Evaluating Artificial Intelligence in Clinical Settings-Let Us Not Reinvent the Wheel

Area of Science:

• Health Informatics
• Artificial Intelligence in Healthcare
• Technology Evaluation Frameworks

Background:

• Healthcare increasingly uses technology, necessitating risk mitigation and benefit maximization.
• Evaluating artificial intelligence (AI) in healthcare requires robust frameworks.
• Existing health information technology (HIT) evaluation frameworks offer valuable insights.

Purpose of the Study:

• To provide an overview of how theory-informed HIT evaluation frameworks can guide AI development and implementation in healthcare.
• To illustrate practical applications of existing frameworks for AI in health settings.
• To support developers, implementers, and decision-makers in safe AI adoption.

Main Methods:

• Conceptual overview of selected theory-informed HIT evaluation frameworks.
• Illustrative examples of framework application to AI in healthcare.
• Focus on pragmatic and stakeholder-driven requirements.

Main Results:

• Existing HIT evaluation frameworks can inform AI development by addressing technology, user, and organizational dimensions.
• Frameworks facilitate the design, implementation, and scalability of AI technologies.
• These approaches support evidence-based decision-making for AI in healthcare.

Conclusions:

• Theory-informed HIT evaluation frameworks are essential for the safe and effective integration of AI in healthcare.
• Applying these frameworks enhances the design, implementation, and sustainability of AI solutions.
• This approach supports maximizing benefits while minimizing risks in health AI.