OHDSI Home | Forums | Wiki | Github

Planning to Add N3C Concept Sets to OHDSI Phenotype Library

The N3C Concept Set Browser and Builder Tool is used by researchers within the National Covid Cohort Collaborative (N3C) to create OMOP Concept Sets. While N3C policy states that researchers are welcome to share “knowledge artifacts” outside of the N3C community, these concept sets are currently stored in N3C’s Palantir Foundry application which is only accessible to N3C researchers.

I’m hoping to add many N3C Concept Sets to the OHDSI Phenotype Library so that my research team at Tufts can easily import them to our Atlas instance and so that they are available to the whole OHDSI community. Ideally, I will create a simple pipeline to extract concept sets as they are created and vetted by the N3C and load them to the Phenotype Library.

I’ve seen the general guidance on contributing to the Phenotype Library (here and here), but I am wondering if there are any special considerations for this use case in which concept sets may already be vetted and have a DOI?

Any thoughts or advice greatly appreciated!

@Gowtham_Rao

I don’t have any guidance to give, but I want to say Colorado would love to support our researchers with well vetted concept sets. @mgkahn and I were just talking about this yesterday! Please keep us in the loop.

1 Like

Hi @kzollove . Sure. Let’s talk and model a workflow. Can you share some examples of the concept sets to be contributed and relevant metadata

@Gowtham_Rao Thanks for the quick response.

Concept sets are notably versioned and linked to “parents” (concept set has versions, a codeset is a specific version), have status (finished, under construction), linked to an informatician and SME, and has a zenodo DOI.

Also, concept sets are linked to Reviews:

Unfortunately I can’t share links directly as N3C requires authentication. Here is some information on an example concept set (Pulmonary embolism):

Concepts (48 OMOP condition concepts):
Export.xlsx (6.4 KB)

Zenodo DOI (https://doi.org/10.5281/zenodo.7705470)

And some metadata for that concept set:


Data and metadata are stored in a filesystem in the Palantir Foundry apps:
Tables in N3C (Concept Set Ontology > hubble_base):

  • concept_set_members: all codesets (concept_set + version) and all concepts. The main data table
  • code_sets: Links codesets to projects, creators, createdAt, etc. Codeset metadata
  • zenodo_doi: Links codesets to a DOI and zenodo publication. This is potential route for citation/attribution

and probably other ones I am missing.

There is a REST API that backs N3C’s Concept Set browser, but it is still unclear to me if it is accessible from outside of the Palantir Foundry infrastructure.

I would like to ask that we clarify what the unit of contribution is.

Is the proposed contribution a concept set or is it concept set expression? Concept set can be thought of as an array of omop concept ids. Concept set expression is an expression that when resolved gives an array of omop concept Ids. This expression is defined by ohdsi circe specifications.

From an application perspective, ohdsi software (eg Atlas or HADES) uses concept set expression as an input.

it can be combination of both concept set or concept set expression. We have found some cases were the concept set express is used and the expression is expanded to include array of OMOP concept ids, it will result, in some case, include certain item we do not wish to include. And for those cases we will include the concept set item in order to exclude.

Please keep me posted. We would be happy to contribute N3C recommended concept set to the Phenotype Library. Wondering how best to contribute if the phenotype was not generated from the Atlas-Phenotype. Thank you.

1 Like

Thanks @stephanieshong . Let’s meet and have a work session together

Good day - We are looking forward to discussing this topic today at 11am EST. True to OHDSI spirit of open collaboration, all are welcome. The meeting is hosted by the OHDSI Phenotype Development and Evaluation workgroup. The meeting details are here

Please see recording . Post to follow.
https://ohdsiorg.sharepoint.com/sites/Workgroup-PhenotypeDevelopmentandEvaluation/Shared%20Documents/General/Recordings/N3C%20and%20OHDSI%20Phenotype%20Library-20230913_110714-Meeting%20Recording.mp4?web=1

t