Stream: implementers
Topic: Secondary data use for observations. LOINC and SNOMED
Michael Hosking (Jul 12 2017 at 20:32):
We are currently looking at building patient based questionnaires (eg. Oxford Hip Score - patient reported pain).
This specific questionnaire is not represented in LOINC as is other questionnaires, such as the PHQ-9 (https://s.details.loinc.org/LOINC/44261-6.html?sections=Comprehensive) - completely LOINC coded for both questions and answers.
For the purpose of displaying information outside of the context of a questionnaire to other clinical users (eg. Observation of "Patient has reported moderate hip pain"), it is known that when using an existing captured concept - the observation concept must hold semantic value outside of the context of the original source (ie. the patient questionnaire).
Currently the best option for patient questionnaire coding:
QUESTION:
- LOINC question code, must match word for word
- IF no appropriate LOINC code, use SMCT Post-coordinated expression to represent question
ANSWER:
- SNOMED CT post-coodinated expression to represent semantic value of observation (ie. body site, severity...)
I would be interested in peoples opinions and views on this.
Lloyd McKenzie (Jul 12 2017 at 21:41):
Seems like a reasonable strategy to me. Only other comment is that it's possible that some questions (or answers) might not be expressible with LOINC or post-coordinated SNOMED, so you may occasionally need to supplement with other code systems, including locally-defined codes.
Michael Hosking (Jul 12 2017 at 22:02):
Thanks Lloyd,
Agreed - we may not be able to achieve semantic matching and alternatives codes (local or otherwise) may need to be used/created (but avoided where possible)
Michelle (Moseman) Miller (Jul 13 2017 at 17:47):
See https://chat.fhir.org/#narrow/stream/implementers/topic/Code.20only.20Observation.20valid.3F (starting with my June 22 comment), which was a recent discussion about using SNOMED in Observation.code to represent a question -- such as vomiting -- and then using a SNOMED qualifier in Observation.value to represent the answer -- such as present or absent. As part of this discussion, I mentioned GF#13318, which I had logged to ask for additional Observation examples that include SNOMED in the Observation.code. While the tracker isn't resolved yet, I think there is general agreement to include such examples and guidance.
Michael Hosking (Sep 05 2017 at 22:22):
Thanks @Michelle M Miller,
I'm looking into this now with our team to understand how this might work. Appreciate the suggestions :)
Last updated: Apr 12 2022 at 19:14 UTC
