OMOP Domain for Assessment Scales (LOINC Survey Class, SNOMED Staging and Scales)

Vocabulary Users
1

@Christian_Reich, @MPhilofsky -

Currently any assessment scale that is encoded in source data using a LOINC code will go to the Measurement table, and any assessment encoded using a SNOMED concept id will go to the Observation table.

Would it be possible to select a domain for all assessments, and assign concepts from both LOINC and SNOMED to the same domain?

Best,

Piper

1 Like
2

Correction -

It’s the opposite.

LOINC survey concepts are currently in the Observation domain and SNOMED staging and scale concepts in the Measurement domain.

3

Ouch. Thanks for catching that.

Can you drop examples?

4

You bet.

A secondary issue is that the concepts for the assessments themselves are both flagged as standard when the same assessment is in both SNOMED and LOINC.

Here are a few examples of the assessment scale/ instrument concepts:

image
image1359Ă—314 23 KB

Here are a few examples of the score concepts:

image
image1409Ă—299 22.2 KB

Thanks for looking at this @Christian_Reich

5

Hi @Christian_Reich -

Where can I learn more about the PPI vocabulary - which organization curates it, etc?

Piper

6

That would be our wiki documentation, section Vocabularies. Specifically here: Vocab. PPI · OHDSI/Vocabulary-v5.0 Wiki · GitHub

Now, surveys, scales, etc. are a mess. Survey subgroup is working on sustainable solutions, so we probably don’t want to introduce anything semi-baked before we have a model (e.g., should it be measurement or observation? standard or non-standard? if non-standard what maps to what?..).

7

Thanks @aostropolets.
This is helpful.

Who do you recommend reaching out to regarding participation in the workgroup?

Best,
Piper

8

There is a universal form to join work groups. Alternatively you could reach @FrogGirl (Nicole Gerlanc) who provides excellent leadership for this group here or in OHDSI Teams :slight_smile:. You could also come to the WG meeting next Monday Nov 25th at 10am EST (link); schedule is also here: Upcoming Workgroup Calls – OHDSI.

9

@aostropolets, Thank you! :slightly_smiling_face:

10

To respond: if you don’t like the domains, another solution is not to create a 3rd but merge the 2 that cause trouble and call it ObsMeasvationment domain. (it is indeed a sore point in OMOP CDM, or any CDM).

It seems the problem you have in screenshots is the violation of principle 4
Unique Standard Concepts:
For each Clinical Entity there is only one concept representing it, called the Standard Concept. Other equivalent or similar concepts are designated non-Standard and mapped to the Standard ones.

let’s consider PHQ-9

image
image736Ă—45 2.27 KB

Let’s roll a dice and make one of them non standard.
Or have some other way of deciding if LOINC or SNOMED wins.
For example which terminology has fewer license/usage legal hurdles.
Or which models things better (has useful relationships)
Or in newly formed OHDSI evidence network (and its concept aggregate counts ) - is winning the adoption battle.

11

Hi @Vojtech_Huser,

Concept ids for the assessment instruments themselves aren’t that important, since the instrument itself is just an physical artifact. It’s not an observation or a measurement. I suspect folks may be using it in cases where there’s no concept for the corresponding score. :scream:

Changing the domain for those might help prevent bad behavior (i.e., using the concept ids for the assessments instead of concept ids for the scores) (cc-@Christian_Reich)

What we need and use:

  • Scores (SNOMED-CT, LOINC)
  • Questions (LOINC, PPI, other?)
  • Answers (LOINC, PPI, other?)

It’s a lot of work, but given that there isn’t complete overlap between SNOMED and LOINC, I don’t think we universally mark all from one terminology as standard or non-standard. If it’s only in one, it probably needs to be standard?

I don’t think it matters so much which domain we put them in, as long as we’re consistent in where they go, Although in theory it would be great if we had the time and resources to distinguish between scores that are truly measurements (psychometrically valid instruments) and those that are just observations (surveys)

Piper

12

I fully agree with this. Then I meet people who have strong opinions about one way or another :slight_smile:

The real problem you and Vojtech are highlighting is that we don’t have good definition for Meas vs Observation for edge cases. You refer to “true measurements” vs “observations” but I’m not sure the distinction is thought through. If it is by somebody here or in the outside world we’d gladly take it.

Precedence of vocabs for dedup: we started a convo here but should formalize it more on one of the Vocab WG calls.

Re specific deduplications: if any of you have time, it would be a wonderful community contribution. It will enable you to see changes without waiting for the consesus on survey harmonization to be reached. The link above is a good example of such contribution.

13

@aostropolets happy to make a contribution - and have already started the the spreadsheet!

14

Hi @Piper-Ranallo
There have been a couple of attempts to map things in the Survey domain.
You might be interested to look into what we presented to the Survey subgroup - not sure if the recording is still available @FrogGirl

Survey WG.pdf (465.8 KB)

15

@Alexdavv -
Thanks for sharing the slide deck. WOW. This is some really nice work!

Is the Survey workgroup the appropriate place for further discussion about the issues raised in the PDF?

I’m focusing right now on overlap between LOINC and SNOMED. I think it makes sense to de-dup there to start.

Thoughts?
Piper