Presentation (CTI)

OBJECTIVES: There is hope that the use of AI could have the potential to transform healthcare by assisting in different aspects including care decision-support, as well as optimized administrative processes. Nevertheless, there are different barriers to successful implementation. Implementation science offers frameworks to overcome barriers and support integration of new research findings into routine healthcare. The aim of this review was to analyze the existing literature on implementation frameworks regarding their use for AI in healthcare, and identify gaps in current knowledge.
METHODS: The scoping review was conducted in accordance with the JBI methodology for scoping reviews. We searched Medline, CENTRAL, and Embase to identify studies of implementation frameworks for AI adoption in healthcare published since 2020.
RESULTS: A total of 1906 unique publications were found and screened. Established implementation frameworks, including the Non-adoption, Abandonment, Scale-up, Spread, and Sustainability framework (NASSS), the Reach, Effectiveness, Adoption, Implementation, and Maintenance framework (RE-AIM), and the Technology Acceptance Model (TAM), remain in use for the implementation of AI-based innovations as well. Additionally, the complex nature of AI technologies has led to the development of some AI-specific implementation models (e.g., FAIR-AI, AI-QIF, and AI in the NHS framework). Economic assessment was more addressed in frameworks such as the NICE Evidence Standard Framework.
CONCLUSIONS: While implementation frameworks commonly consider themes like stakeholder engagement, workflow integration, regulatory compliance, and AI governance, economic validation is rarely taken into account as an implementation step. In addition, many frameworks are developed from retrospective analyses, in single-site settings, and remain conceptual. Consequently, evidence on the comparative validation of these frameworks in real clinical settings remains limited.