CIMI MTF Minutes 20141009
CIMI Modeling Taskforce - Meeting Minutes
- Stan Huff
- Linda Bird
- Virginia Riehl
- Deepak Sharma
- Patrick Langford
- Sarah Ryan
- Daniel Karlsson
- Thomas Beale
- Joey Coyle
- Gerard Freriks
- Harold Solbrig
- Steve Harris
- Eithne Keelaghan
- Draft agenda for Amsterdam meeting
- Progress report on work items from Oakland
- Models with changes as agreed to in Oakland - Patrick
- Creating and preserving unique identifiers for CIMI extension terms - Harold
- Review of example instance data (see attached file )- Linda
- New business
Detailed Meeting Minutes
Stan: The agenda for Amsterdam - want to review. A review of how things have been going since Oakland. And sample instance data question. Other?
Amsterdam - Draft Proposal
Stan: We have a very rough draft of proposed agenda for Amsterdam... Virginia is projecting. You may not have agreed to do [what is on] this, but you can object.
Stan: The first day... organization and governance and overview for those interested... like Jamie Ferguson and... others. But Virginia has been helping us. For the most part, we have put the content in here with an idea of how much time it would take. And, Linda - we put the time in here but do not know when you are available. So, to start, welcome, orientation... Then discussion of CIMI as part of IHTSDO... 1 ½ hours. Then lunch. Then we are thinking of 2015 plan... It would probably be... how do we build the model content... How build... laboratory... What is our process.? And then tooling approach and plans.
Stan (cont'd): Some of you have been contacted. Figure out tooling and what funding needed for tooling... Draft document I have shared with Harold and Linda. Feedback.... my plan would be to present to Don Sweet... And get input from Don and discussions before Amsterdam about the approach to tooling. So... the meeting... in the context. We are in the very early stage. And we could be sensitive. May be strong feelings on SNOMED side... regards to tooling.... I may not know about... So I caution this group... in the early stage.... but nothing is agreed to yet, especially on SNOMED side. So don't want to disturb sensitivities, especially on SNOMED side.
Stan (cont'd): Tom has been sensitive to... ADL workbench. The clinician bench... Browsing models... and managing curation... The open terminology server. Important question - in our tooling - things available... Don't know if OCEAN or OpenEHR, but terminology server... Then - joint activities with HSPC and SMART, FHIR, HL7-brand associated with FHIR... discussion I had with the VA around SOLOR... More discussion... Daniel?
Daniel: That is the plan - to have something to present at the Amsterdam meeting.
Stan: Then progress on MTF... review for approval... The decisions we have been making in this task force... And update on Harold... and Sarah... and... ? Everett?
Linda: If you can give me notes on what you want me to talk about... MTF propositions. And tooling... have to figure out how fits with RTI for terminology server from...?
Stan: Would be great. If they have a process going on there, we might... IHTSDO... whatever they have for their solution...
Linda: Yes. The preferred option by IHTSDO. But if have other requirements... Most terminology servers support other terminologies as well.
Stan: Virginia - please put Terminology Service on that line.
Stan (cont'd): So - does that look reasonable? Other things?
Action Item: If want to add to Amsterdam Agenda, notify Stan or Virginia
Stan: So if you think of something after this call, tell Virginia or me.
Stan (cont’d): We could have a specific MTF session, or other… So, the second day – Sunday – could get into the same sort of detail and explain to group things we discussed in Oakland… and our conclusions. Do you think more to say on terminology service on another day…?
Harold: IHTSDO SNOMED specific terminology-service. Work around Requirements. We talked in San Francisco – the need for SNOMED (identify?) server… prompt. But – what does it need to do? So I think we should keep that time [on the agenda] unless it gets tight.
Stan: OK. And plan for model content creation. In plan for 2015 – talk about plan for meeting schedule. But main plan, how curate the models… I would rename this “Designing the process for model curation”. We have done some of this, but need to use and finalize content… Have in repository by then… Think we are down to saying – what else do we want to do at the [Amsterdam] meeting?
Stan (cont’d): Would like to get further into… the semantic binding language so we can see the relation between information model and… So what would show us the way to make… The way to make (?) between information model attributes and… terminology model attributes…
Stan (cont’d): Other things to discuss in Amsterdam?
