CIMI MTF Minutes 20140515
- 1 CIMI Modeling Taskforce - Meeting Minutes
- 1.1 Attendees
- 1.2 Draft Agenda
- 1.3 Detailed Meeting Minutes
- 1.3.1 One Model as 'Style of Model' for a Type of Data
- 1.3.2 A process in place to Create Reference Archetypes
- 1.3.3 Build Library of Transformations
- 1.3.4 Need to create some Reference Archetypes and put into a Repository
- 1.3.5 The Key is to Agree how Forward Generation happens from CEMs (Clinical Element Models)
- 1.3.6 Recommendations from IHTSDO on Terminology Binding Strategy - Linda Bird
- 1.3.7 Goal -
- 1.3.8 to start to Generate CIMI Profiles
CIMI Modeling Taskforce - Meeting Minutes
- Deepak Sharma
- Stan Huff
- Linda Bird
- Harold Solbrig
- Mark Shafarman
- Thomas Beale
- Ian McNicoll
- Daniel Karlsson
- Jay Lyle
- Patrick Langford
- Eithne Keelaghan
- Any final discussion of the "minimalist" RM proposal
- Next steps after voting
- Create new BMM file
- Create reference archetypes
- Recommendations from IHTSDO on terminology binding strategy - Linda Bird
- Any other business
Detailed Meeting Minutes
Stan: I sent out the draft proposal. I did not check with Linda - you can refuse if you are not ready.
Linda: I can say a couple of words.
Stan: OK. Minimalist approach that I sent out... optimistic that it will pass. Then next steps... creating BMM and reference archetypes... then models and next big item - finalize term model structures and strategy. So depending on what Linda has to say today, we can plan today and then talk in earnest next time. Anything else?
Stan: I put the short description in the email. The Reference Model will have data types and constructor types. Then we would move to [?CIMI Ref Model?]... The entry style of creating models used in openEHR and 13606... And with option 2, create other reference archetypes... Those would be the parent types to family types that conform to... or other types in IMH.
One Model as 'Style of Model' for a Type of Data
Stan (cont'd): A given model would adhere to the core of one of those ref archetypes. So would be wrong to create panels like openEHR and IMH. So per our previous agreements, we would choose... We would prefer one model as style of model for that type of data.
Stan(cont'd): So that is my brief summary. Discussion? Counter point?
Linda: So for each type of clinical data, could we choose different reference archetype, or would there then be one preferred model?
Stan: One preferred model. One preference for this style of numeric... or all derive from same archetype... or allow us to create models that conform to... Would allow us to make automated ways of mapping...
Linda: OK. Thank you.
Thomas: If we go with minimalist model and don't add context-related properties, then... is it worth putting a standardized set... like property type and... locateable... is it worth putting them in the Ref Model? Because if it is totally ad-hoc in the Reference type...
Stan: Right. I don't think the creation of Reference archetypes... ad hoc... I think we would have discussions. But I could go either way. Is a matter of copying and... making consistent across archetypes... But if we thought there were going to be a lot, then putting in Reference Model would probably...
Thomas: I think... doing it the way you are saying is the way to go.
A process in place to Create Reference Archetypes
Stan: Yes - I think it is controlled development, not ad hoc. So would come together to define Reference archetypes. So would be a process in place to create Reference archetypes.
Stan: So Harold - we talked about all this when you were occupied, so...
Mark: One of the models we were having difficulty with on first type of model were blood transfusions - when have blood of patient from another person who might not be identified, and have family history and... So would this be a special case?
Stan: In openEHR style of modeling, you can model a transfusion reaction. And... would have to end up in 2 different entries. Would end up in composition... But both IMH and openEHR have a way to...
Mark: So have certain types of transforms from one type to another.
Stan: Yes. And one style would be CIMI-preferred style and other would be local. So if entry style was CIMI-preferred, then IMH-style would use translation to get from one to another.
Build Library of Transformations
Mark: So may build library of transformation for those?
Stan: Yes. Want to help implementers... Coming from local to...
Harold: Has someone put together what this looks like in UML? ... Is Section gone?
Stan: We have to look at that. I didn't think we would monkey with participation and... And I don't know about Sections.
Thomas: Like a... [?]
Linda: My understanding... would be Cluster element and data element and data type. I did not think participation was going to be there.
Thomas: I probably got rid of... got rid of Composition and... Anything in Reference Model - easier to work with... unless have to put in Ref Model. So unless different types of participation, I would leave it...
Daniel: I don't think you can leave participation because are on entries and entries are not in minimalist model.
Linda: Yes - participation was moved to being more generic.
Thomas: On... model.
Harold: Where is that?
Thomas: Workbench... BMM file. Core Locateable... Has participation. And... So Linda is correct. I think Entry can be thrown out with relative impunity.
Stan: Yes. I thought we were going to leave those other parts, like participation, alone... The earlier in the model that we can define it, the better...
Need to create some Reference Archetypes and put into a Repository
Thomas: Probably Ian and I would need to do the thing we need for openEHR... So, Stan... you and Joey and Patrick... if you want to ask me... I would do the cutdown BMM, if all agree. We need to create some Reference archetypes... rough draft... and put into a repository. So I'll put a BMM in the CIMI Github Repository. Ian and I will create some reference archeytpes for openEHR. Maybe we can publish them and have a call about this... So you probably know for IMH...
