JMIR Publications

ABSTRACT

Background:

Robust and adaptive artificial intelligence (AI) governance frameworks are essential to ensure the responsible, sustainable and effective integration of AI-based technologies in health systems (HS). Over the past decade, numerous frameworks for governing AI in HS have been developed to foster accountability, support implementation and mitigate risks such as bias, compromised quality of care, data breaches and adverse financial impacts. However, current frameworks often fail to reflect the multidimensional and dynamic nature of AI governance in HS.

Objective:

This study had two objectives: (1) To review and synthesize existing AI governance frameworks for HS; and (2) To propose an Integrative AI Governance Model that identifies key components to guide AI-related policy, practice and research in HS.

Methods:

We conducted a systematic review of AI governance frameworks for HS published from November 2014 to July 2025. Sources included eight academic databases (PubMed, MEDLINE, Embase, ACM Digital Library, Web of Science, Scopus, Social Sciences Abstracts and PsycINFO), grey-literature databases, and the web portals of international organizations. Inclusion criteria covered peer-reviewed articles and reports proposing a framework, guideline, standard or position statement on the governance of AI in HS. We excluded abstracts, letters to the editor, commentaries, essays, viewpoints, conference proceedings, and articles in languages other than English, French, Spanish or Portuguese. Two reviewers independently selected papers, assessed framework quality using the Appraisal of Guidelines for Research and Evaluation for Health Systems (AGREE-HS), and extracted data. A thematic analysis was then conducted.

Results:

A total of 18 AI governance frameworks for HS were identified. Most were published between 2022 and 2024 and were rated as moderate or low quality. The frameworks targeted four levels: international (n = 3), national (n = 4), local (n = 3) and organizational (n = 8). The composition of actors within governance structures varied across levels. Six key governance processes in HS emerged as critical: (1) Needs and/or problem identification; (2) Data governance; (3) Risk assessment and management; (4) Validation and/or evaluation; (5) Maintenance and monitoring; and (6) Integration. Four pivotal relational mechanisms were identified: (1) Ethical principles and/or values; (2) Education and training; (3) Standards and regulations; and (4) Communication. Barriers and challenges included poor alignment across interconnected policy domains and governance levels, the substantial resources required for effective implementation, and the ongoing demands of continuous monitoring of AI-based technologies.

Conclusions:

The findings highlight key elements of AI governance, including structures, processes and relational mechanisms, distributed across the international, national, local and organizational levels. The proposed Integrative AI Governance Model for HS aims to inform constructive policy and practice discussions on how to ensure the responsible, sustainable and effective integration of AI-based technologies into HS. Clinical Trial: The protocol for the systematic review has been registered on the Open Science Framework (OSF) and is available at: https://doi.org/10.17605/OSF.IO/MCBTS