Daniel: Question – So, the semantic binding language – technical… but also discussions on how to apply in CIMI modeling style?
Stan: Yes – I would include. We talked about whether those kind of semantic bindings should be in the model or be independent of the model itself… Depend on how you use it. Will be invoked at run-time... Used in formulating queries… All that kind of discussion. We could talk about that for the rest of the day if San Francisco is a predictor…
Linda: Yes – relates to conversation in San Francisco. But need to separate that discussion with… those needed to query vs. those needed to transform… Those to validate data could be stored with… data?
Virginia: So, is query and transform a subtopic or different topic?
Stan: Yes – I think subtopic in that topic. What else would we want to talk about in Amsterdam? Thomas – can you make it to Amsterdam or participate remotely?
Thomas: I think I can make it
Stan: That would be great. We’d love to have you there.
Thomas: I’ll try.
Stan: Appreciate that.
Linda: Before talk on style… would be good to go over laboratory…
Stan: Yes. What I was thinking… That we would cover our strategy for value-set binding… The (?) binding for the attributes… What else did we decide in Oakland? I would have to go back to my notes. So – those are at least two things we could cover. Linda: And we looked at grouping of laboratory specializations around ordinal and… Might want to go over minimal representation and reference model and…
Stan: We would want to review the lab models that have been created and see if those are following the style we think is best. I don’t know if we want to take on something like a problem list model where things become much more interesting… in SNOMED. Some -- interesting in lab stuff. But value in taking on SNOMED is more visible when… So – I don’t know if take on health issue/problem list… Would be one of the most interesting… relationship between lab models and SNOMED CT.
Virginia: Semantic binding language – could go through Q1 and Q2 on Monday [see agenda].
Stan: Yes – we can use the whole morning for that. We might move problem list down… as test case.
Linda: I wonder if we are going to cover 2 sessions... terminology bindings.
Virginia: Is this correct?
Virginia: Should I reference Problem List here?
Stan: Leave it as it is.
Linda: I am available all Saturday. The Sunday and Monday I am double booked with SNOMED sessions…
Stan: So I treat this as proposed, but we have flexibility and can move you…
Linda: Thank you.
Virginia: Daniel – are you registered? Please let me know.
Daniel: I still have to register. Will do that tomorrow.
Stan: Glad you will be there.
Virginia: I am going to drop off the call – so thank you.
Progress report on work items from Oakland - Patrick
Stan: Next item was follow-up from Oakland. So – Patrick – want to describe, or just show it?
Linda: Yes – show...
Patrick: OK. And I have questions.
Linda: Sorry – I have not gotten back to you.
Patrick: Thank you for all the notes from Oakland. So – these are the archetypes. In the TB… lab… If you look at laboratory test quantitative, Thomas made groupings… The way he modeled it… So, a number of breakouts… Count per volume in blood… These… So I got these working… And I started on the list to change. And I sent note to Michael so get into BM diagram. And I got off my BMM… So no problem for others…
Linda: Did Michael get back to you? Or Harold?
Patrick: No – Harold – we need to have that done.
Harold: Sorry – I did not hear… Talking to Steve…
Linda: Talking about the… model… And you have Enterprise Architect – can you change it?
Patrick: So – how do you want me to do this, Linda? I have your list but there are a ton of different changes.
Linda: I don’t have the list.
Stan: I am interested.
Patrick goes over the list of changes decided on at Oakland meeting
Patrick: OK - Subject of care changes to subject of record.
Linda: That was for…?
Stan: I think that is better. Should have the value of the person whose medical record you are putting this into… So subject of record could be the mother… and could be the fetus or child…
Linda: I suggested Patrick put... If no SNOMED attribute... But if available, Patrick put in to see how it looked.
Patrick: OK - Next. Add constraint to subject of care - a function... Adding an at-code... And the associated SNOMED CT code... I have xx's there because I do not know which one it would be...
Linda: One - that is...
Patrick: Thank you.
Linda: Refer to RTO (?) Browser where you can search the concept ID's...
Patrick: They mostly made sense. So - I added that...
Linda: We should eventually get a function code for... But not think anything in SNOMED for [that]... So [goes] into CIMI extension...
Patrick: So - information subject - remove?
Stan: No - keep it. Need to find a code for it. Harold - need someplace where make id's... CIMI extension... But now we need some way to make permanent identifiers.
