Advancing healthcare AI governance through a comprehensive maturity model based on systematic review
npj Digital Medicine, 2026
This article explores a key challenge in healthcare AI: not just how to develop or adopt AI tools, but how to govern them responsibly in real clinical settings. Based on a systematic review of frameworks published between 2019 and 2024, the authors find that current guidance is often fragmented, uneven, and better suited to large academic centers than to smaller or lower-resource healthcare organizations.
To address this gap, the paper introduces HAIRA, the Healthcare AI governance Readiness Assessment, a five-level maturity model designed to help healthcare systems evaluate and strengthen their AI governance capabilities. HAIRA covers seven key domains: organizational structure, problem formulation, external algorithm evaluation, algorithm development, model evaluation, deployment, and monitoring. Its main value lies in offering a practical, staged pathway that organizations can adapt to their own resources and expertise.
The paper also emphasizes that strong AI governance must go beyond model accuracy alone. It should include fairness, patient safety, workflow integration, transparency, user training, and continuous monitoring after deployment. The authors highlight recurring recommendations such as multidisciplinary governance committees, internal and external validation, shadow deployment before full rollout, and systems to detect bias or performance drift over time.
Overall, the article makes a useful contribution by translating broad responsible AI principles into a more operational framework. It offers healthcare organizations a realistic way to assess readiness, identify governance gaps, and scale AI adoption more safely and effectively.