Closing the implementation-evidence gap using data science: a transferable workflow applied to medical cost analysis.

Rising healthcare expenditure is a global challenge, particularly in ageing populations with increasing chronic disease burden. While many studies identify cost determinants, few provide a reproducible analytic workflow that is transferable across health systems. This study presents a data science-based analytic pipeline to model medical expenditure, demonstrated using a large, publicly available health expenditure dataset. The pipeline incorporates feature engineering, statistical and machine learning models, model comparison, and interpretability analysis. The primary contribution is methodological, emphasising transparency, reproducibility, and potential transferability rather than context-specific conclusions. Although implementation science frameworks inform the interpretation, they are used only as conceptual lenses, and no implementation strategies are evaluated. The findings are therefore intended as forward-looking considerations, illustrating how data-driven approaches may inform future policy design, stakeholder engagement, and technology-enabled strategies for preventive care and chronic disease management, subject to local adaptation.