Proposal for a complimentary file to the MDRS - the "ECRS" ("Edition Composition Reference Set")
Dion has requested a discussion to clarify the best application of the Module Dependency Reference Set (MDRS) - some background reading before our meeting is here:
https://snomed.atlassian.net/wiki/display/FHIR/4.2.1.0+Using+SNOMED+CT+with+FHIR?focusedCommentId=57803275#comment-57803275
Please let me know if you have any questions before our meeting, once you've read through the info, as it would be great for everyone to start on the same page in order to have the most effective discussion on the day.
Thanks very much,
Andrew
Comments
Hi Michael,
Just to clarify - in the UK no one edition/package contains all of the data? Interesting to see that the UK edition is almost an extension of the drug extension (a leaf node of that) rather than the drug extension being dependent on the UK edition (or extension).
For us, our US Edition includes all of the data (see image). Our extensions & refsets (even when created independently) still feed back to the US Edition (which in turn, feeds updated data back to the extension and refsets).
Edited after: Okay I can't add my awesome image to this thread....
Hi Suzy, I uploaded my image to https://pasteboard.co and then inserted it as an "Image from the web" to work around the permissions problem.
In the UK we have 6 monthly UK Edition releases, containing International release files and UK Clinical release files. UK Drug releases are 4-weekly (with interim releases 6 monthly) that just contain UK Drug release files. To correctly describe the module dependency you need the three MDRS files. Here is current model based on April 2018 UK Clinical and Interim Drug with Jan 2018 International data.
Yes, my apologies - when I hurriedly generated my diagram above I accidentally filtered out the clinical extension modules
This issue has come up again for us in connection with the UK releases. The following image is a visualisation of the "inclusion view" of the MDRS content. The nodes represent modules that occur in the source context (and thus represent an Edition in the SNOMED URI sense) and the arrows indicate that all the content from the source Edition is included in the target Edition.
Since there is more than one leaf node, no Edition contains all content. Furthermore, the leaf nodes are not the Editions that are normally thought of as the UK editions.
In the discussion linked above, Dion suggests doing one of:
The sticking points for me are 1. most of the MDRS content in existence uses the module content dependency semantics, 2. the SNOMED URI specification explicitly defines and depends upon the MDRS content providing edition composition semantics.
The third consideration is the impact of change. I know that CSIRO tools depend explicitly on the edition composition semantics, but am under the impression that this is a minority situation and that most other tooling either 1. just uses the SNAPSHOT files, ignores the MDRS context and assumes the SNAPSHOT was correctly computed, or 2. has some other means of determining edition composition for the purpose of computing a SNAPSHOT.
I'd like to propose a fourth option, which is to define a new reference set, the Edition Composition Reference Set (ECRS) which, when entries are present for a given Edition module, is used by the SNOMED URI specification to determine edition composition but, for backwards compatibility, fall back to the MDRS content when/if it doesn't exist.
The advantages of this approach are:
Tooling that uses the MDRS for edition composition semantics and that hasn't been updated to know about the ECRS will continue to work, and continue to experience the same issues they do today.
Tooling that uses the MDRS for edition composition semantics and that have been updated to know about the ECRS will avoid current problems. The required change for these tools is small.
Tooling that uses the MDRS for content dependency semantics can continue to operate as it does today.
Release centres that use the MDRS for both roles can adopt the ECRS to separate the two roles explicitly, but won't have problems if they don't
Release centres that use the MDRS for content dependency semantics can adopt the ECRS with minimal change and effort and no significant tooling impact.
No changes will be required for current or historical MDRS content
Note, it would be desirable for the ECRS to be populated with historical content so that it can be used to build accurate SNAPSHOT releases for old versions from FULL data.