SNOMED CT Editorial Advisory Group (AG)
Teleconference via GoToMeeting
Attendees:
Chair: Jim Case (JCA)
AG members:
Bruce Goldberg (BGO)
Guillermo Reynoso (GRE)
Paul Amos (PAM)
Observers: Cathy Richardson, Donna Morgan, Krista Lilly, Monica Harry, Nicola Ingram, Linda Parisien (LPA), Penni Hernandez, Yongsheng Gao (YGA)
Apologies:
Keith Campbell (KCA)
Minute-taker: Juliet Gole Krarup, from a recording.
Roll Call
JCA welcomed the group and went through attendees, noting that KCA was not on the call.
Conflicts of Interest
None, other than those previously declared
Review of the Agenda
No additions or changes
Approval of the minutes from the Uruguay Meeting
JCA noted that due to the length of the meeting, the minutes had only finished being transcribed the previous week, so he suggested going through the minutes for the next call so that AG members would have a chance to review them.
- AG to review and approve minutes of the Uruguay meeting at the next teleconference.
Laterality
JCA introduced the topic (see screenshot to the right).
GRE said in Uruguay the group had discussed whether to precoordinate all lateralizable structures, and it also discussed a new, fourth option of using a laterality attribute for modeling procedures, findings and other things and eventually assigning a proper role group. He said he and YGA had had an action item to test it, and they had held some meetings together and with other groups (such as the Observables group). Some stakeholders, they found, were in favor of precoordination of lateralizable structures, but he and YGA had decided that the best option was to use a laterality attribute, and it did not need to be a new one.
JCA noted that that option 4 of the attribute had been the preferred solution from the Uruguay meeting, and he too had done some testing and agreed that there did not seem to be an advantage to creating a new attribute.
GRE said that YGA's testing had shown that it was possible to maintain what was already there and avoid precoordination of more lateralized structures, and postcoordination was still possible. The only caveat, he said involved translation of RF2 into OWL. He ageed with JCA that adding a new attribute would make it hard to compute the equivalency.
JCA said that if they were to go ahead with those recommendations, then they needed to come up with some editorial guidance on how to use either (a) the laterality attribute in the role group or (b) the precoordinated lateralized body structure in a concept model when it already existed as opposed to using the attribute. He said his preference was to use the attribute and not the precoordinated lateralized structure unless there was some modeling implication in doing that.
GRE said to be consistent, having the attribute in the role group would be the correct classification.
BGO said he felt that there were distinct use cases for precoordinated lateralized structures, so he did not want to say that they should not be used in the future.
GRE replied that the main problem in creating precoordinated lateralized structures was not creating the concept itself, but having to create all the relationships and hierarchies between the lateralized structures. It would be significant overhead, he said. He said there were different ways of handling it in the new anatomy model, including GCIs, and he suggested that YGA post the findings so the AG could discuss them on another call.
JCA said the AG would need to identify use cases for where additional precoordinated lateralized structures would need to be created, but he agreed that the overhead in preemptively having to create all the lateralized structures to be in the models created a large overhead.
GRE spoke briefly about informal discussions he had had with members of the Modeling AG. He then added that the proposed approach was both forward and backward compatible.
PAM asked about impacts on implementers of the proposed approach. GRE said again that it appeared to be backward and forward compatible, and the Modeling group was stressing the importance of avoinding impacts on implementers.
JCA asked what requests would need to be sent to the Tooling group to allow the editors to implement the proposed guidance? GRE replied updating the allowable domain to represent and enforce RCN across the new tooling. It would not require any new constructors, he said. It would have to be complemented with guidance to the authoring team to avoid using precoordinated lateralized structures. JCA agreed, noting that it could be written into the Editorial Guide: don't use precoordinated structures; use the role group instead.
GRE suggested having a separate call about his and YGA's testing. JCA asked them to write up a summary and post it on Confluence. GRE agreed.
- Guillermo Reynoso and Yongsheng Gao to write up and post a summary of their testing on laterality for use at a later AG call.
- JCA went through some to-do's: draw up a set of the AG's recommendations and the current communications the AG had had with other AGs; write up for the Tooling AG a description of what would have to change for the machine readable concept model to support the use of the laterality attribute in other parts of the terminology; and work with the Content Managers AG to put a priority on when it would start.
JCA said that the next point on laterality was that the proposed approach would add a substantial amount of content, and some users might say it was bloat because they could do the same thin in post-coordination. KCA, JCA said, had proposed separating it out from the core. The concept of modularizing out certain aspects of the terminology was generalizable, he said, to a number of precoordination patterns, so it seemed like a good idea. He asked for opinions on how it should be done: an extension, new module ID, refset that could be extracted out? He clarified that he meant separating out anything that had lateralized content, not just the precoordinated lateralizable body parts but also procedures, etc.
