Boosting precision health: An automated machine learning approach to actionable cardiorespiratory fitness risk prediction in female healthcare workers

Machine learning (ML) is increasingly used to predict health outcomes such as unsatisfactory cardiorespiratory fitness (CRF), a condition linked to obesity and cardiovascular disease. However, its performance relative to traditional methods remains mixed, and key predictors of poor CRF are not well defined. This study compared automated ML (autoML) models with multivariable logistic regression (MLR) in predicting unsatisfactory CRF among female health workers. Data from a hospital-based workplace health promotion program were randomly divided into training and test sets within a development cohort (<i>n</i> = 1,281). An independent validation cohort (<i>n</i> = 550) was used for external testing.