Stream: omop + fhir
Topic: OMOP <-> FHIR mappings
Ward Weistra (Jun 14 2021 at 12:11):
Following up from my earlier summary on Zulip:
Ward Weistra said:
As noticed in the #FAIR stream too, there was a partnership between the OHDSI (OMOP) community and HL7 International announced: https://www.ohdsi.org/ohdsi-hl7-collaboration/
- I only see Charles Jaffe and George Hripcsak in that article. Does anyone know who to reach out to for more information?
- From the few words in the announcement it seems like a joint data model between FHIR and OMOP will be developed. Anyone details on the approach?
Over the years there have been MANY mappings between the two models, but, as indicated by the new ones being developed, they have reusable to a very limited extend:
- All the ones listed here: https://www.ohdsi.org/web/wiki/doku.php?id=projects:workgroups:mappings_between_ohdsi_cdm_and_fhir, including:
- Georgia Tech's http://omoponfhir.org/
- CDMH project: http://build.fhir.org/ig/HL7/cdmh/profiles.html#omop-to-fhir-mappings
- MIRACUM: https://www.devdays.com/wp-content/uploads/2019/12/Sohrab-Hejazi-Map-FHIR-to-OMOP-_-DevDays-2019-Amsterdam.pdf
- Google: https://github.com/GoogleCloudPlatform/healthcare-data-harmonization/tree/master/mapping_configs/fhir_omop
I think one of core problems is that when mapping between two specific data models you lose some information. Depending on mapping OMOP to what FHIR profiles (core resources with or without custom extensions, US core, anything else) you need to choose how to handle that. So, any mapping implementation from OMOP to one particular set of FHIR resources will not easily be reusable by another.
Wouldn't it be more efficient in the long run if we'd model the OMOP data model in FHIR logical models and create one custom set of scripts that transforms the data between those.
Then every next project would only need to map between FHIR logical model conformant data to their profiles of choice, for example with the Mapping Language.
Maxim Moinat (who promised to create a Zulip account) pointed out yet another mapping between FHIR (US Core) and the OMOP CDM: https://github.com/OHDSI/FhirToCdm
Ward Weistra (Jun 14 2021 at 14:23):
May Terry said:
Just for historical context, there was a rather thin IG back in 2018 to map from FHIR to OMOP through the Common Data Model Harmonization (CDMH) project. https://build.fhir.org/ig/HL7/cdmh/profiles.html#omop-to-fhir-mappings. Unfortunately, it stopped in 2018 and is now outdated.
Mareike Przysucha (Jun 16 2021 at 11:55):
Hello everyone. We are currently starting a project which contains at a later stage a OMOP <-> FHIR Mapping on wound data. The project is called ZIEL (https://www.hs-osnabrueck.de/ziel/). At the moment we are currently exploring the data we have and can map, but soon we will come to OMOP and FHIR. As we are just starting, there hardly is any resources to share yet, but we will share once we have anything to share...
Maxim Moinat (Jun 17 2021 at 12:47):
Hi all, I have just joined this omop + fhir thread. Thanks for the invite! As an OMOP/OHDSI expert I have for a long time following progress in FHIR, but since a few weeks actually digging into FHIR data. The profiles we are using are created by the Dutch 'Medical Intelligence' initiative. And we are doing a conversion of flattened FHIR tables to OMOP (yes you read that correctly, we have flat FHIR tables).
Grahame Grieve (Jun 17 2021 at 13:21):
we have flat FHIR tables
That's not a particularly new idea, but nor is it very specific - can you explain more?
Maxim Moinat (Jun 24 2021 at 10:38):
The nested elements are unpacked to column names like subject.identifier.code. In case of a one-to-many relation, like indications of a medication, these are stored in a separate table. I can imagine this is indeed nothing new, but ideally we would like to work with FHIR resources i their native json/xml form.
Grahame Grieve (Jun 24 2021 at 11:04):
well, you might follow the discussion on #analytics on FHIR, though it's slowed at the moment.
Christian Reich (Jul 29 2021 at 00:14):
@Ward Weistra et al.:
Ward Weistra said:
As noticed in the #FAIR stream too, there was a partnership between the OHDSI (OMOP) community and HL7 International announced: https://www.ohdsi.org/ohdsi-hl7-collaboration/
- I only see Charles Jaffe and George Hripcsak in that article. Does anyone know who to reach out to for more information?
We are finally kicking it off. Please see our announcements: https://chat.fhir.org/#narrow/stream/286658-omop-.2B.20fhir/topic/Updates.20on.20FHIR.20.2B.20OMOP.20collaboration/near/247395986 and https://forums.ohdsi.org/t/fhir-omop-collaboration-kick-off/14551. The point is to create awareness in both communities, and help understand how the other community works.
However, we are looking for use cases we want to cover beyond the mapping or conversion. Anything that comes to mind?
May Terry (Jul 29 2021 at 02:13):
@Christian Reich - I'd like to explore an oncology use case that aligns FHIR-based mCODE to OMOP Oncology for observational research. A secondary use case would explore feasibility of mCODE using the OMOP genomic extension and OMOP variant concepts rolled up to cancer types.
Christian Reich said:
_However, we are looking for use cases we want to cover beyond the mapping or conversion. Anything that comes to mind?_
Last updated: Apr 12 2022 at 19:14 UTC
