Summary
Provides consensus practices and evaluation methods for AI-enabled clinical decision support, with a focus on large-language-model and retrieval-augmented systems that deliver evidence-based medical information at the point of care, including evidence concordance, citation accuracy, usability, safety, and transparency.
Healthcare Implications
Developers and health systems should validate outputs against specialty guidelines and reference evidence, verify citations, measure hallucinations and clinically significant errors, preserve clinician responsibility, test workflow usability, and monitor performance and safety after deployment.