Patrick: Will put... here.
Stan: So - we used xxx in LOINC codes and caused lots of stuff on the web to flash pornographic [material]...
Stan: It is a funny aside... No real problem...
Patrick: Yes. Next - terminology binding to id-3.
Patrick: id-3 is subject of information. Oh - that is an error.
Linda: So, at-codes define the values of the...(?) elements and the id codes define the semantics. at-codes are value sets for answers, id codes are answers.
Tom: Yes. id codes are the names, or as Stan says, at-codes are value sets for answers, id codes are answers.
Daniel: This means that in instances of this, there would be the string of at-2... As a value for the participation.
Linda: Participation function...
Daniel: Yes - the value of the function.
Linda: You would not see at-2, but would see UIR and code... filled out.
Daniel: So - this is new... You see that code and not the text string at 1?
Tom: This is an ambiguous area. We can do what we want. Depends on coding structure to create templates. So in Australia, very little... We leave the at-codes as values because are reliable. But could imagine a... generator that would substitute the value set reference into the... set of the model... You could get... or could do get put at-code in, but have to know what coverings the bindings have.
Linda: In Oakland... and what goes into the fields of the terminology-binding. Tom?
Tom: Well - on my screen...
Linda: So - taking that binding reference and knowing how... and code piece... What goes into terminology attribute. That also leads to... Populating the coded-text term attribute... And where is information coming from?
Tom: You mean the text, Linda?
Tom: We could make the binding right-hand-side something more than URI... May be sensible... If got tooling, don't have to... the binding...
Daniel: I guess you would like Swedish terms... but not in other countries...
Linda: So that would be why not add...
Harold: Did we say... not identifying?
Linda: We said for version and not...
Harold: Could you bring up screen for...
Harold: OK - You are talking about the terminology-field?
Action Item: Harold and Tom schedule a time to discuss the URI and the mapping rule from ADL, and resolve before Amsterdam meeting
Harold: This is a sensible representation of the intent of the URI that makes sense to user... Can't go further on that. I need to talk to Tom... We believe we have reconciled in AOM and can... Just the mapping rule then from ADL. I would like to propose that we - Thomas and I - have this resolved by Amsterdam meeting.
Harold: So - Tom - if you can schedule a time.
Linda: Thank you.
Stan: So - Patrick - we are going through your long list.
Patrick: Yes. I see that.
Linda: That is great.
Patrick: So - that is clinical statement. I went through archetypes and removed Null Flavor if it had a cardinality of 1. At some point we will see that.
[Patrick shows on the screen]
Patrick: OK. Observation - had a lot of changes. First change was name - cardinality of 1 renamed to code. OK. I don't have this yet. I will get it renamed to code. And... changed to... And see here - Null Flavor removed.
Stan: That's great.
Linda: Every time I see that "Flavor"...
Tom: Wrong, isn't it? I think in our OpenEHR it is "Flavour" just to annoy people.
Stan: That's great.
Tom: I have been corrected by Dutch or Swedish... I made sure specialize is...
Patrick: I went through and changed them all. Some places it was one way and some places it was the other... So I switched them all... Control of the editor...
Tom: We'll change again when we get it back!
Patrick: OK - change cardinality... 0 to many.
Linda: I think you might have changed identifier from 1 to many...?
Patrick: Is correct now?
Patrick: Next change - so - id501 - to 5014... Created here incorrectly... So 501 to 14... specialization of id-5... elements... So I changed all of them.
Linda: Thank you.
Patrick: And 15 - got changed to specialization of cluster... I had been attempting to... I created another archetype called... action... and observed actions... And I broke out into another archetype...
Linda: Should we call action-set... And then specialize... since we will need other sets later on?
Stan: Yes - I think adding that step would be nice so we can use.
Linda: Yes. Can still call observation-action. Probably action set and other observation... action set.
Linda: And that... later into laboratory... observation set... depending on if need...
Patrick: I will write you to see if correct.
Patrick: OK - next change... We added a number of terminology bindings... for reason, method, status-priority, and results. Linda: I don't think was meant to be one for results... And not intend to be exhaustive for these... Might want to check with Daniel... method... corrected to method... attribute, but Daniel might recommend...
Patrick: Oh yes... 507... should have been 508. This needs to move down 1...
Linda: Daniel - comments?
