CIMI MTF Minutes 20140206

From CIMI
Jump to: navigation, search

CIMI Modeling Taskforce - Meeting Minutes

Thursday 06 Feb 2014 @ 20:00-22:00 UTC

Slides with these minutes

Attendees

  • Stan Huff
  • Steve Hufnagel
  • Harold Solbrig
  • Deepak Sharma
  • Jay Lyle
  • Patrick Langford
  • Daniel Karlsson
  • Sarah Ryan
  • Michael van der Zel
  • Joey Coyle
  • Eithne Keelaghan

Draft Agenda

  1. Joint meeting with SemanticHealthNet March 13-15 in Brussels
  2. Update on AML and current questions and issues - Harold
  3. Panel modelling examples - Thomas (if available), Stan

Detailed Meeting Minutes

An Update on Archetype Modeling Language (AML) - Harold

Harold: This is an update on archetype modeling language (AML).

[slide #1] Title Slide

[slide #2] Archetype Modeling Language

[slide #3] Archetype Modeling Language

Harold: The number-1 bit is what the RFP is all about... To come up with a set of profiles... to represent archetypes... So can use all of the various artifacts to test code... It turns out that it is working out exactly the way we put it.

[slide #4]

[slide #5] - Archetype Modeling Language

[slide #6] - Archetype Modeling Language

[slide #7] - Archetype Modeling Language

Harold: The first one is the... profile... When build Ref Model, the first is a declaration of what we will recognize in Ref Model and what we won't. We won't recognize static diagrams... Let's say that... can have... and say 0-1 in OCL. And unless the... recognizes OCL... So constraint model needs to meet... Gets down to constraining integers... So important that Ref Model and UML have a good understanding of those... Then recent changes that Tom made to ADL 1.5.

[slide #8] - AML Specification Sections

Harold: So, 3 large chunks. The first is the Metamodel... [name-of-someone?] encouraged this... document... exactly what you are trying to do. Second - Profile. Third - XMI.

[slide #9] - Approach

Harold: Five Meta-Models. First one I italicize... Metamodel... The types and constructs to be constrained by... Not good if, for example, you create a... called complex numbers, but ADL not know anything about this... not good. This applies to strings... and dates and times... and core constructs.

[Slide #10] - Approach

Harold: Constraint Model - how one addresses paths and names... The third - Terminology Binding Model. Thomas has done useful work on ADL 1.5. ...We have a normalized representation... archetype... or SNOMED CT. Numerated... using remote terminology...


[slide #11] - Approach

Harold: We talked about this some earlier. ADL defines a core set of... logic... made assertions about... can see in OCL... This is looking to be a task unto itself. Maybe, for the purpose of ADL...

Michael: What are the rules for, exactly?

Harold: Good question. Let me pull up in ADL. The ADL example - does not work for a whole lot of reasons...

[Harold pulls up ADL]

Harold: The other thing.... is... ADL states that it is side-effect free. But the examples so far are predicates on... So, question is whether BMI belongs in AML at all.

Jay: There were also clinical guidelines like Blood Pressure.

Harold: Good point. There are some minimal things, although not know about total validity of that.

Jay: I don't know how frequent that is...

Harold: We have the rules meta-model. For initial submission we are going to leave the meta-model ...

Harold: And the 5th bit... ADL itself carries a huge amount of information. Who, what, where... Not archetype-specific. Dave Carlson did useful work there. Alignment with 11179... We are going to leave that one out... Going to show where the metadata takes off.

[slide #12] - Initial Submission

[slide #13] - Approach

Harold: AOM 1.5 is still a moving target. .. Heisenberg... The question and issues get resolved... Some by Thomas...

Harold (cont'd): What we are producing does not look a whole lot like AOM 1.5. We're giving you methods. Some were to... OCL... We are not modeling a lot of OCL. The Meta-model will have a minimum amount... There is a serious mismatch... There are places where you have very few classes with lots of attributes... and if doing as a choice, you don't fill this out... Take advantage of ordered lists... difficult to explain... tried to do this in pictures... Then some of the work fed back into... Have remained relatively faithful to AOM... If glitches, Thomas fixed.[slide #14] - Approach

Harold: Showing as instances of the Meta-model... So - an example out of document. This is how we would represent as instances of a Meta-model... If we can represent consistent representation of all in ADL, then can be sure is complete. Also, if... then can get rid of... We are in the process of turning... into... I am sure Jay is frustrated. Interesting issues... creating the formal profiles.

