The AMA’s SNOMED CT–to–CPT Mapping Initiative: An Analytical and Practical Review

By Elizaveta Bannova, Billing Department, WCH

What Was Announced

On July 14, 2026, the American Medical Association announced a new interoperability initiative to build and deploy formal terminology mappings between SNOMED CT® (Systematized Nomenclature of Medicine – Clinical Terms) and CPT® (Current Procedural Terminology). The goal, as stated in the release, is to support electronic prior authorization by connecting clinical terminology with administrative coding.

The purpose is narrow and specific: reduce the administrative burden prior authorization places on physicians, cut the barriers it creates to timely patient care, and support the CMS Health Tech Ecosystem’s push to modernize prior authorization and improve interoperability. AMA President Willie Underwood III made a point of framing this as a terminology and governance problem rather than a software problem — modernizing prior authorization, he said, requires trusted, interoperable terminology assets, not just electronic bridges between applications.

On the practical side, the AMA is calling on payers, EHR vendors, clearinghouses, prior-authorization vendors, and standards organizations to join collaborative testing, technical sessions, and workflow validation. The initiative will start with the prior authorization scenarios that offer the biggest reduction in administrative burden and the clearest benefit to patients waiting on care decisions.

Why This Is Happening Now

This isn’t a standards project born out of general goodwill — it’s a direct response to a hard compliance deadline. Under CMS’s Interoperability and Prior Authorization Final Rule (CMS-0057-F), impacted payers — Medicare Advantage organizations, state Medicaid/CHIP programs, Medicaid managed care plans, and QHP issuers on the federal exchanges — must have several FHIR®-based APIs operational by January 1, 2027.

One of these is a dedicated Prior Authorization API, built on HL7 FHIR Release 4.0.1, the US Core Implementation Guide STU 3.1.1, and the SMART App Launch Implementation Guide Release 1.0.0. The same rule sets new decision-timeliness standards: 72 hours for urgent requests, seven days for standard ones, with a specific reason required for every denial.

That’s the real problem the AMA initiative is trying to solve. A FHIR-based Prior Authorization API can move a request electronically in milliseconds, but the request still has to state what clinical situation justifies what billable service. Right now that translation happens manually, inconsistently, or through each payer’s own proprietary rules engine. As CMS and the industry push toward FHIR-based prior authorization ahead of the 2027 deadline, this mapping is meant to make real-time electronic authorization actually work by linking clinical concepts to procedural codes.

Two Vocabularies That Were Never Built to Match

It’s worth being precise about what SNOMED CT and CPT actually are, because they aren’t two dialects of the same language — they’re built for entirely different jobs.

SNOMED CT is a clinical reference terminology: tens of thousands of concepts — findings, disorders, procedures, body structures, organisms — linked by formal logical relationships. It’s designed to represent what a clinician observed or did, at whatever level of detail the record captured.

CPT, owned and maintained by the AMA, is an administrative code set: a closed, numbered list of billable services, each tied to specific documentation and reimbursement rules, updated annually through the CPT Editorial Panel and priced by the RVS Update Committee (RUC). It doesn’t describe clinical reality in the abstract — it’s a payment unit.

Because of that, mapping one to the other isn’t a lookup-table job. It’s a lossy, many-to-many translation between a clinical ontology and a billing taxonomy, and there are several places where it gets genuinely messy:

Granularity mismatch. One SNOMED CT concept can correspond to several CPT codes depending on surgical approach, laterality, or implant use — details that live in the clinical note, not in the diagnosis concept.
One-to-many collisions. Some cases map cleanly to a single CPT code; others map to a set of plausible codes, and the right one often depends on payer-specific or specialty-specific coding conventions that have nothing to do with SNOMED CT.
Two release cycles, one snapshot. CPT updates annually. SNOMED CT (distributed in the US through the National Library of Medicine as part of UMLS) has its own schedule. Any mapping built today starts decaying the moment either vocabulary changes, which is presumably why the AMA is framing this as an ongoing program with pilots and working sessions rather than a single deliverable.
Licensing. CPT is proprietary AMA intellectual property with its own licensing terms; SNOMED CT is distributed free to US users through a national release center. A cross-map inherits the licensing baggage of both sides — something every vendor and payer will need legal sign-off on before integrating it, not just engineering sign-off.

