CIMI MTF Minutes 20130926

From CIMI
Jump to: navigation, search

CIMI Modeling Taskforce - Meeting Minutes

Thursday 26 Sept 2013 @ 20:00-22:00 UTC


Attendees

Linda Bird

Stan Huff

Harold Solbrig

Daniel Karlsson

Gerard Freriks

Joey Coyle

Patrick Langford

Dave Carlson

Rahil Qamar Siddiqui

Eithne Keelaghan - Secretary


Draft Agenda

  1. Review of Cambridge Meeting
  2. Review of SNOMED CT binding slides from Cambridge
  3. Other business

Detailed Meeting Minutes

Stan: Can add commentary and review slide set. My understanding of how take advantage of existing SNOMED... the reason we are doing the relationship binding rather than just value set bindings.

Stan (cont'd): OK - I'll jump in. Feel free to add or correct me. I'll walk through the agenda and say what happened there.


Review of Cambridge Meeting

FHIR and Graham Grieve

Stan (cont'd): We were joined by Grahame Grieve for one hour to talk about FHIR - had a good discussion about... the parts we are doing that is overlapping. FHIR resource... one... overlapping... One thing is, basically, analyze the FHIR models - the attributes that FHIR needs... To be able to represent FHIR's name for this attribute versus CIMI's name for the attribute... The real value is in defining profiles... What Grahame recognizes they need... The FHIR folks are engineers... Will look outside to find...

Action Item - produce FHIR profiles from CIMI models

Stan (cont'd): Action item - to try to produce, in some automated way, FHIR profiles from CIMI models. There is no forcing deadlines now... FHIR hasn't anything formal yet. Just went to ballot as DS2U and it will not be until January... until DS2U... In spite of that, lots of people out implementing... Good for the knowledge, but it means that people... will create things that are not part of FHIR yet. I thought it was a good meeting. Could be value that we can add to FHIR... Graham was kind to come to the meeting.

Stan (cont'd): That is my report about FHIR. Does anyone want to add to that?

Linda: What is the agreed plan... that you mentioned? 100 profiles from models...?

Stan: I... myself and Joey and Patrick to do that...

Linda: Models from InterMountain Health?

Stan: No - from CIMI. Patrick has generated 60 models from lab Mindmaps we created. We have 60 models. Can probably do a lot more than that. Patrick can say what the deficiencies are - what must be corrected. Can compile, but not stylistically or semantically correct yet...

Linda: ?

Stan: We may switch from CEML (?) to ADL rather than Mindmaps, but no... details yet and no date because no connectathon with FHRE that require models at any known time... But we did volunteer... would like to help with this.

Daniel: Work I have been doing... similar status to Patrick's work... and some gaps... not brought in constraints yet or... FHIR from UML model... but I am interested.

Stan: Yes. FHIR is well-spoken of everywhere. John Halamka at HL7 - said FHIR will replace CDA. Then Bob Dolin got up and said "Wait a minute"... But a lot of things people don't like about CDA. Lots have said that this is a lot better technology...

Stan (cont'd): FHIR may be important to IMH's future... May come into play in things we are trying to do at IMH with a new vendor, as well. Other comments?

[no response[

SMART team and Josh Mandel

Stan: The next topic - Josh Mandel from SMART team was there and talked about what they were doing. Their funding has run out... so trying to find other partners... to go forward. Their implementation of Standard API is... been using CDA(?)... Had discussions with FHIR to say - what if we changed payload to be FHIR? But no action items... The SMART guys did not say. An open door... But continue with open-standard API's with information models... not directly committed to using CIMI... but what garners the most from the marketplace... FHIR or... other payload.

Stan: So that is my brief report on CIMI and SMART.

Linda: Thanks, Stan.

Daniel: Question about FHIR. Different frameworks use different ways of composing models... Is there a similar... for FHIR for composing and aggregation?

Stan: FHIR profiling is filling same role in many ways as openEHR profiling. Can compose larger objects out of those pieces... Can make bigger structures out of parts... When do profiling you are constraining... to do major classes...

[Lost connection with Stan]

Daniel: the reason for asking is... FHIR... I've read is to simplify things... make easy-to-implement models.

Stan: Probably - the thing we would learn from creating 100 models is if there are problems. Joey - do you know about composition rules in FHIR? How to make bigger models out of smaller?

Joey: There is a generic thing - can make into anything within constraints. If doing labs, "results" is flexible, so can do... Like we model at IMH... similar to ADL (?), too.


Interactions with Other Groups

Stan: CIMI's interactions with other groups.

