140+ FHIR resources exist, but seven core resources handle 80% of use cases. Understanding their proper use prevents integration mistakes.
**1. Patient** — Identity, demographics. Foundation of everything.
**2. Encounter** — Visits, admissions. Ties clinical data to visit.
**3. Observation** — Labs, vitals, structured findings. Most common resource.
**4. Condition** — Diagnoses, problems. SNOMED CT + ICD-10 coded.
**5. MedicationRequest** — Prescriptions. RxNorm coded.
**6. Practitioner** — Clinicians. NPI and identity.
**7. Organization** — Facilities, health systems. Structural context.
Correct usage patterns
1. References vs. contained. Reference across resources; never contain shared entities. 2. Profile conformance. US Core baseline for US deployments. 3. Terminology bindings. Every coded field bound to specific ValueSets. 4. Bundle for related writes. Related resources written as one transaction.
Common resource misuse
1. Patient with contained MedicationRequest → duplication. 2. Observation without profile → no validation. 3. Encounter without location → workflow gaps. 4. MedicationRequest without RxNorm → cross-drug analytics broken. 5. Practitioner without NPI → identity resolution issues.
Search parameters worth knowing
1. Patient?identifier= — find by MRN, SSN. 2. Observation?category=laboratory&_sort=-date — recent labs. 3. Condition?patient=X&clinical-status=active — active problems. 4. MedicationRequest?patient=X&status=active — current meds. 5. Encounter?patient=X&_include=Encounter:location — visits with location.
Vendor state
All major FHIR servers support these seven resources fully. Focus on profile conformance and terminology binding depth.
Getting the core seven resources right unlocks most FHIR value. Add specialized resources (SDC, PlanDefinition, Task) as use cases dictate.
