OHDSI Home | Forums | Wiki | Github

Icd11 / snomed


Like many, we want to promote OMOP as a model free of rights and licenses.
The use of SNOMED-CT requires a license.
France has not yet paid for a license and does not plan to do so in the coming months.
We would like to evaluate ICD-11 as an alternative to SNOMED-CT. What do you think of that? Is there a working group on this subject?

(Dmytry Dymshyts) #2

Hi @aparrot89
ICD11 as well as ICD10 has monoaxial hierarchy:
for example Diabetic retinopathy,
ICD11 has parent - “Retinopathy” and grandparent “Disorders of the retina”
see here https://icd.who.int/browse11/l-m/en#/http%3A%2F%2Fid.who.int%2Ficd%2Fentity%2F1006882070
Note, that ICD10 parent of Diabetic retinopathy is diabetes mellitus.
SNOMED concept has both parents -
Such a rich SNOMED hierarchy easies building of concepts sets and cohorts.
So, there’s no workgroup, everybody kinda agree that SNOMED is good standard for Condition.
@Christian_Reich can you comment about the licencing?

(Vojtech Huser) #3

@aparrot89 (not sure what your real name is), can you post a script that retrieves the foundation layer of ICD11? This would be crucial for the purpose you propose.

How would you go about getting a CSV file (required for import into OMOP Vocab) for ICD11. I think Dmytry has in mind the ICD11-MMS. That is just one view of the ICD11 knowledge base. https://en.wikipedia.org/wiki/International_Statistical_Classification_of_Diseases_and_Related_Health_Problems#ICD-11_Linearization_for_Morbidity_and_Mortality_Statistics_(ICD-11_MMS)

But ICD11 is MUCH more than just the MMS view that people are used to from ICD10.


@Dymshyts. Thanks for you answer.
I think I agree with @Vojtech_Huser. I think that ICD11 is more than ICD11-MMS Linearization.
An example, if you see in foundation system, I think that VIH is multiaxial because parents are :

And in the MMS-Linearization VIH is only a Certain infectious or parasitic diseases
But ICD11 terminology is in process and like for the OMOP community it is possible to do proposal in contributions sections (need to be logged). I think we should do one for diabetic retinopathy.
They are post coordination as well. Translation is free.
What do you think?

(Christian Reich) #5


Couple things:

  • The license problem is significant. We are trying to work with snomed.org on a solution for OHDSI.
  • I certainly hope ICD11 is overcoming the limitation of the previous ICDs, which really was linear and constituted a statistical classification, allowing only single parenthood. We should definitely take a look and define how we are going to utilize it. We currently also use MedDRA as an additional hierarchical system for Conditions together with SNOMED, so there is a precedent.
  • Ripping out SNOMED and replacing it with ICD11 is akin to an operation on the open heart - everything right now banks on SNOMED-based Conditions and their description logic. We cannot do that just like that.

At any rate, whatever we want to do we would need to create a problem statement and a proposal in the CDM issue list. Can we do that?

  • License : In Paris hospital we want to use OMOP for all our processes. It will be a bedside patient model. So we need the same terminologies for research and care (I think that is not a good choice to have 2 terminologies, once for the care, once for the research). Will the solution between snomed.org and OHDSI address this issue?
  • It really understand that replacing SNOMED with ICD11 will be difficult. But this is the price to be free. And we are speaking about medecine. Open heart surgeries are done every day! ^^
  • Apparently more than 30% of ICD11 foundation concepts are linked to SNOMED-CT (by an ICD10 code or by exact matching string)
    Ok for adding a proposal in the CDM issue list.

(Dmytry Dymshyts) #7

For sure, we can create this initiative on ICD11 imporovement. But first we need to decide wether ICD11 is useful for the community.

And I like the post-coordination.

So, in the patient data you’ll have this complex code.
But, to fit this into the current OMOP vocabulary model we need to build all possible post-coordination combinations, which is kinda cumbersome. We need to think through though.

BTW, when does ICD11 becomes active?


SNOMED CT supports post - coordination. How the OMOP community handles that? Is that difficult?

(Dmytry Dymshyts) #9

Yes, SNOMED does.
what we have in a CDM -
CONDITION_OCCURRENCE.condition_concept_id, which corresponds to exact concept in the vocabulary.
To allow post-coordination we need to create a meta-table, which will contain the combinations.

Something like that:
postcoordination_concept_id | relationship_id | attribute_id
2D1Z&XB32&XC56 | Is a | 2D1Z
2D1Z&XB32&XC56 | Has laterality | XB32
2D1Z&XB32&XC56 | Has Specific anatomy | XC56

This is a big change of the model. I’m pretty sceptical about it (at least for now).

(Christian Reich) #10


As @Dymshyts tried to say, we don’t have post-coordination. Only pre-coordination. We even pre-coordinate ourselves, like in the ICDO cancer diagnosis case. And we are thinking of doing that to be able to map complicated pre-coordinated (AND or OR) source codes, creating SNOMED Extensions.

Why would you need post-coordination? What is the use case?

(Dmytry Dymshyts) #11

I just looked at ICD11, and they have post-coordination as a way of condition coding.
So, I anticipate this in a patient data, that’s why we are talking about that.