[slide #15] - Approach

Harold: And next... take some set of examples... represent as profile... Tells us what we have to do if want a round-trip with ADL. ADL path structure has to be converted and...

[slide #16] - Submission

Harold: Submission being done in RSA. Doing a good job... Text generated using BIRT.

RFP Submission Timeline

[slide #17] - Submission Timeline

Targeting May16 for Final RFP Submission

Harold: Confident will be good enough for initial submission... And... people doing... tools... will walk us through... We are targeting May 16th for final submission for vote in June. And issued as Beta... The rules for Beta-submission... Can't substantively change the spec... It is a good rule.

RFP Submission Team

[slide #18] - Submission Team

Harold: I have done Meta-model writing... Angellique Cortez.... and others.

[slide #19] - Issues and Resolution

Harold: Some of the issues and resolutions. Archetype identifiers... There are a couple of places where this is turning out to be useful. No substantive difference between ... and SNOMED CT. Both terminological resources... Acquire definitions and understand how they map from one terminology to another... Paths are... a real challenge.

[Shows openEHR screen]

Harold: When you constrain, changing name of top-most node... constrain on element id-13...

[Shows another screen in openEHR]

Harold: Over on left-hand-side - the laboratory... has complex constraint observation... is the path... and the constraint... The lab thyroid archetype... The first constraint... Then in the middle we have the path. If introduce... then we can replicate the path... So we are saying that the one with node 13.2 is... Interesting - had not added for the constraints.

Harold (cont'd): What makes it work is that this node-id is the same as this node-id... Have not added a new identifier... So this is just a path... no constraints. So - all archetype constraints have an optional... can reference...

Harold (cont'd): Have not... the way to provide values... The default value... What constitutes... in a particular... May not get this back in the...

[slide #20] - Issues and Resolution

Harold: OCL vs ADL Assertions... separate profile... Regular expressions... part of... What we decided to do... if I move to Ref Model profile.

[Harold shows another image on screen]

Harold: If include... that gives people using it the flexibility... If want to be in CIMI world, then have to... So assume further constraint on Thyroid-function test. I can constrain... Don't have modeled yet...

Harold (cont'd): Not going to talk about identifiers in the model...If 13.1, then parent would be... One of the areas that is an issue... The way Dave Carlson did terminology profile, he separated... the class names... attribute names... Would all be translatable... Interesting... The things that can't be translated... No matter what you speak... But constraints can... But what is missing in Meta-model - the attributes defense... I am interested in what the group thinks. So if pulled UML down... archetype... The class name might be in Spanish, but... A good thing? A bad thing? No terminology identifier that allows translation.

Stan: I think this is an area where... I would put out what you explained, and get some real experience with it.

Harold: I think it will be easier to use what Dave has done already... So non-archetypes... The problem with changing the names, might not understand...

Michael: I think the model should be in one language. Or you have something like Microsoft has... which is impossible to read.

Harold: I understand conceptually, but you are one of the ones who... You show up all the time with models in Dutch. Not confusing because models don't change, but all in Dutch so I don't understand. I don't think huge difference between class in... and class in Ref Model and archetype constraint. Different functions, but... So - I think I will take Stan's advice and... I wish Tom was on the call. He uses Body Mass Index (BMI)... Calculations have bits of code which have... Seems to me that BMI has... outside the scope.

Stan: Yes - I thought - what is the principle... Rule vs. constraint... and is in the name. Constraint ways that I have semantic space that... Going to make semantic space smaller in a child... And what Jay... stating dependencies... start-date before end-date... Systolic higher than Diastolic... Parent can be 0-100, child between 0-30...

Stan: Flavors... pure constraint... Co-dependencies... and BMI - calculate a new thing. I can argue that co-dependencies and calculation of independent variable...

Harold: So - if you did it, would be a separate use-case... which is different...

Stan: Yes.

Harold: Both came with full-fledged UML models. And both OCL and archetype constraint model - strange to deal with... whether pure trees in UML or...

