OHDSI Home | Forums | Wiki | Github

Question on MedDRA SMQ entries in concept_relationship

I tried to locate the code related to building the concept_relationship table for SMQs, but could not see it on
github.com:OHDSI/Vocabulary-v5.0. I am trying to understand why some SMQ entries have ‘Subsumes’ and ‘Is a’ relationship_ids, while others do not.

(Ignoring ‘MedDRA - SMQ’ and ‘SMQ - MedDRA’ relationship_ids)

EG:
38000004	Cardiac failure (narrow):  has neither Subsumes or Is a
38000005	Hepatic disorders : has Subsumes only
38000008	Liver related investigations, signs and symptoms (narrow): has both Subsumes and Is a.

Thanks,
Don

@donohara:

SMQs are stale. We actually want to retire them. Do you have a use case?

Hi @Christian_Reich.

By “stale” do you mean the version of SMQs in the v5 Standard Vocabulary I downloaded in March 2016 does not include the latest SMQs, or do you mean they are passing out of favor in the community?

As far as I know, SMQs (along with MedDRA IMEs) as part of our client’s drug safety analysis processes, and MedDRA is still the primary way to classify adverse events within the FAERs database.

On the MedDRA website,
(http://www.meddra.org/sites/default/files/page/documents/list_of_smq_topics_for_website_march_2016.pdf) the “List of SMQ Topics for Development by CIOMS Working Group for SMQs”
shows 6 SMQs in development/testing.

It seems unusual that of 319 SMQ concepts, 125 have neither Subsumes or Is a relationship_ids.

@donohara:

Correct. SMQ was added way back, and nobody really ever used them. But more importantly: We just created a simple hierarchical relationship between SMQs and MedDRA codes, but that is not correct. There are mini-Algorithms, and some of the MedDRA might even be negated in those.

If you feel we need them and there is a good use case, we’d have to bring it up in the CDM Working Group and discuss how they would be implemented. Very happy to take that on.

@Christian_Reich Do you know if there are any other vocabularies in the v5 release that may have issues similar to SMQs, i.e. being “stale” or incomplete?

You can see when they were updated in the Athena website. If it’s empty, they haven’t been in a long time. For many of them it is irrelevant, becasue they are more administrative (like Type Concepts), but for those that have routine updates you want to see a date in there.

We will go through them in the CDM Working Group and clean it up.

@Christian_Reich
Great! I plan on joining the CDM WG. As my father used to say, if I asked “why”, he said “you just volunteered!”. I’d like to help where I can on these types of issues.

@donohara:

What’s your email address? You keep incognito around here.

don.ohara@evidera.com I just updated my OHDSI profile yesterday.

@donohara :

What are we going to do with your proposal? The problem is this:

  1. Disease definitions like SMQs should really part of a exercise like the one we are planning during the OHDS collaborator face-to-face meeting. In it, we want to create a library of cross-database definitions of diseases. SMQs could be either a good starting point, or even whole chapter of its own.
  2. As a vocabulary classification system it really doesn’t work: Reason is that we cannot properly represent the algorithms. Take “Acute pancreatitis” (concept_ids 38004571 (broad) and 38000022 (narrow)): The algorithm is defined as follows (from the SMQ Introductory Guide V12.0):

2.3.3 Algorithm
To apply the algorithm for this SMQ, a report is considered a relevant case for further review if

  • it includes a term from Category A
    OR
  • it includes at least one term from Category B (the list of laboratory values) and at least one term from Category C (the list of signs and symptoms)

Our hierarchy cannot do AND between different descendants, only OR.

Any good ideas what to do?

@Christian_Reich
Hi Christian - I like the idea of exploring the options during the face-to-face meeting. The initial use case was “The OMOP standard vocabulary contains MedDRA; MedDRA provides SMQs; OMOP Vocab should also contain SMQs”. Enumerating the issues as you have here, and working with the community to identify alternatives such as the library of definitions, is a great way to move forward.

t