In our recent Observable calls we have discussed the results of classification of LOINC Terms with a module of SNOMED CT. One example we've discussed is LOINC Term 'Amino acid:SCnc:Pt:CSF:Qn' which becomes a parent of LOINC Terms such as 'Alanine:SCnc:Pt:CSF:Qn' and 'Cystine:SCnc:Pt:CSF:Qn' after classification. We asked the Regenstrief Institute if these hierarchies were to be expected, and if the more general Terms with "Amino acids" were meant to be grouper concepts as such? Their feedback was:
"...the amino acids concept should not be a parent concept for all the individual amino acid terms. The amino acid concepts are actually pretty old, and I’m not sure anybody measures total amino acids in various fluids anymore, but regardless, these concepts are not meant to represent the parent concept in a hierarchy. In the LOINC tree, we have the part “Amino acids” (the blue highlighted row at the top) as a grouper or header for organizing all the amino acid terms, but you can see the CSF amino acid terms (the blue highlighted rows at the bottom) at the same level as the other amino acid terms:
"
We need to discuss the implications of this and any potential changes in modeling that are needed. Thank you.
3 Comments
Daniel Karlsson
For the discussion tonight here are two things that need resolution:
/Daniel
Daniel Karlsson
For the discussion tomorrow I have created a set of slides:
https://docs.google.com/presentation/d/14gGQ4lFN5EIM-95GJ_R4KKWm_xtgdZc_64uXj-S-_Qg/edit?usp=sharing
I've talked to Stefan Schulz who sent this paper:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3194169/
Daniel Karlsson
In relation to the issue of "total" vs. "non-total substances" I found that at least observables with concentration, ratio, fraction, and rate property types would refer to "total substances" and that there are about 12 500 such observables in the technology preview. About 13 500 inhere in some SNOMED CT substance. Thus, I think it is safe to assume that the issue is relevant for a major part of the LOINC observables. Note that if we would go down the route of adding "total substances" (cf. entire body structures) one can still query for observables using "non-total" substances and get the expected subsumption results.