HL7- Support agreement with CIMI

Stan: Traded drafts of support agreement with HL7... charge $18,000/year to handle logistics. We are working on formalizing that agreement - seems reasonable. Looks like we will do a combination of people... benefactors of... and charge people per meeting... per meeting charge rather than find it in the budget...

IHTSDO

Stan: So - the IHTSDO agreement - David Markwell was there - he interpreted one of the agreements I sent him - no liability if models re misused... But I was saying - current contract does not specify liability (?) He was more comfortable with that. OK to have CIMI executive members sign agreement of CIMI members...

ISO Meeting in Sydney

Stan: I received a nice invitation from Steven K (?) to participate in discussion of ISO13606... will support my travel to ISO meetings in Sydney. This is a positive thing...

SemanticHealthNet

Stan: We've had the suggestion to hold joint meeting with CIMI and SemanticHealthNet ... a letter to Stefan (?)... about 2014 - joint CIMI technical meeting with SemanticHealthNet following... Both these are positive activities.

Health Service Platforms

Stan: Health Service Platform Consortium is continuing. Trying to put together a business entity to adopt a set of standards for interoperable services and applications... Difference being - not just specifying the standards, but to create a business entity so people can build applications and get certified and build marketplace for healthcare applications...

Gerard: Concurrent use in...?... where 13606 is... standard and two other standards... Health Service Platform Consortium... Existing platform that is being renewed... The other is... Continuity of Care.

Action Item: Gerard to send Stan pointers to Health Service Platform Consortium and Continuity of Care documents

Stan: Will you send me pointers to those documents?

Gerard: Yes. To have a business entity that takes care of... is an interesting idea.

Stan: Yes - the idea has generated a lot of interest. Would be provider-led, not vendor-led... Would be the providers, venture capitalists who want to...

Stan (cont'd): We worry about accepting too many people and then we won't get anything done. Also - a fine line - a lot are interested so can influence... Want it to be open, but not too big too fast.

Gerard: Yes - but in closed community can...

NCBO - National Center for BioMedical Ontology

Stan: Last thing - NCBO - National Center for Bio-Ontology... Funded by NIH ... Mark Musen and Ray and others at Stanford. Have BioPortal website... Can go browse lots of medical ontologies. They were interested in putting CIMI models in... Another front end - view a model and download it. Interesting... They are interested in AML/UML representations of model. They feel those using other ontologies will be familiar with UML ontologies and environment... Interested when we have AML environment... Interesting for you, Dave... Bindings into SNOMED and... Bio-portal light(?)... Semantic annotations to this... We are of the opinion that we need semantic annotations... SNOMED... next step... So SemanticHealthNet is very much interested...

Harold: Bioportal has fairly complete CST2 interface... but need to be cautious, especially in SNOMED area... Different from what is in IHTSDO...

Stan: Good to have you, Harold. Correct me if I am lying...

Harold: You don't often lie...


IEC Meeting/Phone call

Stan: We had an executive call this morning. Trying to set up so organization can set-up as benefactors and pay $1000-$1500, and others can participate by attending meeting and paying for meeting - about $250. A formal proposal will come out on that, so you will have a chance to correct.


URI's

Stan: So, Harold - how are you coming along with URI's? Have you had a chance to do this?

Harold: No - I got sidetracked.


Draft from CIMI - CIMI models, meetings

Stan: Next - you all should have received a draft... We added quite a few things for a list of CIMI models... Probably an agenda item for Modeling Task Force... to review...


Medication Terminologies - Discussion with Stephen Chu at Cambridge meeting

Stan: There was a prolonged discussion between Stephen Chu and me around medication terminologies. People have gotten comfortable with SNOMED and LOINC, but there is diversification. The same name drug can be formulated in different ways in different countries. So we need to say how we... in CIMI preferred... Probably can do this in a terminology meeting... Can say what is strategy and what is preferred way for interoperability. People in own country will do as they do because of what is mandated in their country, but we have to say what we will do for... interoperability.

Linda: I would love to be involved in those conversations because meds are of particular interest to me.

Gerard: ...Chinese... It is a work in progress...

Stan: Yes. I think they key thing - need to put up as agenda item in terminology meeting.

Linda: I suggest we look at international terminology model for IHTSDO.

Stan: Yes - May be the best solution. Other contenders are RxNorm. Pros and Cons against it as well.

Linda: I did a survey of...

Gerard: Is more than a drug. Also has to do with pharmaceutical industry and how things are packaged. Wider scope than IHTSDO...

Stan: I'll come back to that.


