When we started designing Glance, we had a choice: build a proprietary data model optimized for our specific use cases, or adopt an existing standard and deal with its constraints.
We chose OMOP CDM. Here’s the reasoning.
The fragmentation problem demands a common language
A typical Medicare beneficiary sees seven providers a year. Each one documents in a different system. Claims data uses ICD-10 and CPT. Epic uses SNOMED internally. Labs report in LOINC. Pharmacies use NDC. A wearable device uses whatever schema its engineers invented.
None of these systems talk to each other natively. If you want a complete picture of a patient’s care, you need a common language to translate everything into.
OMOP CDM is that language. It was designed specifically to make clinical data from different sources comparable — which is exactly the problem we’re solving.
Clinical research has already validated it
The OHDSI network (Observational Health Data Sciences and Informatics) runs one of the largest clinical research consortia in the world. Hundreds of institutions have mapped their data to OMOP CDM to participate in multi-site studies. The model has been pressure-tested against real-world data from EHRs, claims systems, registries, and wearables.
When we build Glance on OMOP CDM, we inherit that validation. Every concept ID we use has been defined and reviewed by clinical informaticists. Every vocabulary mapping has been tested against real data. We’re not reinventing this — we’re building on proven infrastructure.
The Athena vocabulary is the hard part, already done
Mapping source data to standard concepts is genuinely difficult. SNOMED CT alone has over 350,000 concepts. RxNorm has over 100,000 drug entries. LOINC has over 90,000 lab codes.
OMOP ships with Athena — a pre-built vocabulary of 2.4 million concepts across all major coding standards, with curated concept relationships and mappings between them. When Glance normalizes a source value, we’re resolving against this vocabulary using deterministic rules.
That’s years of clinical informatics work we didn’t have to do, and don’t have to maintain. For a team of our size, that leverage is significant.
It makes our analytics actually portable
Because Glance stores everything in OMOP CDM, our quality measure computations, HCC coding logic, and care gap identification all run on the same standardized schema — regardless of what the source system was.
An Epic practice and a Cerner practice with Medicare DPC data can both use the same analytics. The same HEDIS measure definition runs on both. The same reference range evaluation runs on both. That’s only possible because the underlying data is in a common format.
What the tradeoffs are
OMOP CDM has real tradeoffs. It’s designed for observational research, not operational clinical systems, so some of the modeling choices are awkward for bedside use cases. The nursing documentation side of Glance required extensions to handle things like vitals flow sheets and medication administration records that don’t map neatly to standard CDM tables.
The vocabulary mapping process is also complex. Our WhiteRabbit scanner and visual ETL mapper exist precisely because the path from an arbitrary source schema to OMOP CDM is non-trivial. We’ve invested heavily in making that process fast and accurate.
But for our core value proposition — giving providers, ACOs, and individuals a complete, normalized, clinically useful picture of health data from any source — OMOP CDM is the right foundation. It’s the standard that clinical research runs on, and it’s the standard we’re building on too.