Daniel: I will look into that. So I see one thing... Looking into SNOMED browser... the subject-of-record... kind of individual... but there seem to be attributes. Should be rules for how selected and used... So holds together.
Linda: Yes. Subject-of-record are...
Daniel: So (?)
Daniel: SNOMED CT without a dash. Yes.
[I missed a little]
Patrick: So I will go back and make sure it is correct so don’t have multiple columns.
Linda; We talked about - try to make semantic bindings... into an attribute... only time... focal content... Perhaps we can discuss further at meeting.
Patrick: So - that was the last of the observation changes. So - go down to... So - we had observation actions. This one, renamed to code. Data type changes. Null flavor removed. This changed to coded text. And added a number of terminology bindings.
Linda: I wonder if that was confusion... category was 5.0.7 vs. 5.0.8?
Patrick: Yes - look identical but different in observation.
Linda: Would be helpful if consistent.
Patrick: OK - I will rename them so consistent.
[missed a little] Patrick: So - first change was to Poct(?) indicator... 0 to 1. Next... Panel code... remove observation and observation-set. Next - add a terminology binding to specimen...
Linda: The husband...? attribute.
Patrick: Move on to Laboratory test. First change - to remove name. When I changed observation, name no longer appears.
Linda: We changed cardinality to 1... Previously was 1 to * when specialized... You still saw name, but my understanding... is when specialize something with cardinality of 1... doesn't reappear. Tom?
Tom: Yes - it does not reappear.
Patrick: Next - Poct indicator... cardinality of... 0 to 1.
Linda: Yes - never have more than 1.
Patrick: Terminology binding for spec... OK - that is all of the changes to archetypes that are not down into Lab tests. Linda: Also, the list of changes we should talk about. Language - but I will be patient and wait...
Daniel: I see the specimen and information... subject...
Patrick: I had generated from LOINC data and we had gone through... And I made changes to a few... And panels - are generated from LOINC data... I worked on tool that will pull out data... So I have not changed all of these. First question - should we structure the test into these categories...? I can calculate count per volume in blood... It is easy. I can take out and put into... and specialize... and I can take... for(?) - is an ordinal. For (?) - is a quantity. That is what I am working on now. I assumed we wanted to go this route. I focused on this. Question is - do we want to keep this structure? The next set of archetypes will be a lot. 1903 panels... At least 1903 archetypes... So if we keep this structure, I would like to know up front.
Tom: I came into a bit of discussion in Oakland about efficient... So I make those breakouts as an experiment... So if it was me... how bad could it get... I would want to get rid of all those child archetypes. A tuple would be a way of doing it. Did you have further discussion?
Stan: No - not in Oakland. But is the same issue when generate FHIR profiles from these. FHIR said they would make FHIR profile only to where structurally different, and then they... So my bias - I would not go into counts per volume... I would go... physical quantity to the terminal... leaf types...
Tom: Yes. The question is still the same. Whether we will have thousands of leaf analytes... I would look at ways to avoid this... My... leukocytes... We will have to think about or we will build the wrong type of... And Patrick will go crazy.
Action Item: Another topic to talk about in Amsterdam
Stan: This [above] is a good thing to talk about in a face-to-face in Amsterdam.
Stan: Want to do some other way rather than propagate 1000's of similar... But don't want people who use this to know about it... the end users... people consuming model. Whether file or table, for example. You want them to look the same whether you put knowledge a row or a table... Discuss this in Amsterdam.
Daniel: This is what terminologists or ontologists are good at... keeping hierarchies like this in order.
Stan: Yes - If this has come up in other [groups] and there are good ontological ways to do this...
Tom: If... manual or automated... I assume they apply methods of generating the results... So protocol or an M(?) modifier...
Stan: Yes. There are 1 or 2 or 3 things that are that way. There is Method. The other is Specimen. Example - people have white blood count - for common, will pre-coordinate this... pre-coordinate the white count in... and... So at some point, white blood count... might not say white blood count in wound drainage. So, same issue you raised in method... happens in specimen-type. Is a great style question. Comes back to iso-semantic type. Preferred would be break out and post-coordinate, and I agree. But in some cases it is pre-coordinated and... So need to discuss.
Joey: If post-coordinate LOINC code, must have constraints so not put the wrong code in... Can ADL handle that?
