OHDSI Home | Forums | Wiki | Github

Mapping Clarity units

(Evan Sholle) #1

Hi all,

Apparently, we never got around to mapping the units for lab results we get out of Epic Clarity. I have a whole list of units that need to get mapped so we can populate unit_concept_id, but I’m not sure how best to do it, other than manually, which I’d like to avoid if at all possible. I tried using Usagi, but it attempted to map to some concepts that were totally off base. I can’t join to CONCEPT on vocabulary_id = ‘UCUM’ and concept_code = [source_unit_value] because the formats don’t match.

Does anyone have any relevant experience mapping units, especially from Epic? Any approaches that worked for the community would be great to hear about.

(Christian Reich) #2


It may be less awkward than you think. Can you share your lists of units? The other thing is @Vojtech_Huser’s list of common units for LOINC tests. That will help also. It’s a 80/20 job.

(Roger Carlson) #3

I mapped our using USAGI. Roughly 150 of them, so it wasn’t too bad.

Some things to keep in mind.

  1. When you first import the codes, set the Filter by vocabulary to UCUM. That will improve it’s initial guesses,
  2. When you go to match codes, make sure the filters are set again. The Import Filters look very similar to the Search Filters, but they do different things.
  3. If USAGI can’t find a match, try putting a text description from Athena into the Query box to find it. Votech’s list would be helpful here.
  4. Units are sometimes case-sensitive but SQL is not. This can be a problem. To correct it, you have to indicate a SQL Collation in the JOIN
    a, LEFT JOIN
    OMOP.source_to_concept_map AS source_to_concept_map_unit
    ORDER_RESULTS.REFERENCE_UNIT COLLATE Latin1_General_CS_AI = source_to_concept_map_unit.source_code COLLATE Latin1_General_CS_AI
    AND source_to_concept_map_unit.source_vocabulary_id = ‘SH_unit’
    b. However, it takes longer
    c. Thanks to Bruce Hope of UoV Health System for pointing this out.

(Evan Sholle) #4

Thanks to both of you - @Christian_Reich and @roger.carlson. I ended up biting the bullet and doing the mapping manually - as you pointed out, Christian, it’s a Pareto principle thing. I was able to get more than 80% of labs mapped by just doing the first 30-40 by myself, typing them into Athena one by one. But the Usagi tips will, no doubt, come in handy next time I try to use it.