XGBoost and SHAP-Based Analysis of Risk Factors for Hypertension Classification in Korean Postmenopausal Women.

Abstract

In postmenopausal women, the prevalence of hypertension increases sharply, emphasizing the importance of its prevention. This increased risk highlights the critical need for effective prevention strategies specifically designed for this population. To address this issue, the present study aimed to identify easily measurable risk factors that contribute to hypertension in postmenopausal women using explainable artificial intelligence (XAI) and machine learning (ML) techniques. This study conducted hypertension classification by analyzing health checkup data from 3289 postmenopausal Korean women aged 55-79 years, extracted from the 2022-2023 Korea National Health Insurance Service (KNHIS) database, using XGBoost, SVM and ANN. XGBoost was the most effective model (AUC: 92.12%, MCC: 0.71) in hypertension classification. Shapley Additive exPlanations-based feature importance identified age and waist circumference (WC) as the most important risk factors for hypertension. In this study, blood pressure increased with variations in WC, a modifiable risk factor. These findings suggest that WC should be managed more strictly to prevent hypertension in postmenopausal women.