Skip to content

When to Use OMOP

Emory has multiple data platforms. Choosing the right one saves weeks of work. This guide helps you decide when Enterprise OMOP is the right fit — and when you should use Epic Clarity, Caboodle, or the Reporting Workbench instead.

OMOP vs. Epic Resources at a Glance

Enterprise OMOP Epic Clarity / Caboodle Reporting Workbench
Best for Multi-source research, standardized analytics, federated studies Operational reporting, real-time data, full encounter detail Quick ad-hoc operational reports
Data sources Epic Clarity + Legacy CDW (Cerner) + Tumor Registry + OpenSpecimen Epic only Epic only
Coding Standardized (SNOMED, RxNorm, LOINC) Source codes (ICD-10, CPT, NDC) Source codes
Structure Patient-centric, cleaned, de-duplicated Encounter-centric, raw Pre-built report templates
Timeliness Periodic refresh (lags real-time) Near real-time Near real-time
Cross-site Yes — OHDSI network-compatible No No
Vocabularies Hierarchies + mappings built in Manual code lists Manual code lists

Use OMOP When...

  • Comparative effectiveness research

    Comparing treatments, outcomes, or utilization patterns across patient populations using standardized codes and clean data.

  • Federated or multi-site studies

    Running analyses that need to be portable across OHDSI network sites. OMOP's standard structure means your query works everywhere.

  • Linking across data sources

    Combining EHR data with tumor registry (Winship) or biospecimen data (OpenSpecimen) in a single, patient-centric model.

  • Cohort definitions using standard vocabularies

    Using SNOMED hierarchies to capture "all diabetes" without manually listing every ICD-10 code. The vocabulary does the work.

  • Health equity and SDoH analyses

    Standardized demographics, geographic data, and observation-domain social determinants in a research-ready format.

  • You need cleaned data

    Non-events (cancelled encounters, missed medications, future procedures) are excluded. What's in OMOP is what actually happened.

Don't Use OMOP When...

  • You need real-time or near-real-time data

    OMOP refreshes on a periodic cadence. If your study depends on today's data, start with an Epic resource.

  • You need the full encounter narrative

    OMOP captures what happened to the patient, not everything the provider intended. Cancelled orders, missed medications, and charge-only entries are excluded. Use Clarity for the complete picture.

  • Intent-to-treat analysis

    If your study is about what care the provider tried to deliver (not what the patient received), Clarity or the CDW is the better source.

  • Historical demographic tracking

    OMOP's person table is a snapshot — it doesn't track address changes, insurance transitions, or prior name/gender over time.

  • Operational or billing reports

    Charge data, scheduling workflows, and real-time census are Epic's domain. OMOP is built for research, not operations.

What OMOP Actually Is (Under the Hood)

Enterprise OMOP is a structural standardization and semantic harmonization of the data in Epic Clarity and the legacy Clinical Data Warehouse (primarily Cerner). In practice, that means:

Structural standardization
All medications land in drug_exposure, all diagnoses in condition_occurrence, all labs in measurement — regardless of which source system they came from.
Semantic harmonization
NDC codes map to RxNorm. ICD-10 maps to SNOMED. Local lab codes map to LOINC. The vocabulary layer translates source codes into a shared language.
Cleaning
Non-events are removed. Records are de-duplicated where applicable. The result is a curated research dataset, not a raw operational dump.

OMOP is not a replacement for Clarity

It's a research-optimized view of the same underlying data. If something looks unexpected in OMOP, the source of truth is always the Epic infrastructure. See our Known Issues page for documented limitations.

Next Steps

Ready to explore the data model? Head to the OMOP Primers landing page for table-by-table guides, or jump straight to the tables you'll use most: