5 FHIR Terminology Servers for ICD-11 Pilot Programs in 2026

ICD-11 is the forward-looking coding system most US health systems will eventually have to support, even though ICD-10-CM remains the operational standard in 2026. Health systems running pilot programs against ICD-11 face a specific terminology-server question: which servers handle the new structure, the post-coordination linkages, and the ICD-10 to ICD-11 translation cleanly enough for a research-grade pilot. The five servers below have shown up in real ICD-11 pilots in 2026. For the FHIR implementation playbook, the broader reference covers the surrounding patterns.

The Five Servers Showing Up in ICD-11 Pilots

Ontoserver. The most common pick because CSIRO and the ICD-11 maintainers have a long-running working relationship. Post-coordinated codes load and expand correctly, which is the single hardest part of an ICD-11 deployment.

Snowstorm. SNOMED International's server. Has added ICD-11 loading capabilities and a translation extension to bridge between SNOMED CT clinical findings and ICD-11 codes.

HAPI Terminology. Open-source. Pilot teams adopting HAPI for ICD-11 have to do more manual loading and tuning than commercial pickers, but the code is open enough to extend for research-specific needs.

Tx-Server. The reference server. Used in some pilots specifically for its tight adherence to the FHIR terminology spec, which surfaces ICD-11 conformance questions early.

Smile Digital Health Tx. The commercial bundle. Wins on operator effort; loading ICD-11 is largely automatic, and post-coordination support ships out of the box.

The five represent the realistic options for an ICD-11 pilot in 2026; large-scale production deployments will follow once US regulators commit to a transition timeline.

Three Capabilities That Make or Break an ICD-11 Pilot

A health system running an ICD-11 pilot should test three capabilities specifically. Post-coordination expression: ICD-11 supports building codes from a stem and modifiers, and the server has to handle the post-coordinated expression as a first-class lookup target, not just a frozen string. ICD-10-CM to ICD-11 translation: pilot programs typically dual-code for a transition period, and the server has to round-trip between the two reliably. Foundation-vs-MMS handling: ICD-11 has two layers, the Foundation Component and the Mortality and Morbidity Statistics linearization, and the server should distinguish them cleanly so the pilot team can choose which one each query hits.

A six-week pilot against a research cohort with dual coding surfaces the operational realities. The oncology cornerstone covers the broader server-selection framework, and the tribal health coding walkthrough shows the same logic applied to a less-common coding context.

Why Pilot Fit Matters More Than Production Optimization

ICD-11 pilots are about learning, not about scaling. The right server for a pilot is the one that lets the research team explore the new vocabulary's structure without spending most of their time on infrastructure. The five servers above all clear that bar in 2026 pilot deployments, which is the practical reason they made the list ahead of others that may technically support ICD-11 but have not been put through pilot workloads yet.

Pilot programs that test these servers against dual-coded research data surface the operational realities of ICD-11. The post-coordination layer and the cross-vocabulary translation step together separate the servers ready for the transition from those that handle the easy part and stumble on the rest.

Beyond the five servers above, the academic FHIR-implementation community has built tooling that wraps these servers with ICD-11-specific helpers. That tooling is worth evaluating as a productivity layer for pilots that go beyond the basic vocabulary load.

The healthcare-IT teams running ICD-11 pilots in 2026 still consider these five servers the most viable starting points for the transition, ahead of vendors that have not yet exercised the post-coordination layer under real research-grade workloads.