"Should we use OMOP, FHIR, or i2b2?" — every organization starting its healthcare data journey asks this. There is no one-size-fits-all answer. This article helps you pick the right tool for the right use case.
1. Map of healthcare CDMs
2. Comparison table
| Criterion | OMOP | FHIR | i2b2 | PCORnet | Sentinel |
|---|---|---|---|---|---|
| Owner | OHDSI (open) | HL7 (open) | Harvard (open) | PCORI (US) | FDA (US) |
| Year | 2008 | 2014 | 2007 | 2014 | 2008 |
| Primary goal | Multi-source RWE | Operational exchange | Self-service queries | Pragmatic trials | Drug safety |
| Schema | 37 relational tables | Resource (REST) | Star schema (fact + dim) | Flat tabular | Flat tabular |
| Vocabulary | Standard Concepts (normalized) | CodeableConcept (flexible) | Local | Local + standard | Local |
| Open tooling | ATLAS, HADES, DQD, ACHILLES | HAPI, Cerner SMART | i2b2 web client | Build-your-own | Closed |
| Real-time | ❌ | ✅ | ❌ | ❌ | ❌ |
| Federated network | ✅ EHDEN, DARWIN, OHDSI | ⚠ via Bulk Export | ✅ SHRINE | ✅ DRN | ✅ |
| Best for | Observational analytics | EHR / mobile / cloud | Hospital queries | Pragmatic RCTs | Pharmacovigilance |
| Adoption (2026) | 800M patients | Industry-wide | ~200 sites | ~70 sites | FDA-internal |
| Vietnam community | Just starting | Growing fast | Limited | None | None |
3. Core differences
3.1 Schema philosophy
OMOP = a normalized relational DB → optimized for SQL analytics. FHIR = document-oriented Resources → optimized for API exchange.
3.2 Vocabulary
OMOP requires every code to be mapped to a Standard Concept (e.g. ICD-10 → SNOMED). FHIR does not require this — you can keep local code systems if you stay internally consistent.
→ OMOP is more expensive at the ETL stage, but it pays off with much stronger analytics.
3.3 Federated vs. centralized
OMOP lets you run a study without moving data: the R study package runs locally at each partner and only aggregate results are shared. This fits Vietnam very well given the Personal Data Protection Law 2025 (effective 2026-01-01, replacing Decree 13/2023/NĐ-CP), which requires sensitive data to be stored in-country and demands explicit consent for each new processing purpose.
4. Decision tree
5. Hybrid pattern: FHIR + OMOP
This is the most common 2026 pattern:
→ Operational EHRs use FHIR (real-time, web/mobile-friendly), while the research data lake uses OMOP (federated analytics). Read the FHIR ↔ OMOP bridge article.
6. Why OMOP wins for RWE
For comparison: i2b2 has a smaller community and fewer updates; PCORnet is more US-centric; Sentinel is FDA-only; CDISC is for trial submissions.
7. When NOT to choose OMOP
- You only need real-time EHR exchange → FHIR is enough
- The dataset is tiny (a few hundred patients) → direct SQL queries are simpler
- You are submitting a new drug to the FDA → SDTM is mandatory
- US pharmacovigilance → Sentinel
- The organization has no data engineers → OMOP ETL is a non-trivial investment
8. Recommendations for Vietnam
| Organization | Recommendation |
|---|---|
| Large public hospital | FHIR (operational) + OMOP (research data warehouse) |
| Private hospital | FHIR is enough at first; add OMOP when starting RWE work |
| Research institute | OMOP as the centerpiece, ETL from multiple partner hospitals |
| BHYT / regulators | OMOP for the national data lake, FHIR to exchange with hospitals |
| Healthcare AI startup | FHIR + OMOP — FHIR for integration, OMOP for training data |
9. Migration between CDMs
You can migrate from:
- i2b2 → OMOP: Georgia Tech's i2b2-to-OMOP tool
- PCORnet → OMOP: ETL maintained by Duke
- FHIR → OMOP: FHIR-OMOP-on-FHIR (HL7 + OHDSI)
- Sentinel → OMOP: when you need deeper analytics
OMOP is usually the final destination because the tooling ecosystem is the richest.
10. Community resources
- OHDSI Forum — answers within 24 hours
- Book of OHDSI — free textbook
- EHDEN Academy — free courses
- OHDSI Symposium — annual, with a virtual track
- Working groups — Vocabulary, Themis (ETL conventions), AI/ML, Vietnam (none yet — waiting for someone to launch it!)
Conclusion
CDMs are not zero-sum. Most organizations in 2026 use FHIR + OMOP side by side: FHIR for operations, OMOP for analytics. If you are in Vietnam and want to do RWE, train clinical AI, or run public-health research — OMOP is a worthwhile investment.
Next article: Standardized Vocabularies & Athena — the heart of OMOP CDM.