Page tree
Skip to end of metadata
Go to start of metadata

Description

Management of Concept Inactivation - Review of existing inactivation values and production of clear guidance for each identified use case.

Objectives

  • Identify potential use cases 
  • Development of scenarios
  • Review of existing inactivation values and proposal for new values
  • Development of guidance for each Use Case/Scenario
  • Engage the community of practice for feedback
  • Present outcome and gain sign off from EAG
  • Update Editorial Guide
  • Provide guidance for the community of practice

 

Relevant Documents

  File Modified
ZIP Archive History Association QA (1).zip 2019-Oct-08 by Jim Case

1 Comment

  1. Comment from Jeremy Rogers:

    I think my main recommendation would be to start by identifying the full range of ‘data healing pattern’ semantics that exist and their combinatorial logic, and then work backward from there to a set of historical associations that is more explicitly - nay, rigidly - defined in terms of an unambiguous mapping to exactly one of those healing patterns. In principle you could still consider having more than one human readable flavour of association mapping to the same underlying machine readable ‘data healing pattern’, but I’m not sure that’s going to be either necessary or helpful. 


    The most important shift in authoring psyche is from casually heuristic, inconsistent and incomplete explanations for humans as to why stuff has been expunged from the terminology, to formal and complete machine readable statements about how existing patient instance data and queries should be healed as a consequence of those changes. 


    The range of healing patterns isn’t large; there are I think ultimately only 5 core:

    TRUE 1:1 SEMANTIC DUPLICATION – SAME_AS, REPLACED_BY, MOVED_FROM

    OLDID and NEWID are semantically equivalent and therefore freely interchangeable identifiers for exactly the same ‘thing’: all possible instances of data that either already are or that could ever be coded to OLDID in the real world, can be exactly re-coded to NEWID (and vice versa). There is no possible instance of an OLDID that is not a subtype of NEWID, and vice versa. No information currently within the meaning of either OLDID or NEWID would be lost if you transcoded from one to the other, and none would be added.

     1:many SEMANTIC AMBIGUITY – MAY_BE_A


    OLDID has one or more possible interpretations but all possible interpretations are fully covered in combination by a nominated set of {NEWID1, NEWID2…}. For all possible instances of data that either already are or that could ever be coded to OLDID in the real world, exactly one member of the candidate set {NEWID1, NEWID2..} is semantically completely equivalent to the particular interpretation of OLDID in use for that data instance. No information currently within that particular interpretation of the meaning of OLDID would be lost, and none would be added, if OLDID were transcoded to that specific NEWID variant. The reverse operation – transcoding all possible NEWID-coded data to OLDID would however be unpredictably ambiguous as to whether certain information previously part of the meaning of NEWID was necessarily still universally implied by OLDID, and also whether the meaning of OLDID adds information not originally implied by NEWID.

     1:1 SEMANTIC AMBIGUITY – MAY_BE_A


    OLDID has one or more possible interpretations but only one of those possible interpretations can ever be represented and fully covered by a SNOMED concept, with all other interpretations being outside the scope or expressivity of SNOMED. For all possible instances of data that either already are or that could ever be coded to OLDID in the real world, only some may have used the particular interpretation of OLDID for which NEWID is exactly equivalent. For these data instances only, no information currently within that particular interpretation of the meaning of OLDID would be lost, and none would be added, if that OLDID-coded data instance were transcoded instead to NEWID. The reverse operation – transcoding all possible NEWID-coded data to OLDID would however be unpredictably ambiguous as to whether certain information previously part of the meaning of NEWID was necessarily still universally implied by OLDID, and also whether the meaning of OLDID adds information not originally implied by NEWID. All other data instances coded with any other possible interpretation of OLDID can have no meaningful SNOMED data repair.

     1:1 SEMANTIC ABSTRACTION – WAS-A


    OLDID has a single unambiguous interpretation, but only part of that meaning can ever be (and currently is) represented within SNOMED as a unique concept; all other aspects of the particular meaning of OLDID are outside the scope or expressivity of SNOMED. NEWID is the unique SNOMED concept that captures as much of the full meaning encapsulated in OLDID as it will ever be possible to represent in SNOMED. Thus, all possible instances of data that either already are or that could ever be coded to OLDID in the real world, can be abstracted to NEWID but this is a lossy transform, with some part of the original meaning of OLDID being lost in the transcoding. The reverse transform is invalid: transcoding all possible NEWID-coded data to OLDID would entail adding information not already implied by NEWID, except where the item had originally been coded as OLDID.

     1:many SEMANTIC ABSTRACTION – WAS-A


    OLDID has a single unambiguous interpretation, but only part of that meaning can ever be represented within SNOMED as a unique concept; all other aspects of the particular meaning of OLDID are outside the scope or expressivity of SNOMED. Currently, however, a unique concept to fully capture that partially representable meaning does not exist as a pre-coordinated concept in SNOMED. Therefore, instead of a direct link to that single concept, a set of active SNOMED concepts {NEWID1, NEWID2..} is offered that collectively cover that meaning only in combination; no individual member of the set offered completely covers it but rather each covers only even more submaximal fraction of the original full meaning of OLDID. Usually, this set comprises those concepts that were historically the direct ancestors of OLDID. Conceptually, they are all also common ancestors of the missing ‘partial meaning’ concept.

    To this core quintet we can I think add two extra flavours:


    •  A variant pattern of true semantic duplication in which both OLDID and NEWID are, and can in perpetuity legitimately remain, concurrently active - as for example between <findings> or <procedures> with and without their explicit default context wrappers. 
    •  A variant pattern of 1:many Semantic Abstraction, in which Navigational Concepts are re-linked back to concepts with 1:1 semantic equivalence  to the direct ancestor and descendent concepts that they had in taxonomy of their original source terminology (CTV3 and SNOMED RT)