OHDSI Home | Forums | Wiki | Github

Incorporating of NEST data into Common Data Model Harmonization project

At the Sentinel Workshop, Greg Pappas of CDRH had discussed the potentiality of integrating NEST data with Sentinel data. Given FDA’s recent establishment of the NEST coordinating center (source), I was wondering if there have been, are, or will be efforts to integrate data on medical devices with Sentinel data? Currently, the Common Data Harmonization Project is well-underway at the FDA to generate a common data model from the Sentinel, PCORnet, OMOP, and i2b2 CDMs. I wonder if NEST CDM (if it has one?) will eventually be a part of the grand-daddy-of-all CDMs, or can the data be incorporated via the OMOP CDm? The OMOP CDM v5.3 already has a unique_device_ID under the device_exposure domain.

@lilipeng Do you have any information on the CDM Harmonization project? Do you know who is leading it, or participating in it? I can’t find a link to it anywhere. We have developed our own harmonized data model and I am interested in looking at what others are doing.

Hi @Mark_Danese Mitra Rocca at the FDA is leading it with contractors providing IT support. Her contact information can be found via the FDA directory. There isn’t a direct link to it, but there is a presentation on it here. What data model have you harmonized?

OK. Now I get it. I did see Mitra’s presentation. We developed what we call a “generalized data model” (GDM). It can be easily transformed into Sentinel or OMOP. So, in other words, the harmonization comes BEFORE putting the data into OMOP or Sentinel. Then it is a short jump from GDM to those data models. Based on our research, harmonizing data after they have been put into these other data models has important limitations.

The whole reason we did this was to develop a data model for ETL into OMOP. Then we realized that, if we did it the right way, we could have a data model that 1) works fine as it is for our purposes, 2) can be transformed into other data models to take advantage of all of the analytical downstream benefits of those ecosystems. We work with both OMOP and Sentinel data, and this is a much easier way to work (for us). Others may, of course, disagree, or have different ideas.

Preprint available here: https://www.biorxiv.org/content/early/2017/11/10/194597

I know they are harmonizing the 4 CDMs to the BRIDG model: https://www.cdisc.org/standards/domain-information-module/bridg

So it seems like the harmonization method has already been decided.

Has the GDM been tested and adopted by any commercial healthcare data providers looking to convert their data model to the OMOP format?

1 Like

GDM has not been used by anyone but ourselves since it is so new. We are still awaiting reviewer comments on the manuscript.

BRIDG is interesting. Thanks for pointing me to that.

The goal of this project is to enable querying any of the 4 CDMs (Sentinel, PCORnet, OMOP and i2b2/ACT) leveraging an intermediary/über model and generate HL7 FHIR or CDISC standards (SDTM) as the output. This is a FDA-led initiative in collaboration with 3 Institutes at NIH as well as HHS/ONC and is one of the projects within the portfolio of the HHS/Assistant Secretary for Planning and Evaluation (ASPE) Patient-Centered Outcomes Research Trust Fund (PCORTF) program:


BRIDG is a Domain Analysis Model (DAM) developed by FDA, NCI and is an ISO, HL7 and CDISC standard. BRIDG is implemented by a few biopharmaceutical companies, academia, FDA and NCI as the underlying model for Metadata Registry (MDR), rare disease registries, by a large CRO as the basis for a central canonical data model (CDM) to increase semantic consistency and quality across a diverse portfolio of products and services.
For more information about BRIDG, it would be best to visit HL7’s Web site:
http://www.hl7.org/implement/standards/product_brief.cfm?product_id=71
.

t