Hi
Do anyone have any experiences with OHDSI - https://www.ohdsi.org/ ?
I was asked if it has any correlation/similarities/overlap with openEHR. Reading through the web page it seems more like a dataware house solution/spec.
I found this blog with a comment from @ian.mcnicoll : http://blog.thehyve.nl/blog/omop-ohdsi-health-informatics-environment
Appreciate any feedback!
Hi!
glad to hear that someone else has similar thoughts!
The OMOP clinical data model maintained by OHDSI is a relational database model (see: https://github.com/OHDSI/CommonDataModel) and relies on a set of terminologies they call vocabulary (http://athena.ohdsi.org/search-terms/terms).
They intend that you build an ETL to store your data into their OMOP data model. To help with the ETL process they provide some tools that help specifying it. But there are also people out there that use it to persist their data for warehousing.
Intention is to build tools for research and data analysis on this model (see https://www.ohdsi.org/demos/ for some examples. Achilles is probably the most powerful one). They also provide R packages to access the data.
In my opinion it’s not the data model but the tools they built ofn top of it that make it interesting.
I didn’t find any information about someone using openEHR and OMOP together when I had d aquick search about this a couple weeks ago.
Later this year it’s definitely on my list to “play” a little with these two and see what happens…
Best regards
Niklas
When I asked about this during the last Medinfo conference, I heard that some kind of collaboration/integration/thinking was being started. Perhaps @siljelb, @heather.leslie, @ian.mcnicoll or others might know more?
I’m afraid I’m spectacularly unenlightened, but interested to learn more.
They have a free book available at https://ohdsi.github.io/TheBookOfOhdsi/
I got a hard case copy from amazon and found it’s quite interesting to read and learn a bit more - it has exercises about data analysis and how to do it.
Planning of openEHR + OHDSI in China seems to be on it’s way already if I understood things at http://openehr.org.cn/item_text.jsp?itemID=44 correctly via Google translate. @lvxd can you tell us a bit more or provide an update regarding plans/status?
Hi Erik,
Sorry for reply late due to the holiday.
OHDSI has become well known since last year, especially for medical informatics researchers. I was invited to have a talk in one of their workshop last year, and talked the possible collaboration with the guys in OHDSI China. I thought it’s a good thing for openEHR in China since the collaboration can encourage more people use openEHR. We discussed to kick off collaboration from some suitable project, the idea is that openEHR will be used to build the CDR in hospital, and openEHR2Ohdsi will be developed to make the data linked to the Ohdsi network so as to be shared by researchers. We planned to further discuss it after holiday, but it has been delayed due to the virus in China.
Did you have experience of working with Ohdsi? Could you also share them if possible? Let’s see how we can work together in this line.
Cheers,
Xudong
Erik Sundvall via openEHR discourse@openehr.org 于2020年2月5日周三 下午4:56写道：
Great Xudong
I think it is reasonable to have som dataanalytics together with openehr. If ohtdsi works well it would be interesting to share some patterns on how to integrate it.
I only know it by name myself.
Yes, that’s exactly what I thought. I will keep you updated about this collaboration
Bjørn Næss via openEHR discourse@openehr.org 于2020年2月12日周三 上午12:27写道：
Thanks for the interesting response!
No, I have not worked with OHDSI but got very interested after hearing Georg Hripcsak’s keynote at the Medinfo 2019 conference. I would guess that a lot of “openEHR2OHDSI” integration can be done by designing good AQL-queries combined with some transformation/translation/formatting code. Probably some kind of two-way flow where parameters from OHDSI requests are used in the AQL etc. You have probably looked deeper into this than me, what do you think will be needed?
Right now we (Region Östergötland) are busy doing a proper openEHR procurement (see Swedish openEHR procurements & RFIs) and will then install and implement that in suitable parts of our healthcare organization and tackle som prioritized development needs. So we’ll likely not be doing any serious “openEHR2OHDSI” until at earlieast Q4 2020, more likely some time in 2021. But at a “discussion” level we’d be interested in looking at this, so please keep this thread updated with thoughts and findings if possible.
George Hripcsak, gave me permission to share the Medinfo presentation here.
EDIT: But the upload did not work, retrying…
Please try again 
I was lucky enough to watch this great presentation in Lyon, too. You can find the slides here.
In fact, OHDSI and openEHR look like a perfect couple. As a clinician, I can’t be of much help concerning implementation. But @backbord 's explanation seems to point in the right way.
I’ll keep a watchful eye on this topic, and share with you whatever I find concerning it.
We are currently on the process of certification in OHDSI.
We are also updating our tooling to fully support OMOP CDM 
First tentative draft of OMOP CDM as GENERIC_ENTRY classes (I have generated all 37 classes, I attach the 3 more common ones to start a discussion)
openEHR-EHR-GENERIC_ENTRY.visit_detail.v0.adl (13.2 KB)
openEHR-EHR-GENERIC_ENTRY.observation_period.v0.adl (4.3 KB)
openEHR-EHR-GENERIC_ENTRY.person.v0.adl (13.6 KB)
Also very interested in this topic.
I am a database solution engineer from PostgreSQL and Greenplum Community China with more than 10 years big data experience in financial, aviation and IoT industries .
I have some experiences in IoT industry platform for vast data integration and sharing among many vendors.
According to my observation, a uniformed standard will help blooming the adoption while too many standards will be a barrier for adoption, cause too many standards equal no standard.
Hello!
What are the use cases when research projects would benefit most of OMOP? In Finland OMOP is used in 2 University Hospitals, and we also have national openEHR based data exchange platform. The hospitals have however mapped legacy data to OMOP, but it does not seem very efficient to do all these mappings, but rather use openEHR-OMOP mapping if it existed.
Yes, in my opinion, a full openEHR->OMOP transformation could be achieved by annotating the archetypes/templates with the equivalent OMOP class . By getting fields from the openEHR RM it should be possible to get the minimum data. If we have other “label” terminology bindings to snomed or loinc we could do this even without the annotation to the OMOP class.
I also think that other approaches, like the one @ian.mcnicoll did for the Lab results, are more directly translated into OMOP table rows, but both are possible in my opinion.
And as for use cases, I see this as a minimum common model for research, but obviously if every place in the research network ‘understands’ openEHR (i.e. can execute AQL) you have more fine grained information at your disposal.
There might be an opportunity to get some support for this work from a university Comp. sci dept - would anyone be interested in helping me (this all pro-bono!!).
Sure, I wanted to even implement some things for the transformation