Regulation — healthcare-specific
CMS AI Interpretive Guidance — State Operations Manual Appendix A
CMS translated its Conditions of Participation into AI-explicit survey language for Appendix A (hospitals). The surveyor question shifted from "do you have an AI policy" to "show me the audit output that confirms your AI performs equitably across patient groups your hospital serves."
What the guidance actually asks (paraphrased survey language)
"Does the hospital audit AI-driven clinical-decision-support tools for adverse-impact patterns across protected classes and clinical patient groups, using measurement independent of the tool's own self-reported metrics?"
CMS AI Interpretive Guidance — representative survey prompt for Appendix A
"Is there evidence that the hospital's medical staff and quality-improvement structure review AI-driven decision outputs on a defined cadence, with escalation and pause criteria when patterns of concern emerge?"
CMS AI Interpretive Guidance — QAPI + medical-staff intersection
What this means in plain English
The surveyor is not going to ask if you have AI. They will ask what you know about how it performs on subpopulations you actually serve, and how you learned it. Two follow-ups:
- Show the measurement. It must be independent of the vendor's own dashboard.
- Show the response protocol. Freeze/hold criteria + escalation targets + medical-staff involvement.
A vendor screenshot is not evidence. A signed independent-verifier attestation is.
What triggers the exposure in the sample
The ED-throughput dashboard stayed green throughout the 45+ day drift window. The hospital's operational measurement did not surface the H3 differential. The Snapshot did. That gap — between vendor/operational instrumentation and independent measurement — is exactly what the CMS surveyor's follow-up question is designed to expose.
What the $499 Snapshot shows against this rule
- Independent measurement of the AI's per-group behavior (distinct model family, distinct math, distinct retention)
- Documented threshold + severity classification + first-drift-day — the timeline evidence
- 3 fix-first items including escalation targets + freeze/hold criteria — the response-protocol evidence
- Signed independent-verifier declaration — the attestation the surveyor will look for on the file
See how the Category-C determination is derived →
$499 Snapshot. 3 business days.
The specific record the CMS surveyor is trained to ask for — independent-verifier evidence, not vendor self-attestation.
Buy $499
Snapshot credit applies to Baseline ($2,500) or Enterprise Attestation ($35-55K) upgrade within 30 days.