Speaking the Same Language: Common Data Models and the Promise of Shared Evidence

Why the same question gets different answers

Ask two hospitals the same clinical question, say, how often a particular medication is followed by a particular complication, and you will often get answers that cannot be compared. Not because the medicine differs, but because each institution records its data in its own structure, with its own codes, its own conventions, its own quiet assumptions. The information is there. It simply does not speak a common language, and so the knowledge locked inside each database stays local.

This primer is about the unglamorous work that lets dispersed health data produce shared evidence: the common data model. The argument is that standardizing the structure of observational health data, rather than centralizing the data itself, is one of the most consequential and least visible advances in modern health research, and that it carries a lesson about infrastructure that reaches well beyond the clinic.

What a common data model actually does

A common data model is an agreed-upon way to organize observational health data, the records generated during routine care such as electronic health records and insurance claims, so that the same analysis can run identically across many separate databases. The most widely adopted in observational research is the OMOP Common Data Model, maintained by the open-science community known as OHDSI, the Observational Health Data Sciences and Informatics collaborative.

The model does two things at once. It standardizes structure, defining how a diagnosis, a prescription, or a procedure is represented in tables and fields. And it standardizes vocabulary, mapping the many different code systems institutions use into a shared set of concepts, so that a condition recorded one way in one hospital and another way in a second hospital resolves to the same idea. Once data is transformed into this common format, a study written once can be executed in many places, and the results can be pooled or compared because every site answered the question the same way.

Federation: keeping the data home

The feature that makes this matter for trust is federation. In a federated network, each institution keeps its own patient-level data behind its own walls. What travels is not the records but the analysis: a study runs locally at each site, and only the aggregate results leave. Nobody assembles a giant central pool of identifiable patient data.

This is a quietly profound design choice. It means a research network can generate evidence across millions of patients and dozens of institutions while every institution retains control of its own data, its own governance, and its own accountability to the people that data describes. The standardization makes collaboration possible; the federation makes it trustworthy. The two together are what let a question be asked at scale without anyone surrendering custody of what is theirs.

The evidence this unlocks

The payoff is reproducible, large-scale evidence about what actually happens in routine care, as distinct from the narrower world of the clinical trial. Observational research on standardized data can characterize how patients are treated across whole populations, estimate the effects of one treatment against another, and predict individual risk, all using analytic methods that have been validated and shared across the community rather than reinvented at each site. The OHDSI community has published hundreds of peer-reviewed studies built on this shared model, and national and multi-site networks across several countries have harmonized their clinical data into it to support both domestic and international collaborative study.

The honest qualifier matters too. Standardization is never finished. Mapping local data into a common model takes real effort, data quality varies, some clinical domains are better represented than others, and the model itself keeps evolving as researchers find what it cannot yet capture. The promise is large, and so is the maintenance it demands.

The lesson that travels

Here is why a piece about data models belongs in a health-and-wellbeing series rather than only a technical journal. The common data model is a working example of a general principle: shared knowledge becomes durable and trustworthy when the infrastructure underneath it is standardized in structure and governed at the edges. The value does not come from any single database. It comes from the agreement that lets many databases answer the same question, combined with the arrangements that let each one keep custody of what it holds.

That combination, standardization that enables collaboration and governance that preserves trust, is the pattern the Foundation works to build wherever communities and institutions need to learn from shared information without losing control of it. Observational health data is one of the clearest places to see it succeed. The same question, asked the same way, across many places that each remain accountable to their own people: that is what speaking a common language makes possible, and it is worth the unglamorous work it takes to get there.

References

1. OHDSI. Observational Health Data Sciences and Informatics. International research network with a coordinating center at Columbia University. https://www.ohdsi.org/

2. OHDSI. Standardized Data: The OMOP Common Data Model. Open community data standard for the structure and content of observational data, with standardized vocabularies. https://www.ohdsi.org/data-standardization/

3. OMOP Common Data Model documentation. CommonDataModel (OHDSI GitHub). Specification of CDM versions, tables, and conventions. https://ohdsi.github.io/CommonDataModel/

4. A scoping review of OMOP CDM adoption for cancer research using real-world data. npj Digital Medicine (2025). Federated model lets data holders retain patient-level databases; distributed network analyses across sites. https://www.nature.com/articles/s41746-025-01581-7

5. Rappoport N, et al. Kineret: Israel's Largest Hospital Network Transformed into the OMOP common data model for collaborative research. PLOS ONE (2025). Multi-center harmonization supporting intra-national and international collaboration. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12574879/

6. This Week in OHDSI / Weekly OHDSI Digest (2025). The OHDSI community has published more than 950 peer-reviewed studies using the OMOP CDM or OHDSI tools. https://www.ohdsi.org/thisweekinohdsi/