The document "Concept Inactivation - Proposed Amendments" presents the work performed by the Concept Inactivation Working Group to review and analyse the current suite of inactivation reasons and historical associations with the objective to:
- Clarify the criteria for use of existing inactivation reasons and historical associations.
- Propose new inactivation reasons to fill existing identified gaps.
- Propose new historical associations to fill existing identified gaps.
- Define the purpose of each historical association and explain how authors should apply them and users interpret the impact on their data.
- Provide additional guidance for authors/editors.
- Recommend actions that should be taken when the concept to be inactivated has itself been the target of previous historical associations.
Instructions for review:
The layout of the main document gives an outline of the current definition, history, and current usage for each inactivation reason. These are presented so that you can see the context in which we have identified the key issues as they present themselves to authors/editors as they attempt to use the current suite of inactivation reasons and associated historical associations. With this background in mind, we have then presented our first draft proposals to improve the current process along with some additional guidance on their use so as to put the proposals into context.
In reviewing the document we would like you to read the sections up to the "proposals for change" and let us know if you think we have identified all of the issues. Then read the proposals and comment on whether you feel that they address the issues. First to help the authors/editors become more consistent in their application of the inactivation process e.g., are they more likely to choose the correct reason for inactivation and then be provided with the correct suite of historical associations to complete the process. Secondly, will the changes we have made make it easier for end-users and vendors to understand the reason for the inactivation of a given concept and the implication for them of using the historical association allocated.
Please either use the comment box below or send your responses directly to Paul Amos ( pam@snomed.org )
I hope that makes some sort of sense.
We appreciate your review of this document and your feedback on the proposed changes. Please complete your review before Friday 19th March 2021. Thank you.
7 Comments
Paul Amos
Following discussion at the EAG meeting, an addendum to the Concept Inactivation - Proposed Amendments will be provided to give further examples using the proposed new inactivation reasons and historical associations. The document is now added above.
Paul Amos
Concept Inactivation - Proposed Amendments v0.5 includes an edit to the last page that removes the incorrect link to a web-based form for feedback. Please provide feedback either on this page or directly to pam@snomed.org.
Paul Amos
Resolving Ambiguity – Options for Consideration
Following the last EAG meeting members were asked to consider the 3 options for resolving ambiguity for concepts where not all of the elements that make up full semantic equivalence could or should be offered as clinically suitable replacements.
We would be grateful if you could review the above and provide your views and comments In the comments boxes below.
The following simple example is used to demonstrate each option:
3094009 |Vomiting in infants AND/OR children (disorder)|
..has 3 potential meanings:
Vomiting in infants
Vomiting in children
Vomiting in infants and children
The sum of these 3 meanings gives complete semantic equivalence with the concept 3094009 |Vomiting in infants AND/OR children (disorder)|, however, from the perspective of clinical use “Vomiting in infants and children” is not likely to be helpful as a statement in the individual patient’s record as they cannot be both an infant and a child at the time when the record entry was made.
In order to resolve the inactivation of the ambiguous concept 3094009 |Vomiting in infants AND/OR children (disorder)|there are 3 potential solutions:
Option 1: Representing complete semantic equivalence with “POSSIBLY_EQUIVALENT_TO”:
3094009 |Vomiting in infants AND/OR children (disorder)|
Inactivated as “Ambiguous”:
POSSIBLY_EQUIVALENT_TO new concept |Vomiting in infants (disorder)|
POSSIBLY_EQUIVALENT_TO new concept |Vomiting in children (disorder)|
POSSIBLY_EQUIVALENT_TO new concept |Vomiting in infants and children (disorder)|
As a second stage:
new concept |Vomiting in infants and children (disorder)|
…would be inactivated as “Erroneous” with an annotation; “Vomiting in infants and children is an erroneous concept because an individual cannot be both a child and an infant at the same time”.
There is then a judgment to be made, whether to provide a potential replacement or not. Given that the concept is deemed “Erroneous” it would be acceptable not to provide a suitable replacement that is of equal or less specificity than the original, however, it may be more helpful to use a historical association of “REPLACEED_BY” and point to the concept 422400008 |Vomiting (disorder)|.
This approach clearly indicates to the user and system implementer that 3094009 |Vomiting in infants AND/OR children (disorder)| is an ambiguous concept and that it is possible to provide full semantic equivalence using individual concepts to exactly represent each element of the ambiguity and from which the user must choose one as a target replacement.
If the user does not know whether this was intended to be vomiting in an infant or child, they would choose |Vomiting in infants and children (disorder)|, at which point the system would then indicate that this concept had been inactivated and replaced by 422400008 |Vomiting (disorder) as the replacement.
This approach intends to make explicit the thinking behind the inactivation and its resolution in a way that is both machine-processable and transparent to the end-user.
There is an impact on the author/editor in having to create and then inactivate the concept that is never to be released as an active concept but much of this process could be supported by enhanced tooling.
Option 2: Using a combination of both “POSSIBLY_EQUIVALENT_TO” and “POSSIBLY_REPLACED_BY”:
3094009 |Vomiting in infants AND/OR children (disorder)|
Inactivate as “Ambiguous”:
POSSIBLY_EQUIVALENT_TO new concept |Vomiting in infants (disorder)|
POSSIBLY_EQUIVALENT_TO new concept |Vomiting in children (disorder)|
POSSIBLY_REPLACED_BY 422400008 |Vomiting (disorder)
An annotation would need to be added; “This concept cannot be made fully semantically equivalent using only clinically useful “POSSIBLY_EQUIVALENT_TO” concepts and therefore a clinically relevant “POSSIBLY_REPLACED_BY” alternative has been provided for that element that is thought to be clinically in error”.
