POST /v1/claim-proof) is planned, not live.
Underwriting / “audit-free wrap” requires a named reinsurer policy—field stays null until then.
What ClaimProof is for
Pre-submit / pre-pay integrity with cryptographically sealed digests and adversarial pressure-testing—reducing RADV exposure and reserve pressure. Not eliminating clawback liability by pure math.
Runtime receipt
Merkle root over evidence leaves + signed content hash at evaluation time. Hashes only on any public anchor path—no PHI on the wire.
Promotes to CREB
On FLAGGED/REJECTED or dispute, promote into CREB / CREB-RADV / AREB profiles through existing seal paths. CREB is not obsolete.
Shadow by default
Same model as Horizon: shadow → advisory → intervention. Intervention never auto-onboards.
Say / do not say
Say
- Continuous claim proof at evaluation time
- Immutable hash chain + optional digest anchor
- Adversarial self-audit before sample selection
- Indemnity only when a named underwriter policy exists
- CREB remains the court packet
Do not say (until true)
- RADV liability drops to zero on the balance sheet
- Claims are mathematically unassailable / audits obsolete
- We run a digital twin of CMS
- 99.98% certainty = clinical necessity proof
Schema & samples
- Public sample: /samples/claim-proof-sample.json
- JSON Schema v0.1 and OpenAPI stub in
docs/implementation/schemas/ - Build plan:
docs/implementation/SA3_CLAIM_PROOF_BUILD_PLAN.md
Relationship to Horizon Aegis & RADV
Horizon Aegis is the production MA readiness wedge: continuous shadow RADV on V22/V24 (the models current CMS RADV audits use). V28, MEAT, and B11 stay ready, not active. Seal paths are CREB-RADV, AREB-RADV, PIEB-RADV, and RREB-RADV — profiles of the four bundles, not a fifth engine. ClaimProof is the broader claim-level runtime envelope that can carry those RADV HCC profiles plus general MCO detection traces.
Detection is a commodity. Proof is the moat. ClaimProof is the receipt at decision time—not a reinvention of court evidence.