GRE said that according to YGA's analysis, that would be about 80,000 concepts. He suggested that it might be a solution looking for a problem. At the moment there were only about 700 concepts that had been modeled using lateralized anatomical structures. He said he was in favor of extensions for additional content to help implementations, but the trick would not be in creating the concept, it would be in all the modeling that would have to be done to ensure that the body structures had the same model as the generic ones. They would have to connect all of the concepts, otherwise there could be isolated siblings.
JCA said he wanted to take the entire anatomy hierarchy out of the discussion by saying no more lateralized structures would be added except under exceptional circumstances. The use of the attribute to model the other hierarchies would be preferential. As long as the anatomy is correct, there is no risk of unrelated siblings because they rely on the accuracy of the non-lateralized relationships.
GRE said that in that case, there would be a need to retire the precoordinated anatomical structures. There were about 1100 of those, he said. JCA replied that that could be done gradually and with advice from the IHTSDO Members. The Members could also be presented with the option of whether it should be integrated into the International Release or put into a separate extension.
BGO said he saw use cases for still having precoordinated lateralized structures. JCA replied that he was not alone, which is why the plan included retention of the lateralized structures until they had advice from the Members going forward. In the short term the editorial guidance would say that they should not use them, and over time the content that already used lateralized structures as part of the models would be revised to use more consistent modeling structures. BGO agreed with that.
LPA made a comment about use of lateralized structures in diagnostic imaging in Canada, but due to some background noise it was difficult to hear her. JCA asked her to write up the use case as she had done for unilateral. She agreed to that.
- Linda Parisien to write up use case on lateralized structures in diagnostic imaging in Canada.
JCA concluded the discussion by saying that the AG would return to it in a future call.
Unilateral Concepts
JCA introduced the topic by going through the text on the screenshot to the right.
He said he thought the group might want to revisit the option of retiring the concepts as ambiguous primarily because it did not have a current solution for using the concepts. They remained, he said, in the situation hierarchy to avoid incorrect inferences in the findings hierarchy. Or, he said, they would have to make them siblings that would not be connected to right, left or bilateral at all.
GRE said that last option was probably the least bad option. Making them siblings even if they would not have a way of distinguishing bilateral from unilateral was the lesser problem until a good solution could be found. That solution might be a technique, modality or way the procedure is done, he said. Previously they had tried to do it using the lateralizable structures, but it could be about modeling the action or modifying the action in some way. They had not explored that so far, he said.
JCA noted that the issue was indeed much more prevalent in the procedures hierarchy than in the findings hierarchy because of the way procedures are named. JCA noted that there are three potential solutions that they might want to ask the Members about:
- Retire the concepts as ambiguous (and this was the most radical solution).
- Add them to the findings hierarchy but keep them as siblings to the lateralized structures, which unlinks them from bilateral, left and right.
- Leave them where they are.
He suggested asking Members for input, arguments, and use cases for one or more of those solutions.
- Ask Members for input, arguments and use cases on the 3 potential solutions proposed for unilateral concepts.
BGO referred back to modeling unilateral in procedures. He said he did not see a use case for having a structure that was unilateral. He suggested using the "has laterality" role with a qualifier value of unilateral. JCA asked him to write up some examples.
- Bruce Goldberg to write up some examples of how one might model unilateral using the "has laterality" role with a qualifier value of unilateral.
WAS-A
JCA and GRE introduced the topic (see GRE's input in screen shot to the right). GRE said WAS A was a compromise for concepts dating from Jan. 2010 and did not make sense in RF2. He said there were two options to deal with WAS A relationships: (1) retire the WAS A, leaving the target concept without a historical relationship; (2) provide a link to a supertype of the concept being retired. He said stakeholders seemed to have been okay with removing WAS A and linking it to a higher level to retain the position in a hierarchy.
JCA said he wanted only one solution, so he thought moving it to the next proximal concept in the hierarchy was a reasonable compromise. He asked what others thought. BGO said it sounded reasonable. There were no objections. JCA said it would have to be added to the Editorial Guide.
Decision: The AG decided to add to the Editorial Guide guidance on retirement of WAS-A relationships. The WAS-A should be removed and the target concept should be linked to the next proximal higher concept in the hierarchy.
- IHTSDO author responsible for updating the Editorial Guide to include the guidance about retirement of the WAS-A relationships by linking the target concept to the next proximal higher concept in the hierarchy.
Standard Time for AG Calls
The AG agreed to hold calls on the last Monday of each month from 18-19:30 UTC.
- Add to Confluence calendar and send calendar invitations to AG members for calls on the last Monday of every month, 18-19:30 UTC.
"Current X" Concepts
JCA introduced the topic (see screen shot).
GRE said it was similar to "present" findings. He suggested they be discussed together. The group agree to that.
- "Current X" and "Present" artifacts to be discussed together and dealt with at the same time.
Adjournment
JCA adjourned the meeting. It was one hour long.