Portavita Web Site

Stan: So - had a nice discussion with Portavita about the website. First - informative information... meetings... minutes... Second - to access models. Browse and download. Things are progressing.

Stan(cont'd): Want to make a link to ... version of IMH website. Patrick and Joey have made version of our... 60 models... Gives people something to see and browse...

Stan (cont'd): Also - want links to other... openEHR... and multiple tabs. One tab would be the catalogue function - type a series of key words... would see... against metadata of model. If you want to see, then go to site and see and browse... Whether openEHR or other... So that looks close to putting up and making available in a number of weeks. Will be able to browse those 60 models... Also - a site that Sarah has been providing them... Also, goals and... So that was a positive interaction. Any questions?


Intellectual Property Issues

Stan: Next was a discussion of intellectual property issues. We said - we'd like to find an Apache or Creative Commons or other license we could use for CIMI. We tracked ideas - people showed what they knew.

Action Item: Nicholas and Stan to research IP licenses for CIMI

Stan: Nicholas and I will do homework to read these and see what might work best for CIMI. That would cover people's use of... in CIMI. Question - if created something new or a derivative of... the question - who is the owner of that intellectual property? An unresolved question...

AML

Stan: Next item - I gave a report about AML... Representative, with slides from Dave. I did a poor job of posing this, Dave. Maybe we can get you to step through these, Dave...

Dave: Thanks for showing them, Stan. Sorry I missed it.

Project Manager for AML Document for OMG

Stan: On AML... We - I mean Dave has done nice work around tooling, solving... But we need a document to submit to OMG as standard. What came up - we would like to have a project manager. I asked for this and a woman from Accenture volunteered to do that. I thought I sent her your name, Harold. She is ready to work now, and I described to her how we would like to create the document... and someone to track who is responsible for issues... and work with people to get the part of the document done... So - very excited to have someone to do that.


Action Item: Harold to talk next week with volunteer project manager for AML

Harold: We will... talk next week.


Catherine Hoang at VA - Dave back on contract

Stan: Another thing - I bumped into Catherine Hoang... and she said "I worked for 2 weeks to get Dave back on contract". So she is focused on it and knows what is needed... and I stressed how important it is to get the AML stuff done...

Dave Carlson: Catherine is program manager over at the VA... She came into that after managing... Healthcare Standards over VA. Also - VA is once again a... member. So should allow VA to be formally a co-submitter of AML...

Stan: Good news. Hopefully... It would be helpful to have a project manager... and part of your work is related to that... It will be a good thing.


Michael and Syntax Instances

Stan: ... syntax instances... Michael is doing good work on that...

Linda: Different types of representation... Might...

[I missed a little here]

Meetings in 2014

Stan: Next - discuss how CIMI wants to meet in 2014. 2 general meetings and 4 technical meeting. Two of the general meetings would be coincident with the technical meetings... May meeting of HL7... and other... in association with IHTSDO meeting. In 2014... will be in Amsterdam in October. In the first of the year and in summer... will be technical. In 2014 - joint meeting with SemanticHealthNet... and summertime. Would be a meeting - like to hold in Australia or Singapore or... Need to send out as email to folks in that area to see if international meeting... Want to set 2014 meetings ahead of time so people can plan so can add days for CIMI meetings with other meetings. So that is my meetings report.


SNOMED CT Binding Strategy for CIMI Models - from CIMI Cambridge Meeting - Sept 2013

Stan: So I would like to go through the slides we went through at the meeting. Good idea to explain... Dave has been one of the principle strategists and technical wizards in SNOMED. Two years ago, he became part of the IHTSDO staff. He was the staff lead for innovation and implementation. Has continued to do the same work, but... The idea is - innovation should happen everywhere. Dave is a great resource to have if you can get him at the meeting.

Slide #1 - SNOMED CT Binding Strategy for CIMI Models

Slide #2 - CIMI Use of SNOMED CT

Stan: So - the ways we are using SNOMED CT... Using actual content in SNOMED to do translation of pre-coordinated model content to post-coordinated model content...

Daniel: I would like to question that - can discuss after...

Stan: OK.

Slide #3 Selected CIMI Policies, Decisions and Milestones

Slide #4 Terminology

Stan: [reads slide #4] SNOMED CT is the primary reference terminology... LOINC is also approved as a reference terminology... CIMI will propose extensions to the reference terminologies when needed concepts do not exist... CIMI Has obtained a SNOMED extension identifier...

Stan (cont'd): As Dave talked about - "public good" use agreement for use of SNOMED CT in CIMI models. Dave said that everyone would be able to do that and not have to have a "public good" use agreement to do this... But people who made... for production use would need to have... from country or IHTSDO.

