Using Machine Learning to Predict Hypertension from Lifestyle Factors: A Study Using NHANES Data

Hypertension, a leading global health concern, often goes undetected until it manifests itself as severe cardiovascular complications. Although prior research has demonstrated the utility of machine learning in hypertension prediction, many models rely on clinical or laboratory-only variables that limit generalizability and accessibility in the real world. This study addresses this gap by exploring the application of machine learning models to predict hypertension using publicly available survey data from the National Health and Nutrition Examination Survey (NHANES). We develop and evaluate predictive models employing random forest, gradient boosting, bagging decision trees, and Logistic Regression using lifestyle, demographic, and medical history variables. A complete feature set with 56 features and a reduced set with 20 features are analyzed to assess the impact of feature selection on model performance. The results indicate that gradient boosting with reduced features achieve the highest accuracy (73.78%), demonstrating the potential of optimized feature selection to enhance prediction efficiency. These findings highlight the utility of machine learning in enabling cost-effective and early risk assessment tools for hypertension, paving the way for more personalized and preventive healthcare strategies.