EHR development in 2026 follows specific practices that separate maintainable from technical-debt-heavy implementations.
**1. FHIR-native storage from day one.** Skip proprietary + facade duplication.
**2. SMART on FHIR for all auth.** Internal service-to-service too.
3. Reference implementations before custom. Use LHC-Forms, Inferno, etc.
4. Declarative over imperative. StructureMap for transforms, ConceptMap for translations.
5. Conformance testing in CI. Inferno on every deploy.
6. Terminology as its own subsystem. Not embedded in application code.
7. Metrics from day one. Per-resource-type observability.
Complexity accumulators to avoid
1. Custom internal + FHIR facade. 2. Custom auth vs. SMART. 3. Custom form renderer vs. LHC-Forms. 4. Manual conformance testing. 5. Terminology mappings in application code.
Investment sequence
1. FHIR-native storage (baseline). 2. SMART for all auth (day 1-30). 3. Reference implementations (day 30-60). 4. Declarative transforms (day 60-90). 5. CI conformance (day 90+). 6. Terminology subsystem (ongoing). 7. Observability (ongoing).
Team skills
1. FHIR spec fluency (all team). 2. HL7v2 integration (1-2 people). 3. Terminology understanding (1 person). 4. Auth/security expertise (1 person). 5. Backend engineering (all team). 6. DevOps/observability (1-2 people).
EHR development in 2026 is engineering discipline more than architecture debate. The seven practices above cover most of the difference between successful and struggling deployments.
