patients and to assess the predictive value of WWI. Methods A total of 1,070 elderly patients with hypertension were retrospectively enrolled. Feature selection was performed using Least Absolute Shrinkage and Selection Operator (LASSO) regression. Eight machine learning algorithms were developed, including decision tree (DT), k-nearest neighbors (KNN), random forest (RF), Light Gradient Boosting Machine (LightGBM), logistic regression (LR), multilayer perceptron (MLP), support vector machine (SVM), and extreme gradient boosting (XGBoost). Model performance was evaluated using accuracy, precision, recall, F1-score, and the area under the receiver operating characteristic curve (AUC-ROC). SHapley Additive exPlanations (SHAP) were applied to interpret feature importance, and Decision Curve Analysis (DCA) was used to assess clinical utility. Results The LightGBM model demonstrated the best overall performance (accuracy: 0.857, recall: 0.905, F1-score: 0.915), with superior discrimination and calibration compared with the other models. DCA indicated that LightGBM and LR yielded the highest net benefits across a wide range of threshold probabilities. SHAP analysis identified a history of cardiac disease, age, and WWI as the most important predictors. Notably, WWI contributed more than BMI, suggesting a stronger association with abnormal fat distribution and cardiovascular risk. Conclusion WWI is a more effective anthropometric indicator than BMI for predicting ischemic stroke risk in elderly patients with hypertension. WWI showed substantial predictive contribution and may provide additional information beyond BMI for ischemic stroke risk assessment in older patients with hypertension.The LightGBM-based model demonstrates strong discriminative ability and clinical applicability, offering a promising tool for early screening and individualized intervention.