Tom: Probably not now. Would have to have tuple-constraints... Talking about tuples of specimen and methods and analytes... WiIl be paths that stick into... The pick the right LOINC code... hard to explain... Define the combination... Needs to be an extra thing in ADL... but would be quite powerful.
Tom (cont'd): I should stress - it is beyond this workbench... it is AI [artificial intelligence]... Next generation.
Daniel: People are using these LOINC codes and... without constraints today... but is doing more than what is expected.
Stan: Yes - comes back to the level of interoperability. And you are right... goes beyond what HL7 is trying to do... and FHIR.
Daniel: We have the LOINC to HL7... mapping... a few 1000 codes... in a month... And would be good... if look into it use to... So we have our attributes "tagged"-up. For example, the (?) attribute of the observable...
Stan: Yes. This is the stuff to get excited about. Talking about the clinical-side rather than technical-side... So I am excited...
Tom: Now - need to come to some conclusion.
Stan: Yes. There are smarter people... others... Heather and Ian and Hugh and Sam Heard... Probably... could get...
Sam: So - how are we doing on your list, Patrick?
Patrick: We could go through 1 or 2 of...
Stan: Did you do urine color or... where you are binding to set rather than to...
Action Item: Patrick to do urine color or urine appearance and present this in Amsterdam
Stan: Might be fun to do 1 of those before Amsterdam. Would be nice to do... I think we know how to do it...
Patrick: OK? Do you have a preference?
Stan: Urine color... or urine appearance.
Patrick: OK. I'll send you an email on that.
Patrick: OK - leaf node...
Linda: Should we have language discussion now?
Linda: We had a discussion in Oakland whether... Stan - want to say something? You were saying that you did not think there was a need to specify a [language?]... And we were... at composition level rather than data element level.
Linda: And we said we should talk to Tom or Daniel or...
Stan: Yes... so don't have to put language on each individual terminal...
Daniel: I think should be a question for European patient-summary... I don’t think I have ever seen a record in 2 languages.
Gerard: I have seen in many languages... in Belgium.
Stan: The patient record could be several languages... But in 13606... could contain several languages?
Gerard: Is unusual, but can't exclude this.
Tom: We had in openEHR, but... French and Flemish... Norwegian... Some languages - you think you are looking at one but it is the other. Now - I would not... Give an example but not sure has to be... But if pull and is Norwegian and comes back Danish... I don’t know if matters in CIMI. Can't imagine you will ever constrain it.
Gerard: I agree with you, Thomas.
Stan: I would not take out of reference model. I would have constrained it out so don’t see in every leaf type. I hate to bring up the whole idea of conductance(?)... One a whole panel... no... I would specify... this EMR is in 1 language, unless... So if came to a case where specified, then would override. But don't want to have to specify in every instance. Is related to HL7 conductance... Hope we don't have to get...
Gerard: I agree. In CIMI, we don't need, but in real life, we do.
Tom: Trouble makers... are a country of 4 million people. Lots of problems with Norwegian... Microsoft has... Annoying technology problems...
Stan: So if my intent was to say... Even though we have composite and individual entries... I would like to start with the assumption that entry was in a language... I don't know if we need to constrain it out. Two things of concern... Don't want people who are creating instances to have to fill in each time. But also - if coding, don't... then to misinterpret and think they need to specify... redundantly... would rather hide the complexity... you guys tell me.
Linda: I was wondering about the actions in response to that. Don't think we have language at the clinical level. So maybe add language to... take out. Is that what we are talking about?
Stan: Yes - other?
Gerard: For CIMI - not a problem.
Daniel: [can't hear]
Action Item: To think about the above discussion about languages over the next week, and then talk about this in Amsterdam, and finally take a vote on it.
Stan: So - think about that over the next week. Maybe talk about it in Amsterdam. And then take a vote on it.
Linda: Other things - perhaps get to these next week.
Stan: Want to express thanks to Patrick and Linda and Thomas. Great to see these models at this level of detail. I am grateful to see this wonderful work at this level.
Patrick: Yes - this is a question Thomas had. So I can go forward. Right now is a headache. Have master and sub-branch now. How about I merge back into Master and... Generate more of LOINC data... So I can go forward?
Tom: Alright with me.
Stan: Other? Stand adjourned. [end-of-meeting]