OHDSI Home | Forums | Wiki | Github

Clinical trial data conventions for OMOP CDM

Happy Friday, everyone!

We the Clinical Trials working group have submitted our proposals for OMOP CDM and Standardized Vocabularies conventions to capture clinical trial specific data.

See this link - https://github.com/OHDSI/CommonDataModel/issues/358

I know this is summer holidays time, and there’s coronavirus and the prep for the US Symposium, but if you are in the Clinical Trials world, do take a look at our proposals, and give us a thumbs up or raise questions.



The working group focused on transformation of clinical data in SDTM format to the OMOP CDM. A lot of clinical data are in SDTM format because of the requirements from regulatory agencies. Making it easier to map SDTM data to the OMOP CDM potentially opens up a large cache of clinical data for collaborative data science research. Hope to get the community’s feedback on this document so that it is broadly applicable for getting the most use out of clinical trial data.



Thanks for the proposal. @clairblacketer and the CDM group is a little under water with the documentation revision before the Symposium. But: As with all proposals if there is no controversy you can start implementing no problem, and when the time comes that we ratify it you will already be there and collected some useful experience with real data.

Thanks, @Christian_Reich. Glad to hear it! We had our first workshop today to kick off our first implementation :slight_smile:

This is a great initiative.

On related note, here is one way to represent study visits (next to EHR visits)
When data has imported EHR data combined with research visits


1 Like


I don’t see the “clinical study visit” in the new Type Concept vocabulary. Should we add in a type concept_id for the clinical study data? As @Vojtech_Huser mentions above, it is good to separate the clinical study data from data of other origins.

1 Like

There is Case Report Form.

Is “Clinical Study” above the same as “Clinical Trial Study”? Not sure but to me they are two different things. Also, the “Clinical Study visit” concept is marked as non-standard.

Also, isn’t there a a draft proposal on new visit types to represent clinical trial data in the link Sonia shared above?


It’s not that simple. A visit in clinical trial is the planned interaction between patient as defined in the protocol. OMOP Visits are configurations of healthcare delivery (who comes to whom, is there a bed, what kind of service level). So, they are orthogonal. For example, you can have inpatient and outpatient visits in OMOP, but how would that square with the clincial trial visits? You would need inpatient or outpatient screening visits, inpatient or outpatient follow-up visits, etc. To build these clinical trial visits into the OMOP hierarchy you would have to create children for each (or many) OMOP Visit with one of the clinical trial visit. It would create lots of visits: OMOP Visits * trial visits. Probably an overkill.

I would recommend creating Observation concepts and record them at the visit_start_date, essentially observing “this is a xyz trial visit”.

what happened with that statement? :slight_smile:

Good points, Christian - very good points. You are correct that nearly all (or all) CT visits would also into the classic visit type classifications. How would you then suggest we do that without losing a critical the CT semantic (e.g. screening, follow up, treatment, unscheduled) while avoiding (as you correctly suggested above) doing a spaghetti hierarchy?

As I said. :slight_smile:

You don’t need observation. In this case, VisitTypeCID and VisitCID duplicate now the OP vs IP information somewhat. So you can just use ClinicalStudy value for VisitTypeCID and still have your full hierarchy (for study or for RWD) inside your VisitCID. ( Sorry for using CID and non standard column names; just for the sake of rebellion.) But I agree that the observation trick works quite well and I like it.