Summary
Provides consensus practices and evaluation methods for AI systems that compare requested services, procedures, or medications with payer medical-necessity criteria, addressing criteria accuracy and versioning, matching performance, transparency, fairness, safety, workflow integration, and human review.
Healthcare Implications
Payers, providers, and vendors should validate criteria matching against current policies, document source and version information, test false matches and subgroup effects, preserve qualified human review and appeals, monitor changes, and prevent automated matching from becoming an unreviewable basis for denying care.