OHDSI Home | Forums | Wiki | Github

Can we store the same information in multiple domains?

Implementers
#1

Hello.

According to the discussion in:

https://forums.ohdsi.org/t/question-about-operator-concept-in-observation-domain/20836

the attributes of cancer should be under the “Measurement” domain.

As a result, we are looking into remapping some of our codes that have previously been mapped to Observation, to change them to Measurement.

For example, here is an entry from a Breast Cancer pathology report:

Anterior Margin in Millimeters (mm): 0.7mm

Previously we have this information mapped to:

Concept id:   3168340	
Concept name: Distance from malignancy to anterior surgical margin in excised surgical specimen of breast
Domain id:    Observation

And now, we are looking into remapping it to:

 Concept id:   4160787
 Concept name: Distance of invasive carcinoma of breast from closest surgical margin in excised breast specimen
 Domain id:    Measurement

When our team is discussing this case, we wonder, what if we keep this same piece of information in both places in the OMOP tables - “Observation” and “Measurement”. Information-wise, this does not seem to be wrong. And from the search perspective, technically speaking, this may provide more venues for the subject person to be identified for the cohort discovery.

I wonder if anyone has tried to store the same piece of information in multiple domains? Will this approach cause any issue later? I hope to hear feedback from the community.

Many thanks! :grinning:

#2

Not an expert here but:

Because there’s 2 concepts, I’d map to both if you are certain these are the same things are the same thing. From vocabulary perspective, it seems there should really only be one concept that represents a single thing, specifically to avoid the problem you’re running into which is ‘which concept should I be looking for when I want this clinical idea?’ If the answer is more than one concept, you run into confusion with standardized analytics.

Look at those 2 concepts in athena:

https://athena.ohdsi.org/search-terms/terms/3168340
https://athena.ohdsi.org/search-terms/terms/4160787

(links were broken in preview so I’m formatting them above)

The measurement one is from SNOMED, the observation one is from Nebraska Lexicon, which is strange because I haven’t experienced a lot of cases of concepts from that Nebraska vocabulary. But my perspective on this matter is that if the Observation is duplicated with the Measurement one, they should probably de-standardize the observation one and maybe make it a concept synonym instead.

#3

Friends:

The attributes of cancer should be recorded as Cancer Modifiers, so the correct concept is Anterior Margin. We have not had the chance to de-standardize all those SNOMED, Nebraska and LOINC codes that are overlapping.

It is wrong, because we are not optimizing that. We are optimizing the analytical use cases. Having the same information in two potential places means the analyst has to query both to make sure nothing is missed. Instead, we want to have every information to have an unambiguous representation and table destination. We are not there yet, but we should certainly not go into the opposite direction.

Hope this helps.

#4

Hello. Many thanks for all the feedback, and the correction. I will let our team know. Thank you! :grinning_face_with_smiling_eyes:

t