Advancing health equity in proactive health management: from data underrepresentation and algorithmic bias to a closed-loop governance framework

This narrative review in Frontiers in Public Health synthesizes 108 peer-reviewed studies (Jan 2015–Feb 2026) to argue that structural data gaps in patient-generated health data, driven by social determinants, systematically underrepresent high-risk groups in AI training data. That underrepresentation cascades through model development and deployment, producing biased predictions and misallocated resources. The authors propose a closed-loop governance framework linking data, algorithm, and equity oversight to advance health equity in proactive health management.