To Negate in Vocab or by Attribute? - Phoenix WGM discussion (Jan. 2013)
Vocabulary WG sessions (extract from minutes)
7 Q1 Tuesday CVX OID discussion
7.5 Negation in Vocabulary or by Attribute?
There has been considerable discussion on a point of principle, originating in Pharmacy.
Consistency is sought so the options are:
- Should we have a preferred way of handling negation, either by explicit use of a negationIndicator, or purely through vocabulary?
- And secondly if we can some to agreement on the first question should we make that this be the recommended FHIR way of doing things.
A long thread has been progressing on this topic and it is clear that there are many difficult issues related to resolution.
In what circumstances can you rely on the terminology to all of your negation?
What are the options and where should you and should you definitely not use a particular method.
Where the design of your model and the intent of its use has implications to the way you use the terminology in the model there is a piece of metadata associated with the coded elements called terminology guidance. This could be used by the modeller who is evaluating the use cases – in these cases they can indicate that in this model in this class the use of the indicator is to be use/or the terminology is to be used – by provision of this guidance you provide a consistency for that specific model element. Where both are required, there would be two different models which differ only in their terminology guidance.
The result of the recent email thread discussion provides significant guidance on the benefits and risks of different approaches and this information could be summarised and offered as guidance which assists in making decisions about the appropriate place for these choices. In some cases the implementer has to make these choices.
Hugh Glover recommended In the first instance, the HL7 Wiki have a site to work through the options and what is the preferable solution in specific situations. This could be built from the existing email string. TermInfo intended to deal with this issue and could represent as a hot topic in TermInfo.
The issue that people wouldn’t know that this is where to go to find this information. TermInfo is currently thought of largely as SNOMED CT and V3. There is a need to change that image. How can this be achieved?
Action: Set up a Wiki to provide guidance on negation use cases. Rob Hausam will include this in the TermInfo area.
There is a further need to consider how this will progress over time.
Action: Mention to MnM that we are setting up this guidance site
Action: A summary of this discussion should be sent to the Vocab List – Rob Hausam
8 Q2 Tuesday TermInfo
As an attribute is deprecated in the RIM and now this is action negation and value negations this is not true in CDA as it still uses the single approach.
The discussion in Q1 continued.
What does the negation really mean? What are the limits of the metalevels – how would you say something – I did not send a message about whether I noted that the patient did or did not have this.
You could say:
- I did not send a message
- I did not send a message about X
- I could have sent a message
There are significant issues related to conformance – if it an element is not populated it can be thought that they weren’t able to populate it but it could be that they were able technically but did not have the information with which to populate the information.
There are people who can model SNOMED CT terms correctly but who do not understand description logic in full.
A library set of guidance to deal with the use cases
The methodology for us to figure this out is to reduce all the models in the closed world to triplets which will represent the common way of thinking about this problem.