Stream: snomed
Topic: Blood Test Results
Olivia Morrow (Nov 12 2019 at 15:56):
can anyone confirm which code and semantic type/top level concept in snomed should be used to capture a blood test RESULT (as opposed to ordering the blood test). For example the appropriate code to capture the results of a neutrophil count test?
Rob Hausam (Nov 12 2019 at 16:58):
I believe it should be an observable entity. Not sure that it exactly fits for a count, but the closest one that I see to your example is 57309003 |Neutrophil production, function (observable entity)|.
Michael Lawley (Nov 12 2019 at 19:58):
If you want a RESULT, then I would expect a finding. Some descendant of 118245000 |Measurement finding| such as 165516004| Neutrophil count normal| or 165517008| Neutrophil count below reference range| etc, for example
Rob Hausam (Nov 13 2019 at 00:18):
@Michael Lawley I expect the intent for this example is to record the numeric result of the neutrophil count, rather than the derived finding. Maybe @Olivia Morrow can confirm.
Olivia Morrow (Nov 13 2019 at 00:30):
Thank you both for your reply ! Yes it would be a general code to record the numeric result irrespective of whether it is normal or out of reference range .
Is it also correct t then that the “procedure” code would be used to order the blood test rather than denote the value of the result ?
Michael Lawley (Nov 13 2019 at 01:37):
Sorry, I always get confused by the extra level of indirection - naming (describing) the result value, not providing an actual result value.
Robert McClure (Nov 16 2019 at 16:49):
@Olivia Morrow In general, we've recommended the use of LOINC to represent the test (observation) that was done for which you have a result. So your example of Neutrophil could be 26499-4 Neutrophils [#/volume] in Blood . You would use SNOMED CT to report an impression of that result such as those list above.
Michael Lawley (Nov 18 2019 at 00:16):
@Robert McClure I think you need to be clear on the "we" there - I don't believe there is international agreement on this (no LOINC in UK for eg)
Olivia Morrow (Nov 29 2019 at 15:44):
We are building profiles for blood test results. We get input from many different labs and are trying to align our SNOMED codes with those that are used by the lab/NHS for the sake of interoperability. If one lab for instance reports using the code 1011481000000105 (Estimated glomerular filtration rate using creatinine Chronic Kidney Disease Epidemiology Collaboration equation per 1.73 square metres (observable entity)) and another lab uses 1020291000000106 (Glomerular filtration rate calculated by abbreviated Modification of Diet in Renal Disease Study Group calculation (observable entity)) would it be advised to make two separate profiles to capture which equation type was used by the lab? or would you aim to simplify, making a single, more generic profile for eGFR and then populate data there with values from all labs?
Grahame Grieve (Nov 29 2019 at 17:50):
there's no single or authoritative answer. And the answer, if there was one, would be very contextual; I would expect that in most cases, those GFRs would have the same meaning (accounting for reference ranges) but in others, I would care very much about the difference
A different question is: why would you have 2 separate profiles - what would that achieve? how would the data be structurally different? (even if they had different meanings in a given context)
Olivia Morrow (Dec 03 2019 at 14:09):
there's no single or authoritative answer. And the answer, if there was one, would be very contextual; I would expect that in most cases, those GFRs would have the same meaning (accounting for reference ranges) but in others, I would care very much about the difference
A different question is: why would you have 2 separate profiles - what would that achieve? how would the data be structurally different? (even if they had different meanings in a given context)
Thanks Grahame! Could an alternative way be to use the "parent" concept "egFR" within the observation profile and then have two different codes for each of the two different calculations as part of the "method" ?
Grahame Grieve (Dec 03 2019 at 19:02):
sure. a profile with a value set for the code would certainly be my starting expectation
Lin Zhang (Oct 11 2020 at 11:43):
A short answer:
Use LOINC for the Question (the order/observation <lab report item>) and SNOMED CT for the Answer (i.e. the result value).
There is an long-term official agreement between them:
Last updated: Apr 12 2022 at 19:14 UTC
