AI-Enabled Digital Health Tool Implementation in Clinical Practice Settings: Umbrella Review

AI has the potential to improve clinical decision-making, efficiency, and patient outcomes, yet adoption in routine clinical practice has been slower than anticipated. While prior studies have identified factors that influence AI implementation, a comprehensive synthesis grounded in implementation science frameworks is lacking.

This study aimed to qualitatively and comprehensively describe barriers and facilitators to AI implementation in clinical care settings using an implementation science determinants framework, the Consolidated Framework for Implementation Research (CFIR), and to identify AI-specific determinants not fully captured by CFIR.

We conducted an umbrella review of reviews examining determinants of AI implementation in clinical practice. In accordance with the PRIOR (Preferred Reporting Items for Overviews of Reviews) reporting guidance, we systematically searched PubMed, Scopus, and CINAHL for English-language reviews published between January 2013 and July 2025. Reviews were included if they addressed barriers to and facilitators of implementing AI-enabled digital health tools in clinical care settings. Determinants were deductively mapped to the updated CFIR and inductively analyzed to identify emergent AI-specific constructs. Review quality was assessed using the Joanna Briggs Institute critical appraisal checklist, and overlap among primary studies was evaluated using the GROOVE (Graphical Representation of Overlap for Overviews) tool.

Twenty-two reviews published between 2021 and 2025 were included, encompassing studies conducted from 1990 to 2025. Reviews varied in design, including systematic, scoping, rapid, narrative, and literature reviews. Overall review quality was mixed, with 9 (41%) reviews rated at 100% quality and 4 (18%) at 50% or lower. Primary study overlap was slight, with a corrected covered area of 0.4%. Determinants of AI implementation were identified across all CFIR domains. Prominent barriers at the innovation level included limited evidence of clinical impact, poor interpretability, complexity, lack of adaptability, and insufficiently representative data. Facilitators included perceived relative advantage, co-design with clinicians, and evidence of local clinical benefit. At the implementer level, low AI literacy, limited confidence, time constraints, and concerns about professional displacement were common barriers, while training and technical support facilitated adoption. Inner setting barriers included inadequate infrastructure, limited resources, poor workflow compatibility, and weak governance structures. Outer setting barriers included unclear policies, laws, and liability frameworks, whereas partnerships and cross-sector collaboration were frequent facilitators. Inductively identified AI-specific determinants included transparency, trust, accountability, model bias, data governance, model maintenance, robustness, generalizability, existential threat, and concern about loss of the human element in care.

AI implementation in clinical settings is shaped by complex, multilevel, and AI-specific determinants that extend beyond traditional digital health considerations. Implementation science frameworks can be adapted to incorporate these constructs. Tailored, theory-informed strategies addressing technical, organizational, psychological, and regulatory factors are essential to support safe, equitable, and sustainable integration of AI into clinical care.