Observational Health Data Sciences and Informatics

This is an old revision of the document!

---

Proposing person: Vojtech Huser

OMOP CDM via vocabulary axiom of one concept per entity enforces semantic standardization. Only one standard concept per entity.

Yet in some instances, the vocabulary violates that axiom.

OMOP CDM, compared with Sentinel CDM does not enforce strongly only one way to represent patient weight (for example). Similarly, unites for a given concept are not standardized, yet standardizing that at analysis time may be sub-optimal.

This proposal argues for CDM implementation guide document that would put additional constraints on CDM data and maintain a formal specification for that.

This is on different level than existing syntax standardization. http://www.ohdsi.org/web/wiki/doku.php?id=documentation:cdm:single-page

In such implementation guide, deeper semantic standardization specs would be documented. (e.g., patient weight is expected as this concept, and expected in this units).

A site could have by syntax a valid CDM dataset, yet analyses would not produce the same result.

For example, CDM site that has non-standard units of measures for weight.

This proposal does not argue to put all possible standardization to ETL. Certain logic will always remain to be done at analysis time (by analytical code), however, there are multiple CDM desired features that do not fit existing CDM spec document that is on table (domain level) and not on concept level. (e.g., measurement concept)