Observational Health Data Sciences and Informatics

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
Proposing person: Vojtech Huser

**Title: Deeper semantic integration**

discussion: http://forums.ohdsi.org/t/standardizing-weight-data/2921

**Use Case**

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. (LOINC term for weight and SNOMED term for weight)

OMOP CDM, compared with Sentinel CDM does not enforce strongly only one way to represent patient weight (for example). Similarly, units for a given concept are not standardized, yet standardizing that at analysis time may be sub-optimal. Sentinel CDM has expected standard units defined for a subset of lab resuls. (feature requested by researchers) 


**Additions**

This proposal argues for CDM implementation guide document that would put additional constraints on CDM data and maintain a formal specification for such concept level standardization for a selected high-stakes subset of concepts.

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


In such implementation guide (IG), deeper semantic standardization specs would be documented. (e.g., patient weight is expected as this concept, and expected in this units). Such IG could allow 2 (or multiple) units in some instances.

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. (we all know about some NASA famous events around that)

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 for LDL cholesterol (plus desired standard units).