Jay: In adopting the archetype-driven approach, we seem to be going around... rigor of compositional styles...
Thomas: I think we are. If we on this call all had our way, then I think we could... But we haven't managed to do that... But even if we did, then would be hard to get rest of world...
Jay: Just asking if that is what we are doing.
Stan: Yes. As Linda said eloquently, we are moving the dead horse...
Linda: A Hungarian expression.
Stan: So - the votes may be contentious, but... remarks... model following openEHR or IMH style... We would know in either of those styles... So vote comes down to... just have to pick one. We have talked about... Would talk about concrete... But would still be relatively contentious vote, but would be concrete types... style of model they are part of...
Thomas: The openEHR style... we would try to make that... 13606... Would make improvements on openEHR and 13606.
Thomas (cont'd): Secondly - the way we would do labs... I have to talk to Ian, but probably be more like IMH-style... So the difference from the way we currently do it... serum sodium or other... we would probably do machine conversion from [?] of lab analytes to... of reference archetype. Now you will say - we will do in IMH...
Ian: We are definitely going to be much closer to IMH style for... this is a good point to get patterns closer to what you do there.
Thomas: Will be clinical-group mentality. Will look at ours and see Hematocrit... and yours and see data...
Stan: I think all of this debate will be helped with concrete-proposed models... level... so I hope there are things we can do in style and... with better understanding from both sides on how can be done. But back to your comment, Jay, we've moved dead horse to different street...
Harold: Whether model should be ecumenical model... or the CIMI model... So we are voting on ecumenical and then can come up with stronger CIMI model... So I see this as absolutely necessary now that I have listened to it.
Stan: I like the positive spin on it. Other questions?
Stan: OK. So, Thomas - say what you will do.
The Key is to Agree how Forward Generation happens from CEMs (Clinical Element Models)
Thomas: Ian and I will knock up some partial Reference archetypes for 13606 and openEHR... Need others to look at this... If Patrick can... So get those built might take discussion... or magic... They key is to then agree on how forward generation happens from CEMs and... Have to be specialized archetypes from those Reference archetypes... So every type of archetype is a level 1 archetype... One generation down...
Harold: We'll find out, but... We may have to revert to one level down the line, but may make this easier...
Thomas: Yes - Joey and Patrick... may have a way to make it easy... I'll have to write a bit of code and see if do-able or not ugly... So every archetype going forward is a child of the Ref archetype.
Action Item: Thomas to make a new BMM and 1 Archetype
Stan: Is that something you could get done before next call? A new BMM and 1 archetype?
Thomas: Yes - I think.
Harold: Do you have access to Enterprise?
Thomas: I'll need Michael or you to...
Harold: I'd prefer Michael, but...
Action Item: Email Michael for help and to get rid of Classes
Stan: Michael is interested in doing that. He is anxious to help. We can ask him... to see if he has time.
Thomas: So we can email him... to get rid of classes... Should be a 15minutes operation.
Stan: Yes - so we will do that. OK - Anything else about next steps on Ref model?
Thomas: If there is, I will put on the list...
Recommendations from IHTSDO on Terminology Binding Strategy - Linda Bird
Stan: OK - Linda - so we go to you...
Linda: I don't have too much to say. You may be aware that last year there was an implementation... providers' scheme and Rahil and ... and I worked on. So we are moving that forward. So terminology for binding is... So define languages for query... Used for terminology-binding. Will be based on (?)-syntax...
Linda (cont'd): The other ting - Costina from Denmark is working on the... and on an introductory to... A Use-case specific guide to terminology-binding. So is a range of styles for terminology-binding. So if your objective is to manage [?], then simple approach to binding. But if you want semantic approach like CIMI, then need to take advantage of... So make simple for simple cases and also the IHTSDO-approach. Also looking at [?] and concept-model connect expressions. An example is... If we have an observation... Terminology-bindings to observation, we have... attributes and... We are saying that this is OK to create the model how you see fit.
Linda (cont'd): But then... in those areas where you want pre-coordination... So this is in progress. If face-to-face in June, then I can present at that... Can define the IHTSDO approach so it is compatible with what CIMI is doing.
Stan: Thank you. And - Yes - we would like to work together on this. I have seen draft on Expression-constraint syntax... Can we share this?
Linda: Yes - I can send the URL around. It is undergoing a review. Comments from CIMI would be good.
Stan: That would be great. We can have homework for subsequent discussion.
Stan: Are there other documents that would be relevant?
Linda: I am working on presentation... end of June I will present internally... Then I can show...
Stan: Thank you for that. Other questions or comments for Linda?
Stan: OK. Any other business for today?
to start to Generate CIMI Profiles
Stan: Alright. I look forward to the next step. We are making good progress. My goal is to get some good CIMI content out there... and start to generate CIMI profiles...
Harold: So what is the deadline on the vote?
Stan: Next Wednesday. I set that so we will have the decision by next week's conference call.
Stan: OK - we stand adjourned.