OHDSI Home | Forums | Wiki | Github

What Is the Status of Interoperating between OMOP and SENTINEL?

Hi all,

This is a vague question spawning out of curiosity but what is the current status of creating studies that can be simultaneously ran across OMOP and SENTINEL Common Data Models?
Has there been any research done on interoperating between these models?
I am looking further into methods for standardization and was wondering if any standards had been investigated to interoperate between these models.


~ tcp :deciduous_tree:

@TheCedarPrince (or is the right address “Your Highness”? :slight_smile: )

Good question. Conceptionally, the models are similar. They both are closed-world systems (fully normalized with all possible values predefined and absence of negative values, except in lab tests), which allows to calculate rates of things buy dividing the number of records meeting some set of criteria by the number of all records they are taken from (the entire database, or another, wider population also defined by inclusion criteria).

Obviously, I believe OMOP is the much cleaner, expandable and computationally more favorable solution, but still, both function in similar ways.

The problem is that despite the same principles, they are too different in their detail. No table and field name or definition matches, and the use of predefined codes or concepts to carry the content are different. Which means, all inclusion criteria for cohort definitions and all methods are rendered incompatible.

The FDA, NLM, NCI, NCATS and ONC put together the CDM Harmonization project to create a way to overcome that hurdle, either by harmonizing the data or harmonizing the queries. It made some progress on a small set of use cases. But the future and prospects of such an endeavor remains unclear.


I think an additional question would be OMOP and PCORnet. PCORnet was largely built off of the Sentinel CDM. However there are many PCORnet sites that ETL into OMOP and then ETL from OMOP into PCORnet. There has probably been more thought into interoperating between OMOP and PCORnet then the OMOP Sentinel interoperation. Even PCORnet and Sentinel are not interoperable. For example, PCORnet has a placeflag logic of EI (emergency department to inpatient) to which the Sentinel CDM does not.

I love standards and think you can actually learn a lot about your data going through any ETL process but alas we should not spend all of our time ETL’ing.


Haha, @Christian_Reich , no need for the formalities here! :wink:

Thanks for that link!
I had found that link earlier but was curious if there was anything more recent as their last update was from 2020.
I’ll double back and check out that final report.
Does our CDM workgroup talk about interoperating with different CDMs at all?

That is an interesting bit of information @Kevin_Haynes and I agree on what you say

Standards are where the magic is at but then actually getting away from standards to do fundamental research is critical.
It’s a fine line to walk.