Stan (cont'd): And CIMI would create a Terminology Authority to review and submit concepts to IHTSDO as appropriate. [see slide #4]

Slide #5 Isosemantic Models

Stan: A simple example around Hematocrits.

slide #6 Isosemantic Problems

Stan: I went to a different model... Similar to yours, Linda. The idea that you could have people entering on a problem list... and then post-coordinated model... So would make association between models... part of same family... And then... broken out... And make same thing implicit... About patient... So could say - this is about patient - is implied by context.

Gerard: I see post-coordinated model - one of the options... But preferred model is what we talk about with semantic interoperability.

Stan: Yes, and we said, as a policy, we prefer the post-coordinated...

Gerard: Same as I advocate in SIAMS.

Stan: Good. And next slides...

slide #7 - Isosemantic Models

slide #8 - Semantic Interoperability

Stan: Statement from CIMI Executive Committee... [reads slide #8]

Stan: These are things we have already said. Then I went into this example...

slide #9 Isosemantic Problems

Stan: One thing we are trying to enable...

[lost Stan's connection]

Stan: So this is from SNOMED...

slide #10 - From SNOMED CT Relationships

Stan: Suspected Breast Cancer is a finding of breast... and if you look at malignant tumor of breast, it breaks down to... [see slide] So this is compatible... from SNOMED. So at modeling level, have pre-coordinated findings, and in post-coordinated - the actual...

slide #11 Model with relationship bindings

slide #12 Model with relationship bindings

Stan: So in particular instances...

slide #13 From SNOMED CT Relationships

slide #14 Model with relationship bindings

Stan: ...where subject-relationship context is subject-of-record.

Stan: So - two things. We should look at SNOMED and understand the relationships. Could point to nodes we want. The relationships we... SNOMED-compliant... Take the best... So - I'll stop here. Daniel?

Daniel: Go to slide #2. One reasonable interpretation is - could use a [CIMI?] model for... That is fine. But IHTSDO has different constraints than CIMI. Models should be compatible... So is limited expressiveness... Don't want that limitation for CIMI models. CIMI will always have...(?) that can never be expressed in SNOMED CT.

Stan: I think we are in agreement... This example focused on the SNOMED semantic part of this, but also... when entered... What is association to other problems? All of that is part of this, but not from SNOMED.


Daniel - Presentation

Daniel: I posted a link - a Google presentation - but if you could [bring it up]...

Google Docs Presentation

Daniel: You see CIMI Model in gray and corresponding... SNOMED CT attributes and value sets in two columns on right-hand-side of picture. You can map... to body location side to laterality... This could be done using the CIMI attributes to SNOMED attribute... if structure is the same.

[Next slide]

Daniel: We have an alternative... not mapping... is wrong... Cannot have laterality on...

[Next slide]

Daniel: Has to be another layer - the constructor binding - define some external resource that does the mapping... Not use the CIMI attribute to SNOMED mapping... This way there is no need to have any similarity between CIMI structure and any structure in concept model... so have a way to...

Daniel (cont'd): So - having this kind of constructive binding... so if not always want to conform to...

Stan: I disconnected a second... I understand the first 2 slides... Here - still using...?

Daniel: Not using... mapping from... But this is not used to determine the semantics...

Stan: Some kind of knowledge represented in a...

[Lost connection to Stan]

Daniel: It is a ... I think I borrowed from something David Markwel did 5-10 years ago... The current node, which is... and go from current nodes to... CIMI path... of this... The path in information structure and pick up value... and do the same with laterality. Free to...

Daniel (cont'd): Main point - this seems to be a requirement for doing this kind of comparing ISO semantic models with different levels of pre and post-coordination.

Stan: I like it. I intended exactly what you described here. That's great.

Gerard: Is what you propose an analog to SemanticHealthNet...?

Daniel: I think there has been one meeting in SemanticHealthNet... and ways to realize - this has not been discussed. This goes beyond what has been presented by Stefan and (?) in the...

Gerard: I agree. It goes beyond, but an analogy.

Daniel: I don't think I agree, but could take a lot of explaining.

Stan: Other questions or discussion?

[No response]

Dave Carlson - AML and Model Transformation Tools

Stan: Dave - we have 30 minutes for what you have working with AML and model transformation tools.

Dave: Yes - if you have slides.

Dave's slides #1

Dave: This first slide. The goal of AML is not only to provide a guide to UML types but also other model formalisms. A top-level goal... Submit OMG standard for AML and be able to get a draft to OMG in December. And a question for Harold - was there... time frames?

