Bi-dimensional health space mapping: machine learning analysis of population health dynamics in Korean and Dutch cohorts.

Abstract

Health spans a broad spectrum, encompassing various biological and lifestyle factors. The complexity of biological systems necessitates for integrating diverse factors into a unified biomarker. We constructed a health space model that highlights metabolism and oxidative stress as key indicators for tracking healthy aging and mapping health trajectories. To ensure cross-ethnic relevance, we used data from the Dutch Nutrition Questionnaires plus and Korean National Health and Nutrition Examination Survey (KNHANES) cohorts. Our approach combines machine learning with logistic regression, applying a least absolute shrinkage and selection operator penalty to propensity score-matched datasets. External validation using an independent KNHANES cohort showed strong performance (AUC = 0.959 for metabolic stress; 0.973 for oxidative stress), confirming model reliability. These findings support the health space model as a holistic tool for monitoring physiological stress. Our research advances personalized health monitoring and offers a foundation for precision nutrition strategies aimed at reducing chronic disease risk.