OHDSI Home | Forums | Wiki | Github

EHR data to OMOP CDM Work Group

workinggroup
themis

(Sam Martin) #182

@MPhilofsky included my scoring and resisted the urge to add new topics… That said, there is a growing question (tension) that I’m trying to address and I cannot help but believe it has been encountered, and hopefully addressed, elsewhere.

Topic:
How do you effectively offer OMOP to users who’s perspective or analysis is represented by a world bound by a terminology such as ICD-10.
Example: Their population definitions, metrics, quality checks… are all rooted in a single terminology; so when presented with OMOP data, they say “can you show this as ICD-10?” Yet, they want to use OMOP because of its ability to deliver more expansive datasets.

I can understand how they would ask that, but it seems to bypass (or miss) the value inherent to a well established CDM like OMOP.

  • Who else has encountered this?
  • How have others addressed it?
  • Is there another thread/group I should be asking this?

Thanks!


(Christian Reich) #183

Answer No 1: There is no such a thing as ICD-10. Well, there is, except there are many. Every country has their own. The American version, ICD-10-CM, is the largest with over 100k codes. The original WHO ICD-10 has less than 20k. So, obviously it’s not the same. But even the codes that overlap are not always the same between the versions.

Answer No 2: ICD-10 is suboptimal for research, because it has a lousy one-parent only hierarchy. SNOMED has a way stronger hierarchical organizations, plus other links connecting the concepts.

Answer No 3: There are conditions in other terminologies. For example, HCPCS M1062 “Patient immunocompromised” is a Condition not stored in ICD-10. Internationally you have plenty of those.

Happens often with people who have used their definitions for a long time and don’t want to change their ways. I have empathy.

Just map them using the mapping tables provided by OHDSI. We have all ICD-10s from most versions mapped over.


(Melanie Philofsky) #184

If I had to guess, I’m sure all US EHR data holders have encountered this resistance from within their organization at some point in time. @cukarthik and friends at Columbia have been using OMOP for a long time. He might be able to provide more insight.

Education, education and then more education :slight_smile: The power and beauty of OMOP is in the standardized data model and the terminology harmonization. The one and only one standard concept_id to represent an idea, the hierarchies, the classification concepts, domains to identify the table where the data lives, and the ability to represent your source data ‘as is’ with a 2 billionaire source concept_id make the OMOP CDM amazing!
*The ability to collaborate with like minded people all over the world is implied :slight_smile:

I haven’t delved into Conditions, but I have started an educational series for Drugs. Similar to ICD10CM codes, folks love the NDC codes, but they aren’t standard! The series includes the OMOP basics of mapping source codes to concept_ids, CDM layout, OMOP Vocabulary tables, etc.; a library of SQL queries including how to obtain all drugs with a specified ingredient or all drugs under an ATC classification concept; a video on how to use Athena to identify drugs, ingredients or classifications of interest; a deep dive on our source data and how it fits into the CDM; future plans include Atlas, DQD, and other OHDSI tools. And it is all catered to our local users. Colorado is still in the implementation phase of this learning initiative, so the results aren’t final. RxNorm and Drugs have been well received. Say goodby to the huge time waste of string text searches on multiple source data tables!

Thank you for not adding more topics, @samart3! The EHR WG does have meetings without an agenda, so folks can bring any idea/concerns/questions to the WG for discussion. And the forums are the best place to discuss topics that involve more than EHR specific data.


(Don Torok) #185

FYI to @MPhilofsky reference of a library of SQL queries. The link to an online version is https://github.com/OHDSI/QueryLibrary. Or you can install R package locally from https://github.com/OHDSI/QueryLibrary.


(JD Liddil) #186

In our experience getting folks to change things regarding their EHR reminds me of the Charleston Heston “my cold dead hands”. Even the place we are partially owned by has told us to take a walk.


(Sam Martin) #187

@Christian_Reich thank you for the answer and also thank you for calling me out on my use of “ICD-10.” I was attempting to be generic, but should have been more specific (ICD10CM).

I think much of it is exactly as you said, they are using familiar population definitions (by habit or mandate) and it has limits their perspective.

@MPhilofsky your perspective was also very helpful.

This just increases the importance for us to implement the OHDSI tools for consuming OMOP and building cohorts. Given that our All of US is the primary consumer of our OMOP data, I’ve not needed to focus on socializing/training on end user access to Atlas, Athena and other tools. Without those it will be difficult for our teams to make the leap to OMOP as their native frame for defining populations.


t