Stream: implementers
Topic: Barthel Index for the assessment of daily living skills o...
Luise Oeppert (Aug 05 2021 at 07:12):
Hello,
in nursing there is an assessment based on an index (Barthel Index). This index is divided into a scale, which can only have the values 0, 5, 10 and 15.
An example: It is evaluated whether the patient can get up allone from a chair. The score can be 0, 5, 10 or 15.The scores are linked to different meanings, but these are not to be presented. A value set is therefore excluded.
We would like to use an observation and map the result (one of the values 0, 5, 10 or 15) into value[x].valueQuantity. Is it possible to say that in value[x].valueQuantity only the values 0, 5, 10 and 15 are possible?
If not, is another type suitable, perhaps value[x].integer or value[x].string?
Mareike Przysucha (Aug 05 2021 at 07:32):
Hello there,
in LOINC there is a panel for the Barthel index (https://loinc.org/96762-0/). The codes in the panel also provide answerLists. Each answer in these lists has a human readable answer, a score and an answer ID. The answerLists can then be used as Value Sets in a valueCodeableConcept.
I hope the answer is helpful.
Jose Costa Teixeira (Aug 05 2021 at 12:14):
@Luise Oeppert Why are ValueSets excluded? in our case, this is done with one VS because that is what we mean:
Whether the patient can get up alone from a chair - the answer is not a score, the answer is a set of values ("yes", "yes, with help", "yes, with more help", "no"). Each corresponding to a score.
Those are the answers you give, that is what you monitor.
"5" is not a measurement of anything, it is just the score you assign to the answer "yes, with help"
Jose Costa Teixeira (Aug 05 2021 at 12:15):
example (sorry, in french): http://build.fhir.org/ig/hl7-be/riziv-inami/Questionnaire-mkatz-questionnaire-form.html
Luise Oeppert (Aug 05 2021 at 12:39):
Thanks for your answers.
As an implementer, I also think that using valuesets is correct here. Especially because there is coding available in Loinc.
I am trying to convince my team to use the existing coding and not only use the scores.
John Silva (Aug 05 2021 at 14:08):
I remember a similar situation where the clinicians on a team I worked with wanted the values because they used it in some other calculation. (there were even situations where different scoring systems could be used; I think it was Pain Scale and the values had to be normalized). It would still be possible for some other logic to perform calculations based on the score values while still presenting the questions and answers in the coded form.
Jose Costa Teixeira (Aug 05 2021 at 14:22):
we had a discussion around the values and their numeric meaning. It went like this:
Jose Costa Teixeira (Aug 05 2021 at 14:28):
We _could_ say that the "value" for "can't out up of chair alone" is 15. But normally what you are asking is not "is it 10 or 15" but "can the patient get up or not at all".
Then there was the idea "ok, can we make it a coded element where the code 15 has display "can't out up of chair alone". What happens if you use another scale of atuonomy or if the algorithm gets reviewed and there the code is 20? you'd have a mess in terms of transition, legacy data...
So, the code is "cannot-get-out-of-chair-unassisted" or "inability-to-get-up-alone" or "SNOMED 111222333444555" if it exists.
The code remains decoupled from its end purpose, data can be reused.
For such scoring, what we have is a lookup/expression: "if the code is X, add 10 to tscore; if the code is Y, add 15"
Lloyd McKenzie (Aug 05 2021 at 15:15):
There's a standard extension (http://hl7.org/fhir/StructureDefinition/ordinalValue) that can be used to communicate the numeric value for the code
Mareike Przysucha (Aug 05 2021 at 17:24):
I also think ART DECOR exports value sets with this extension.
Last updated: Apr 12 2022 at 19:14 UTC
