Application of machine learning algorithms in predicting new onset hypertension: a study based on the China Health and Nutrition Survey

• DOI: 10.1265/ehpm.24-00270
• Publication Date: 2025-01-09T22:15:05Z
• Authors: Manhui Zhang; Xian Xia; Qiqi Wang; Yue Pan; Guanyi Zhang; Zhigang Wang
• Keywords: new onset hypertension; chns; machine learning algorithms; prediction; Public aspects of medicine; RA1-1270; Research Article

Hypertension is a serious chronic disease that can significantly lead to various cardiovascular diseases, affecting vital organs such as the heart, brain, and kidneys. Our goal predict risk of new onset hypertension using machine learning algorithms identify characteristics patients with hypertension. We analyzed data from 2011 China Health Nutrition Survey cohort individuals who were not hypertensive at baseline had follow-up results available for prediction by 2015. tested evaluated performance four traditional commonly used in epidemiological studies: Logistic Regression, Support Vector Machine, XGBoost, LightGBM, two deep algorithms: TabNet AMFormer model. modeled 16 29 features, respectively. SHAP values applied select key features associated A total 4,982 participants included analysis, whom 1,017 developed during 4-year follow-up. Among 16-feature models, Regression highest AUC 0.784(0.775∼0.806). In 29-feature performed best an 0.802(0.795∼0.820), also scored MCC (0.417, 95%CI: 0.400∼0.434) F1 (0.503, 0.484∼0.505) metrics, demonstrating superior overall compared other models. Additionally, selected based on AMFormer, age, province, waist circumference, urban or rural location, education level, employment status, weight, WHR, BMI, played significant roles. model first time predicting achieved among six tested. Key be determined through this algorithm. The practice further enhance predictive efficacy diseases factors diseases.