Where This Fits in the Bigger FHIR Picture

The mapping is meant to plug into the existing Da Vinci Prior Authorization Support (PAS) and Coverage Requirements Discovery (CRD) implementation guides — the HL7 FHIR specs that already define how a prior authorization request gets transmitted, but not how a clinical concept in the EHR turns into the procedure code a payer’s rules engine expects. This aligns with the FHIR implementation approaches and CMS priorities already in motion.

Da Vinci PAS solved the pipe — the transport layer. This mapping is an attempt to solve what flows through it: making sure the data moving through the Prior Authorization API is coded correctly on both the clinical and billing side without a human coder bridging the gap in real time.

That matters because the 2027 deadline is going to produce a lot of FHIR APIs that are technically compliant but clinically shaky underneath, if the coding logic feeding them is inconsistent from payer to payer. A payer-specific convention sitting on top of an inconsistent SNOMED-to-CPT crosswalk would just recreate today’s chaos inside a nicer API — faster, but no easier to audit.

What’s Actually Strong Here, and What Isn’t Settled Yet

What works about this approach:

Putting the mapping under one governing body instead of letting every payer or vendor build its own crosswalk avoids the usual interoperability failure — everyone solving the same problem slightly differently and calling the result a standard.
Starting with a handful of high-friction prior-authorization scenarios instead of trying to build a universal crosswalk on day one is sound engineering. It produces testable results fast instead of trying to boil the ocean.
Because the AMA owns CPT outright, it can keep the mapping synced to each annual code cycle in a way a third-party vendor never could.

What’s still open:

The conflict-of-interest question. The AMA owns CPT — a meaningful licensing revenue stream — and is now also positioning itself as steward of the crosswalk that determines how clinical data turns into billable CPT events. Payers and vendors will reasonably ask about update transparency and whether this creates new licensing costs on top of existing CPT fees.
Accuracy at the edges. The scenarios causing the most prior-authorization friction today — high-cost imaging, specialty procedures, oncology regimens — are also the scenarios with the messiest one-to-many code relationships. Early pilots will likely show that a meaningful chunk of real cases still need a rules engine or a human to sort out, mapping or no mapping.
Maintenance governance isn’t public yet. Nothing so far specifies release cadence, error correction, or whether the mappings will be published under an open, auditable methodology or kept as a closed AMA data product.
Adoption isn’t mandatory. CMS-0057-F mandates the transport APIs. It says nothing about this specific mapping. Whether it gets used depends entirely on EHR vendors and payers choosing to build it into their systems — which is why the pilot participation the AMA is asking for matters more than the announcement itself.

What Different Stakeholders Should Actually Do

Health plans: Treat the mapping as a candidate input to your rules engine, not a replacement for it. Validate it against your own historical authorization decisions before trusting it in production, and get licensing terms clarified with the AMA before committing to integration work.

EHR and health IT vendors: The mapping has the most value at order entry — where a SNOMED CT-coded problem or order can suggest the right CPT code(s) for a pending prior authorization through the Da Vinci PAS/CRD workflow. It makes more sense as an upgrade to existing coding-assist modules than as a new standalone product.

Physician practices and billing staff: This doesn’t remove the coder from the process anytime soon. It’s aimed at cutting friction on prior-authorization-specific scenarios; general coding accuracy, modifiers, and documentation still fall on people. Practices should keep an eye on AMA pilot invitations — early participants will have more say in which use cases get prioritized.

Standards bodies and Da Vinci implementers: This mapping becomes a content dependency for PAS/CRD workflows. Watch for updates that formally reference AMA mapping artifacts as a recommended or optional data source.

Where This Leaves Things

This initiative isn’t really a new technology — it’s an attempt to close a semantic gap that the CMS-0057-F mandate exposed. The industry has agreed on how to move prior-authorization data electronically by 2027, but there’s still no shared, authoritative way to translate clinical documentation into the procedural codes that authorization decisions actually run on.

Whether this becomes real infrastructure or just another crosswalk that payers quietly route around with their own logic depends on how well it holds up in the messiest specialty cases, how transparent the governance and licensing terms end up being, and whether EHR vendors actually build it in before the 2027 deadline. The pilot testing the AMA mentions in its own release is the thing worth watching — much more than the press release itself.

Sources