Machine Learning–Based Prediction for Incident Hypertension Based on Regular Health Checkup Data: Derivation and Validation in 2 Independent Nationwide Cohorts in South Korea and Japan (Preprint)

• DOI: 10.2196/preprints.52794
• Publication Date: 2024-11-05T15:47:36Z
• Authors: Seung Ha Hwang; Hayeon Lee; Jun Hyuk Lee; Myeongcheol Lee; Ai Koyanagi; Lee Smith; Sang Youl Rhee; Dong Keon Yon; Jinseok Lee
• Keywords: Preprint

<sec> <title>BACKGROUND</title> Worldwide, cardiovascular diseases are the primary cause of death, with hypertension as a key contributor. In 2019, led to 17.9 million deaths, predicted reach 23 by 2030. </sec> <title>OBJECTIVE</title> This study presents new method predict using demographic data, 6 machine learning models for enhanced reliability and applicability. The goal is harness artificial intelligence early accurate diagnosis across diverse populations. <title>METHODS</title> Data from 2 national cohort studies, National Health Insurance Service-National Sample Cohort (South Korea, n=244,814), conducted between 2002 2013 were used train test designed anticipate incident within 5 years health checkup involving those aged ≥20 years, Japanese Medical Center (Japan, n=1,296,649) extra validation. An ensemble was identify most salient features contributing presenting feature importance analysis confirm contribution each future. <title>RESULTS</title> Adaptive Boosting logistic regression showed superior balanced accuracy (0.812, sensitivity 0.806, specificity 0.818, area under receiver operating characteristic curve 0.901). indicators age, diastolic blood pressure, BMI, systolic fasting glucose. dataset (extra validation set) corroborated these findings (balanced 0.741 0.824). model integrated into public web portal predicting onset based on data. <title>CONCLUSIONS</title> Comparative evaluation our against classical statistical distinct studies emphasized former’s stability, generalizability, reproducibility in onset. <title>CLINICALTRIAL</title> <p/>