This approach appears to give the same end result as option 1 above, however, the end-user is shielded from the detailed judgment that the author/editor made to identify full semantic equivalence of the inactivated concept. How they then identified and replaced that element that was deemed not to be clinically useful. It also makes it more difficult for implementers if the resolution of ambiguity is sometimes achieved with “POSSIBLY_EQUIVALENT_TO” alone and sometimes in combination with “POSSIBLY_REPLACED_BY”.
3. Using “POSSIBLY_REPLACED_BY” alone:
3094009 |Vomiting in infants AND/OR children (disorder)|
Inactivate as ambiguous
POSSIBLY_REPLACED_BY new concept |Vomiting in infants (disorder)|
POSSIBLY_REPLACED_BY new concept |Vomiting in children (disorder)|
POSSIBLY_REPLACED_BY 422400008 |Vomiting (disorder)
The issue with this approach is that it implies that it was not possible to provide semantically equivalent concepts for ANY part of the concept 3094009 |Vomiting in infants AND/OR children (disorder)|. This is of course not true as we have already established that 2 of the 3 elements can be provided with semantically equivalent alternatives.
We would be grateful for your views and comments.
Kind regards
Concept Inactivation Working Group
Keith Campbell
I'm adding these comments on behalf of Penni Hernandez:
Option 1: Representing complete semantic equivalence with “POSSIBLY_EQUIVALENT_TO”:
** Is it necessary to cover a non-sensical meaning for a concept already designated as ambiguous, do all three cases need to be represented on would the first two be acceptable. Would also eliminate the need to create and immediately retire a concept (I think that is what this is saying). Is it really an error to not represent “vomiting in infants and children”? It seems like an acceptable solution to just use the first two and not try and represent a non-sensical meaning just to cover all possible permutations of the original concept. There is probably use cases where this wouldn’t be acceptable.
Option 2: Using a combination of both “POSSIBLY_EQUIVALENT_TO” and “POSSIBLY_REPLACED_BY”:
** question about the name of “possibly_replaced_by” past tense, sounds like you might not have done it, but I think it means “possibly replace with”. Is this a relatively new association type or do I just not remember it already? This seems like an acceptable solution, since it seems like an ambiguous inactivation would need to be reviewed by a human no matter what and seeing two different types of associations wouldn’t seem like an issue.
** It doesn’t seem to me that any of these associations denote “semantically equivalent concepts”, even “possibly_equivalent_to” so I’m not sure this option wouldn’t be suitable and perhaps simpler.
Monique van Berkum
The three options provided offer four possible replacement codes:
In addition to the problems already pointed out with the third code, the first two codes are also problematic because of the plurals. Instead, they should be:
Since all three parts of the original code had problems, and because of the complexity of trying to decide whether to point an already ambiguous concept to multiple options that involve “possibly equivalent to” and/or “possibly replaced by”, I think that option three is adequate and simplest. I also prefer it because it avoids any claims of “possible equivalency”.
Therefore, I agree with the comment above and would go with option three:
POSSIBLY_REPLACED_BY new concept |Vomiting in infant (disorder)| (singular as opposed to plural)
POSSIBLY_REPLACED_BY new concept |Vomiting in child (disorder)| (singular as opposed to plural)
POSSIBLY_REPLACED_BY 422400008 |Vomiting (disorder)
Paul Amos
Monique van Berkum, thank you for your comments.
Firstly, apologies for the example chosen. I agree, as per our editorial guidance plurals should be removed, however, the example is intended to demonstrate a principle, which is that for a given ambiguous concept it may be that not all of the individual elements can or should be provided with semantic equivalents.
Where a semantic equivalent for a given element is available we are suggesting that "POSSIBLY_EQUIVALENT_TO" is the appropriate historical association as is the case for an ambiguous concept in which all of the elements have appropriate semantically equivalent concepts. However, where an ambiguous concept includes a mixture of elements, some that can be provided with semantic equivalents and some elements that cannot. Where we are unable to provide a semantically equivalent concept to a given element we must indicate that the best we can do is provide a replacement that is either of similar or less specificity than the element it replaces. It is the "POSSIBLY_REPLACDED_BY" historical association that serves that purpose.
Options 1 and 2 superficially achieve the same outcome. However, missing out on the interim stage of inactivation with replacement along with the annotation explaining why the concept should be inactivated means that we have an incomplete audit trail.
I hope that helps explain our thinking.
Paul Amos
Keith Campbell, thank you for posting the comments from Penni Hernandez.
The "POSSIBLY_REPLACED_BY" historical association is new and is based on the existing "REPLACED_BY" historical association.
The "REPLACED_BY" historical association represents an alternative concept which in the judgment of the author/editor is either of the same or less specificity than the inactivated concept and which might be appropriate for specific use cases. This clearly differentiates it from the use of "SAME_AS" for duplicate concepts in which full semantic equivalence is expected. An end-user who encounters an inactivated concept that has a historical association of "REPLACED_BY" can either accept the judgment made by the author/editor or look for an alternative to satisfy their use case.
The "POSSIBLY_REPLACED_BY" historical association has been created to cater to those situations in which a single inactivated concept may be replaced by one of 2 or more concepts. An example is where a single organism after genomic investigation is found to be not a single organism but a number of different organisms of which for any given instance the end-user must choose whichever is applicable. Therefore, the addition of "POSIBBLY" indicates that there a number of options from which the user must choose one.