Healthcare AI Learning
Course overview

Technical specialist lab · Module 1 · 75 min

Reading FHIR: Resources, JSON & References

Read a FHIR resource as a modelled clinical object rather than arbitrary JSON: ids versus business identifiers, the terminologies that supply coded clinical meaning, references between resources, searchset Bundles, and the version question you ask before any integration.

After this module you can

Open a Fast Healthcare Interoperability Resources (FHIR) payload from an unfamiliar system and read it structurally: what each resource is, how it identifies itself, what it points at, and what is still missing before it can become model input.

  • Read a FHIR JSON resource as a modelled object, not arbitrary JSON
  • Distinguish a resource id, a business identifier and a Reference
  • Recognise common Patient, Observation, Condition and Encounter fields
  • Tell a classification (ICD) from a clinical terminology (SNOMED CT) from an observation identifier (LOINC), and read them inside a CodeableConcept
  • State what AI-assisted coding can propose, and where human validation stays mandatory
  • Follow references between resources to assemble patient context
  • Explain what a searchset Bundle represents
  • Ask which FHIR version, profiles and implementation guide a system supports

Querying for these resources over a FHIR API is planned for Module 2, which is not yet released. Here you learn to read what comes back.

Step 1

“We have FHIR”

An AI team is told the hospital's platform is FHIR-enabled, so integration is solved. Six weeks later the model still cannot be given a usable input: the endpoint returns hundreds of Observations with no clinical grouping, the HbA1c results carry two different codings, half the Conditions have no onset date, and nobody can say which encounter a value belongs to.

What FHIR gives you

A shared resource model, defined data types, stable field semantics, references between resources, and an interoperable exchange format. Two systems can talk about a blood pressure the same way.

What it does not give you

Task relevance, a curated cohort, guaranteed completeness, consistent local coding, or an AI-ready feature table. A conformant server can still return data that is unusable for your specific question.

Version reality check

  • FHIR R5 (5.0.0) is the current published core release of the specification.
  • Many production deployments still run R4 or R4B. SMART App Launch 2.2, a standard app-launch and authorisation framework for FHIR systems, is published against R4.
  • R6 is a ballot/development sequence, not the production baseline you should assume.
  • Examples in this module use stable, R4-compatible shapes that are familiar in production systems.
  • Before integrating, inspect the target server's CapabilityStatement — the FHIR resource describing its supported resources, versions and operations — and the implementation guide it claims to conform to. Never assume a version, a profile or an extension is supported.

Representational State Transfer (REST) is the most common way to interact with a FHIR server, but it is not the only one. FHIR also defines messaging, documents and operations, and resources are routinely exchanged as files or bulk exports. Treat FHIR as a resource model plus exchange framework, not as “an API style”.

Imaging mostly lives outside these resources: pixel data uses Digital Imaging and Communications in Medicine (DICOM), exchanged over DICOMweb, with FHIR ImagingStudy acting as the pointer. It is treated here as an adjacent standard only.

Lab progress

0 of 7 steps complete

Sources & evidence · 6 sources

This module cites standards and specifications.

Content reviewed: September 2026. Publication dates of the individual sources are shown in each citation.

  • HL7 FHIR R5 (v5.0.0, STU) — current published specification

    The core specification. R5 (v5.0.0, STU) is the current published release; R4 and R4B remain in production use — verify the release implemented by the local system. Release-pinned; checked 10 September 2026.

    Open source

  • HL7 FHIR — directory of published versions

    Which sequences exist, what is published, and what is still in ballot.

    Open source

  • HL7 FHIR — Resource: Patient

    Field definitions, cardinality and the identifier element.

    Open source

  • HL7 FHIR — Resource: Observation

    code, status, effective[x] and the polymorphic value[x].

    Open source

  • HL7 FHIR — Resource: Bundle

    Bundle types, entries and search mode.

    Open source

  • HL7 FHIR — References between resources

    Relative, absolute, contained and logical references.

    Open source