Study Proposes Integrative AI Governance Model for Health Systems

Per the JMIR preprint by Alami et al., the authors conducted a systematic review of AI governance frameworks for health systems covering November 2014 to July 2025 and searched eight academic databases, grey-literature databases, and web portals of international organizations (Alami et al., JMIR Preprints #87448, 2026). The preprint reports that current frameworks aim to address risks such as bias, compromised quality of care, data breaches, and adverse financial impacts but frequently fail to reflect the multidimensional and dynamic characteristics of AI governance in health systems. The paper proposes an Integrative AI Governance Model that identifies key components intended to guide AI-related policy, practice, and research in health systems (Alami et al., 2026).

Per the methods section of the JMIR preprint, the review included peer-reviewed articles and reports sourced from PubMed, MEDLINE, Embase, ACM Digital Library, Web of Science, Scopus, Social Sciences Abstracts, and PsycINFO, alongside grey literature and international organization outputs. The manuscript synthesizes conceptual elements across frameworks rather than presenting a new empirical evaluation of deployed AI systems; the proposed integrative model aggregates recurring governance domains, mechanisms, and actor roles observed in the literature (Alami et al., 2026).

The Duke-Margolis white paper, "AI Governance in Health Systems," frames governance as a process of aligning innovation, accountability, and trust and documents contributions from a multi-stakeholder working group including health system privacy, data science, and engineering leaders (Duke-Margolis Institute for Health Policy, 2024). Reporting by Duke-Margolis emphasizes operational governance activities such as cross-functional working groups, risk assessment workflows, and alignment with existing clinical governance structures, which map onto many of the components the JMIR preprint synthesizes.

Industry observers note that synthesizing governance concepts into an actionable integrative model is a common next step after fragmented guidance emerges. Health systems attempting to operationalize AI governance typically confront integration problems across data governance, clinical validation, procurement, and regulatory compliance; the literature synthesis in Alami et al. consolidates these recurring domains into a single framework that can aid comparative evaluation and policy design.

For practitioners, the combined corpus (JMIR preprint and Duke-Margolis white paper) consolidates conceptual governance building blocks-such as stakeholder roles, risk assessment, monitoring, and accountability pathways-that frequently appear in health-sector guidance. Editorial analysis: Standardizing terminology and domain boundaries, as the JMIR preprint attempts, reduces friction when health systems translate high-level principles into clinical workflows, procurement criteria, and audit processes.

Observers should track whether the JMIR preprint undergoes peer-reviewed publication in J Med Internet Res (the preprint lists J Med Internet Res 2026;28:e87448) and whether subsequent case studies or toolkits operationalize the Integrative AI Governance Model. Industry context: Metrics that would show uptake include published implementation case studies, alignment of institutional policies with the model's domains, and the development of practical artifacts such as checklists, templates, or evaluation rubrics by health systems or regulators.

The JMIR preprint is presented as a preprint under review and the authors indicate the work synthesizes existing frameworks rather than reporting new empirical validation. The Duke-Margolis white paper represents a practitioner-oriented synthesis with contributions from invited working-group members and was funded by Duke AI Health, as stated in the white paper.