I’ve tried to use Usagi for UCUM mappings, but mapping seems to be not very good even when constraining the search to UCUM vocabularies. How do you usually map UCUM?
In particular, I want to know which source UCUM units has a mapping, as little discrepancies to the standard UCUM code have been detected.
Usagi seems to be failing to match to the UCUM code, probably because they are too short or have lots of special chars
UCUM is special, because the concept_name field contains the spelled out units. Nobody captures them that way. The usual abbreviations are in concept_code, but Usagi is not looking there. You may be better off doing it by hand, usually people don’t have that many units to map.
Can you show us?
1 Like
I mean in the source UCUM codes vs the real UCUM ones. UCUM codes in OMOP are correct. The only thing I’ve found is some valid UCUM codes as per https://ucum.nlm.nih.gov/ucum-lhc/demo.html not present in OMOP vocabularies. For example mcg/dL or /mcL
As Christian mentioned, we do mapping manually, but also before that we join sources on concept_code directly to UCUM
I have added this as a potential enhancement of Usagi Allow searching on concept_code · Issue #134 · OHDSI/Usagi · GitHub.
That said, simply joining the unit abbreviation on concept_code will often give you a really good mapping already. (even if the source is not UCUM)
1 Like
Main problem is with things like iU/dL which is not correct (should be [iU]/dL) but it’s close enough where Usagi would potentially pick them up