Stream: implementers
Topic: Codesystem vs conceptmap
Medi Harsini (Sep 29 2020 at 19:46):
Is there a recommended way to implement the code maintenance between the two? How do we maintain the code on conceptmap if it gets changed in the codesystem. Perhaps having the reference on the codesystem to conceptmap is required
Michael Lawley (Sep 29 2020 at 22:01):
Can you describe the kind of changes you're talking about?
Medi Harsini (Sep 29 2020 at 22:44):
Michael Lawley said:
Can you describe the kind of changes you're talking about?
Any concept.code changes which impacts a code translation which is already defined under a different concept map. This may be pattern, but a concept.code change requires a cascading updates on all the mapping components (including concept map) to ensure the changed code is changed there too (either delete or upsert or a complete new code)
Michael Lawley (Sep 29 2020 at 22:51):
In a properly manage code system, code meanings should not change. But, in general, there's not really any alternative than finding the changes in the code system and then manually updating the map.
If your code system includes codes with meanings like "other", then merely the addition (or retirement) of a different code can change the (relative) meaning of your "other" code and consequentially require a map update.
It is ultimately a manual process.
Michael Lawley (Sep 29 2020 at 22:53):
Some code systems publish change artefacts that can help in the maintenance process. For example, SNOMED CT publishes a set of "Historical Association" reference sets that indicate potential replacement concepts and reason for retirement.
Medi Harsini (Sep 29 2020 at 23:20):
@Michael Lawley thanks it makes sense. Thought it would be easier if there was a pointer instead of the actual value on the concept map.
Medi Harsini (Sep 29 2020 at 23:29):
On a different note but related; I’ve got a number of data consumers in which each of them has a number of hospitals as the clients. Those hospitals requires to send us data via those consumers. None of them have FHIR based codesystem. I’m planing to create codesystems defined in my FHIR server for each of those hospitals and then create a concept map for each of them to my codesystem. Perhaps for gender, we have have 10 gender codesystem, each belong to a hospital and one which is mine. So 10 code system and mine (11) and one concept map for all together Mapped to mine. I will then use this conceptmap to map the values before it’s written to downstream systems and ingested. Would you see a fault or perhaps anti pattern in this design ?
I’m just taking advantage of having a subject expert matter and to check my designs*
Appreciate your insights there
Michael Lawley (Oct 05 2020 at 03:51):
One question is whether you can also preserve the original codes. The other is how you deal with lossy mappings and the downstream implications. Gender might seem like an easy case, but depending on whether your codes and the source codes actually refer to biological sex or gender identity can become quite complex. A lot depends on what these downstream systems and the subsequent use is going to be.
Last updated: Apr 12 2022 at 19:14 UTC