Harold: No.

Dave: OK - by December meeting. And ISO-11179-3 MDR (?)... The binding of meaning and value sets to model of meaning.

Harold: I'm sorry - we submitted a motion that deadline would be Feb 24th.

Dave: So - Munich meeting for delivery of draft?

Harold: Yes.

Dave: I focused the presentation on... Reasonable first draft of UML profile... Rather than documentary... discovery... Certainly not 100% coverage of what is expressible in ADL, I put emphasis on... runtime models in JAVA.

Dave: I made quite a lot of progress of... that Harold made and... So from that, I cleaned up a few more things, Harold merged a couple of primitives. I tool UML and generated... and a full JAVA API and... Run through a JAVA library... The first step towards building...

Dave: So - then given some CEML-models from IMH, Joey sent to me. I imported about 200 models... Hematocrit - derived from standard lab observations, and... I also imported into UML model... that includes the Blood Pressure panel...

Dave (cont'd): So I have all associated types. Getting structural aspects in place... Derived from CIMI model...

Dave's next slide.

Dave: In order to create archetypes for... model, best to create on... for ... pattern. So we looked at - take pattern and create archetypes. First prototype approach - create new reference model for clinical element, import CEML models as archetypes on this reference model. Important to... AML not tied to any reference model. So I prototyped both approaches...

Dave (cont'd): Issues I came up with I keep coming back to container entry problem. A panel can contain... but also can contain modifiers and qualifiers as well as a... and statements. I have not come up with a way to... An obstacle...

Dave's Next Slide.

Dave: Screen shot of tooling we have been using... The model for the CEM pattern archetypes... An archetype for... to statement... for CIMI entry. I have elements or attributes within this... Part of the AML modeling style... You can see...each of these is a subset of... Has a relationship of 0 or more items contained in an entry... So instead of archetypes that collectively represents the pattern of models...

Dave's next slide.

Dave: So using... CEM Models (from CEML)... Import several 100 archetypes. Work in progress pending... Parse convert CEML constraints, convert to OCL(?) and/or AML representation.

Dave: So constraints working to how to convert to... Also the IMH models use vocabulary key... Many of those can map to SNOMED CT... So - we could import... The second point in this - finish mapping the CEM data types to CIMI data types...

Dave's next slide - a screen shot.

Dave: Having imported from CEM/UML to... BP panel shows... Body position... Systolic BP measurement... Each of the relationships refers to another... BPs derived from another cluster... Each of these archetypes is derived from one of the CEM on... And body position restricts the components.

Dave (cont'd): Pretty good job of bringing in structural... Can do good job of... derived archetypes... Guides you towards...

Dave's next slide.

Dave: From proof of concept - can take models and using ecor to JAVA, can generate... completely JAVA library... Can see JAVA class called VitalSignsPanel and Blood Pressure Panel... VitalSignsPanel has BloodPressurePanel... So can programmatically create instances of... creating instances of CIMI model... within Eclipse modeling frame one can customize...

Dave's next slide - XML serialization to...

Dave: I programmatically construct... export to XML-based on CIMI reference model. I look at tag names - these are CIMI model tag names... serialization of CIMI model...

Dave's next slide - Conclusions.

Dave: Full-model driven life-cycle. CIMI model, itself... life cycle through... TBD - Generate FHIR profiles from archetypes. Stan - you said Patrick is generating ADL(?) from lab models... I don't know if you... If we can agree on use of CEM pattern archetypes... CEM to UML...

Patrick: I have just worked from Mindmaps.

Dave: OK - if we have correct transformations... still question of inserting those... for FHIR, interesting question - what the CEM...

Patrick: CML to FHIR - we need to talk to Joey about this. He had an epiphany about this.

Joey: Yes - the way I do my other trans(?)... I could convert CML to CIMI models. But no direct to CIMI...(?)

Stan: Yes - would want to adhere to CIMI reference model... So other patterns and data types would be enforced.

Dave: So - that is it. Gives me confidence of how to implement CIMI model...

Stan: That is great. What we want to do in AML proposal. Tell me - we talked about 11179(?). Did you do terminology binding in this?

[Dave shows slide] - Body laterality; location; Screen shot of BloodPressurePanel archetype (UML)

Dave: So what I did - I took the key and I created a class... Signed that as the meaning... stubbed out the concept class... Binding the meaning into the model... When there is a constraint for a value set, I would want to do the same approach... Enumeration of ... stereotypes. Create stubs of these and apply the value of that...

Stan: Other questions?

[No Response]

[end-of-meeting]