Stream: terminology
Topic: radlex
John Moehrke (Apr 06 2021 at 15:43):
Is there somewhere that the fhir core build can access the radlex valueset? The ImagingStudy resource today points at a download
<valueSet value="http://loinc.org/download/loincrsna-radiology-playbook-file/"/>
is there a way to get this available in the fhir spec?
Lloyd McKenzie (Apr 06 2021 at 16:27):
@Grahame Grieve @Rob Hausam
Rob Hausam (Apr 06 2021 at 16:32):
@John Moehrke Do you want to have RadLex in the FHIR spec, or available in tx.fhir.org? I'm thinking it's probably the latter, but I don't know for sure what you're thinking. I think that should be doable (I'm not sure what the current status is).
John Moehrke (Apr 06 2021 at 16:36):
@Jonathan Whitby @Christopher Lindop ?
Robert McClure (Apr 08 2021 at 23:37):
@John Moehrke Sorry but this makes no sense. That download is not a value set by my standards. Have you looked at the csv file (among all the files in the zip) and can you tell me what are the concepts in the expansion? Where is this specifically used in the resource and who let it pass inspection?
Robert McClure (Apr 08 2021 at 23:41):
Found it. What a mess. it is on ImagingStudy.procedure.value[x] @Grahame Grieve @Lloyd McKenzie Please tell me this should be generating errors and explain why FMG let this pass?
Robert McClure (Apr 08 2021 at 23:45):
If I was going to guess, this should be a value set of the set of LOINC codes in the first column of the csv file.
John Moehrke (Apr 09 2021 at 00:33):
Hence I ask terminology for help
Elliot Silver (Apr 09 2021 at 02:49):
If I recall, at the time RadLex was proposed as the value set (DSTU2?) it was maintained outside by RSNA outside LOINC. (I'm sure someone who has been around longer or has a better memory can correct me.) Since then, appropriate values were created in LOINC, and mappings created. My guess is that the binding hasn't been changed since.
There is (apparently) a ConceptMap, that I think is from the old values to the new ones. On a quick search, I don't see how to get the RadLex subset of LOINC, but it's probably obvious to a LOINC expert.
Elliot Silver (Apr 09 2021 at 02:57):
Oh, look this LOINC page identifies http://loinc.org/vs/loinc-rsna-radiology-playbook as the canonical URL. (There is also a http://loinc.org/vs/loinc-rsna-radiology-playbook-core; I don't know the distinction.)
Lloyd McKenzie (Apr 09 2021 at 04:21):
@Robert McClure Why would you expect this to be an error? There's no requirement that the canonical URL for a value set resolve to an actual value set. We would have spit out a warning that we couldn't resolve the URL and look it up, but it wouldn't have been an error. Obviously, it IS an error, as it's not actually talking about a ValueSet at all, so a tracker item to get it fixed would certainly be in order.
John Moehrke (Apr 09 2021 at 12:43):
Elliot Silver said:
Oh, look this LOINC page identifies http://loinc.org/vs/loinc-rsna-radiology-playbook as the canonical URL. (There is also a http://loinc.org/vs/loinc-rsna-radiology-playbook-core; I don't know the distinction.)
Thanks @Elliot Silver for giving me the canonical URL. I tried that, and it still does not expand. It has that canonical URL in the structureDefinition, but when one clicks on it to see what the valueSet is defined as, it gets the loinc login. I was hoping for the kind of page that explains the valueset
John Moehrke (Apr 09 2021 at 12:49):
It is likely more proper to use the canonical URL, then the web site download URL. but I don't know.
John Moehrke (Apr 09 2021 at 13:19):
Would really like if radlex was maintained by LOINC or RSNA as a ValueSet resource. would it be horrible for us (hl7) to define a ValueSet resource with the proper description and such (the normal elements in ValueSet resource filled out), with a defined canonical as given on loinc.org? This is what we have historically done with DICOM and IHE - having since moved to using ValueSet resources that DICOM and IHE maintain and publish.
Isn't there some vocabulary workgroup mechanism like this? Is this HTA? Is this UTG? Is this ABC? (I can't keep track of acronyms in this space).
Robert McClure (Apr 09 2021 at 14:37):
Why would you expect this to be an error? There's no requirement that the canonical URL for a value set resolve to an actual value set. We would have spit out a warning that we couldn't resolve the URL and look it up, but it wouldn't have been an error. Obviously, it IS an error, as it's not actually talking about a ValueSet at all, so a tracker item to get it fixed would certainly be in order.
FHIR error - no because FHIR doesn't care if the value set makes sense. Human error - yes, because the the file that downloads automatically has multiple choices with no clear indication as to what is supposed to be a valid value.
Robert McClure (Apr 09 2021 at 14:44):
@Elliot Silver - Yes, you found the playbook FHIR resource - excellent. I think that is what should be used in the binding and it aligns with that first column in the file.
@John Moehrke
Would really like if radlex was maintained by LOINC or RSNA as a ValueSet resource. would it be horrible for us (hl7) to define a ValueSet resource with the proper description and such (the normal elements in ValueSet resource filled out), with a defined canonical as given on loinc.org? This is what we have historically done with DICOM and IHE - having since moved to using ValueSet resources that DICOM and IHE maintain and publish.
Isn't there some vocabulary workgroup mechanism like this? Is this HTA? Is this UTG? Is this ABC? (I can't keep track of acronyms in this space).
It is a value set resource but it's not one that renders in html nicely, which is what I think you want. Of all the entities responsible for making that file look pretty to humans, it's not HTA or vocab I can tell you that. @Lloyd McKenzie what do we need to ask of LOINC to add to the resource file so that anyone that accesses it would see a pretty file too?
Lloyd McKenzie (Apr 09 2021 at 16:41):
Well, having the zip include a ValueSet instance in either XML or JSON (or ideally both) would be nice...
Last updated: Apr 12 2022 at 19:14 UTC
