Back to all articles
Interoperability

What FHIR R4 adoption really requires of a legacy EHR estate

6 min read

Most conversations about FHIR adoption start with the API and end there, as if bolting a FHIR façade onto an existing system is the whole job. In practice, the API is the easy part. The harder work is underneath it: whether the data your system holds is structured and coded well enough to be exposed as clean, well-formed FHIR resources in the first place.

The gap between a FHIR endpoint and FHIR-native data

A system can expose a Patient or Observation resource without the underlying data ever having been captured in a structured way. Free-text diagnoses, inconsistent identifiers, and locally-invented code lists all still work over HL7 v2, because v2 tolerates ambiguity that FHIR consumers won't. The result is a FHIR API that technically conforms to the specification but returns resources too inconsistent to build real integrations against.

This is why platform architecture matters more than API tooling. On Excelicare, every clinical entry (a diagnosis, a medication, a result) is captured as a coded, structured record from the point of care, using SNOMED CT and the platform's own longitudinal data model. FHIR resources are a projection of that underlying structure, not a translation layer bolted on afterward.

What this means in practice for an integration team

  • Resource identifiers stay stable over time, because they're derived from the core record, not regenerated per-request.
  • Terminology bindings (SNOMED CT, and mapped ICD-10 where needed) are consistent across every resource, not dependent on which module produced the data.
  • Bulk export and real-time subscriptions draw from the same underlying data, so they never drift out of sync with each other.

None of this is a criticism of HL7 v2. It's still the right tool for a lot of real-time messaging, and the platform supports it in parallel for exactly that reason. The point is narrower: if FHIR adoption is the goal, the conversation has to start with the data model, not the API surface.

Where to start

If you're evaluating a platform's FHIR maturity, ask to see a CapabilityStatement and a sandbox before you ask about endpoints. Whatever comes back, whether consistent, coded, complete resources or a thin wrapper over legacy tables, tells you more about the real integration effort ahead than any feature list will.

Want this applied to your service?

Book a demo and let's talk about what's relevant to your specialty.