Harold (cont'd): I want to mention the slot-notation. Our intent is to change the archetype-slot. Will reference a value-set... Our assertion will be... if includes... If in that value set, can be plugged in... Going to move away from assertion-grammar... Thomas... using strings.

Harold (cont'd): One other thing. This is a sample of output. Deepak.. majority of submissions will come out of RSA...

Questions about Harold's presentation - Meta-data

Stan: Great. So - back to start where you were going through the 5 things... Approach... So the 5th... What was not clear to me - is this Meta-data on the models you are creating or on the instances?

Harold: Our submission, we stop at... Authored Resource. All around this we put into #5. While important, it is not going to break...

Stan: So, additional question... As far as CIMI is concerned, some of these elements will be required for our processes to work. And not putting on here, so a... And can we agree in CIMI what...?

Harold: If you can come up with a few essential elements quickly...

Stan: When created...

Harold: Yes. When you get into a rich... as this, it is going to take a lot of time to do and will be an interest in generalized... OMG... 11179... Meta-data... Wrap all this Meta-data around. A couple of options - if minimal set, then we can include.

Harold (cont'd): Another solution - Tag value pairs. Then... for reporting ADL, same as reporting CIMI...

Stan: Yes.

Harold: Hard part is... time.

Stan: This is amazing work from my perspective. Appreciate everyone - especially you, Harold...

Harold: AOM and ADL - I have learned lots. And Thomas has given invaluable advice.

Stan: Other questions?

[No Response]

Email Harold if want to look at AML

Harold: Available to all with OHT logon. But is in RSA format. If you want to take a peek, email me. In a week will have first draft document.

Steve H: And the slides?

Harold: Who takes minutes?

Michael: Eithne. You can send them to me... or to Eithne.

Harold: I have to go.

Representing Panels - Options narrowed down to 2 Main Options - Thomas's examples

Stan: the other thing. Two weeks ago we had narrowed the options for representing panels to 2 main options. And Thomas offered to show... and make examples to show how it would work. He did not get done for last week, but almost for this week. And Thomas was not able to be on the call today, but sent examples. And Michael - you can be a big help. So - we can all go through them together. You should have received an email. Can you read it?

Response: Yes.

Stan: So - the Ref Model Michael put up is not legal, although EA did not complain.

[Thomas' email and picture in email]

[Stan reads Thomas' email and shows picture]

Stan: So, Michael - does this make sense?

Michael: I was struggling with red lines. I though what Linda did was correct. But what he does is more than splitting... So I am not sure yet. Especially the top red line - implies that you can get anything into entries. I did not think that was the intention.

Stan: I think you are right. Not the intention. It looked correct to me. Distinguish compound from individual entries. Individual entry - only item and...

Michael: I couldn’t figure out what is UML-legal.

Stan: Maybe we said as much as we can on this call about that. Understand what Thomas intended. Making... a child of core-locatable does things we did not intend.

Michael: Entry has a data-link to item. But we give compound entry also a link to entry. So compound... Not UML-entry. But I don't think what Thomas did with red lines is a solution.

Stan: OK.

Michael: Is Linda on line?

Stan: No - is in Copenhagen getting acclimated to new job with IHTSDO.

Michael: OK. I will look at it.

Stan: And next thing... I guess that part of the tree is right. You have individual entry and compound entry. So I am not sure he did what his red line showed. Not show composition a part of... Not know if I am qualified to say anything about this. Others have comments?

Stan (cont'd): And his last example shows archetypes built on this Ref Model. The examples are not showing what I thought... Is this saying that body height and body weight are individual entries? And body weight down here... Have a compound entry with a body weight. So - I apologize. I don't know enough to comment on what Thomas said. And we have a lot of static coming from someone...

Stan (cont'd): OK - well, I think we are done. I don't know enough to explain what Thomas sent. Unless some of you can explain...

Patrick: There are examples from the diagram, but they do not really show the possibilities. So your example is a possibility, but not being shown here. So not all of the possibilities... Can put... down as entry, but have to do and see if workbench has trouble.

Stan: OK. Anything else?

[No response]

Stan: OK. Next week, we'll have more examples, and Michael will have had a chance to talk to Thomas. And we will be able to say the representation of panels in CIMI... the preferred style.

[end-of-meeting]