Use First, Trust Later? Exploring How Health Care Providers View the Gaps Between AI's Regulation and Its...

Area of Science:

• Health informatics
• Artificial intelligence in healthcare
• Regulatory science

Background:

• Current regulatory frameworks for artificial intelligence (AI) in healthcare primarily emphasize premarket approval.
• This focus may inadvertently limit the consideration of AI's full life cycle and its potential for real-world adaptation.
• The dynamic nature of AI necessitates a broader regulatory perspective beyond initial approval.

Purpose of the Study:

• To highlight the limitations of a premarket approval-centric regulatory approach for AI in healthcare.
• To emphasize the need for continuous oversight and quality assurance of AI systems by healthcare providers.
• To underscore the importance of managing AI's real-world adaptiveness and life cycle potential.

Main Methods:

• Qualitative analysis of current regulatory trends in healthcare AI.
• Review of literature on AI life cycle management and real-world performance.
• Conceptual framework development for ongoing AI oversight in clinical settings.

Main Results:

• Regulatory focus on premarket approval overlooks AI's evolving nature and real-world performance.
• Healthcare providers-cum-deployers are essential for continuous quality assurance and oversight.
• Early adoption phases are particularly critical for managing AI adaptiveness and ensuring patient safety.

Conclusions:

• Ongoing oversight and quality assurance by healthcare providers are imperative for AI in healthcare.
• Regulatory strategies must evolve to encompass the entire life cycle of AI, including real-world performance and adaptability.
• A balanced approach integrating premarket evaluation with continuous post-deployment monitoring is necessary for safe and effective AI integration.