Implementation Frameworks for Artificial Intelligence Translation Into Health Care Practice: Scoping Review

Area of Science:

• Health Informatics
• Artificial Intelligence in Healthcare
• Implementation Science

Background:

• Artificial intelligence (AI) offers significant potential for healthcare improvement.
• Healthcare professionals face challenges in integrating AI into daily clinical practice.
• Existing implementation frameworks may not adequately address AI's specific requirements.

Purpose of the Study:

• To identify and analyze existing implementation frameworks for AI in healthcare practice.
• To understand the current state of knowledge regarding AI implementation in healthcare.
• To highlight gaps in current frameworks for effective AI integration.

Main Methods:

• A comprehensive scoping review of multiple databases (Cochrane, Embase, MEDLINE, PsycINFO) was conducted.
• Publications from 2000 onwards, in English, focusing on AI implementation in healthcare were screened.
• Thematic analysis was performed using the Nilsen taxonomy and the NASSS framework.

Main Results:

• Only 7 articles met eligibility criteria; 2 provided formal AI implementation frameworks.
• Identified elements aligned with the NASSS domains, but no single framework was comprehensive.
• New domains crucial for AI implementation were identified: data dependency, shared decision-making, human oversight, and ethics.

Conclusions:

• Understanding AI implementation in healthcare is nascent.
• Existing frameworks require adaptation or new development to guide AI integration.
• Further research drawing on implementation science is essential for developing robust AI implementation frameworks.