CIMI MTF Minutes 20130124
CIMI Modelling Taskforce - Meeting Minutes
Michael van der Zel
- Reflection on Scottsdale meeting
- Review and plan action items
- Comparative analysis spreadsheet template
- Archetype Object Model
- Spreadsheet template
- Serialisation format
- CIMI Modelling Style Guide
- Review of Demographics Comparative Analysis spreadsheet
- Next week (Modelling Taskforce)
- Table of Contents for the CIMI Modelling Style Guides
- Party/Demographics Models
- 2 weeks
- Review of CIMI Reference Model
Detailed Meeting Minutes
Linda: Need a week to do a Prep, examples of CIMI Ref Model, before begin. Action items.
Joey: Modeling for Lab Report. If FHIR not do lab report - should we talk about in Tuesday or Thursday meeting?
Linda: Tuesday - Terminology Model implementation. Whether talk about now - why not?
Joey: OK - talk about before we end the meeting.
Linda: OK - in [part of Agenda] Review and Plan items.
Linda (cont'd): So - relfection on Scottsdale. Colleen normally asks people to talk about what they thought of the meeting and what to do better next time. So, if you have something to say... If you can let us know what you liked...
Galen: I liked seeing tools working rather than just talking about them - extremely helpful... Snow Owl, John Gutai's tool. Very pleased. Maybe premature - would like hands-on modeling/tooling sessions with tool-builders. Hard to do over the phone.
Linda: Great idea. Easier when tooling is in place.
Jay: Anything concrete is good. Another example - is for terminology binding. Better ways to interpret - as soon as get into a tool the sooner they will be complete.
Joey: I thought - people were coming to realization that UML models and CIMI Reference Model of UML are similar... so can exist together. Dave's presentation showed that. Coming together... and I liked the microphones, too.
Linda: They controlled the conversation a bit.
Mark: I learned a lot. Interested in figuring out how CIMI models will interact with other groups. How make more computable and... on human intervention. So I am interested in FHIR and guidance for how CIMI can work with other models.
Linda: Patrick - your first meeting?
Patrick: Yes - I liked the demonstrations so I could come up to speed, although time-limited. The use of the mics was really good, although we need a few more. And looking forward to working with the group - will be fantastic.
Peter: I think it went well in all respects. It was organized well, ran well, and moved at the right speed.
Daniel: One thing - want more of the detailed modeling discussion. More effective face-to-face than over the phone.
Eithne: Enjoyed the demonstrations. The meeting was well organized. I was very happy with the use of the mics - especially during second and third days.
Stan: I liked the meeting - especially the parts you chaired, Linda. I enjoyed the tool demonstration. Hope we can choose tools and do modeling in meeting, or at least review. I was happy with Mayo - Thanks to Chris Chute and Harold for setting up. No downside. Always anxious to get to next milestone... get to terminology-binding... further discussion on the topic.
Linda: Do you have comments about meeting?
Michael: Not yet - if I remember something I will tell you.
Linda: I agree with what all said - good meeting. Great to see all there. Hopefully we'll work toward implementing models and having tools in place. Thanks.
Linda (cont'd): Want to review and plan action. Start talking about FHIR - Joey?
Joey: So - we committed to do with FHIR - create profiles restricting the Lab resource based on what we have done. They cancelled for January - changed to May - but now not in May - will be doing the document. Graham said we will be receiver of lab report - lots of work - why did we agree? If only... and sent to be stored in server up there... lots more work. So - what should we do? Linda and Stan got emails [regarding this].
Stan: What we are trying to accomplish - trying to find people who will consume what we are making. Will be different in each situation and will change over time. I know Graham said he does not want us to make resources. I think over time he will say - want someone to make resources and creation of profile. So we may need to talk to Graham. I don't see CIMI as a Connectathon implementer. We are trying to provide common profile and... So - ask Graham - who is going to define their profile? Each will do their own? So if these are the right ones, then a different... Maybe - consolidated CDA templates... creating executable version of models... correspond to Meaningful Use... health models... or better tie-in to Australia. We want people who consume the model. Or maybe Intermountain Health... or folks at Kaiser - what could CIMI provide that would add value to what they have? I can have a conversation with Graham... make sure he understands what CIMI is doing. Not only application developer at Connectathon.
Joey: So - want limited resource in FHIR. If 80% have attributes - then add to model. Graham - straw-man specimen. Good for Graham to see our... and he can get an idea of 80/20 rule by looking at what we have. The Connectathon is implementation focused - receivers and [senders?] of data... Clients taking instance-data and send off. Need lab info system to do this and CIMI probably not want to do this. If... then fine - but ton of implementation work. Beyond the scope... Huge amount of work. Simpler - participate in documents and participate in Connectathon with this. Stan - what do you think?
Stan: Not excited about setting up lab system. But I'm encouraged that they need that infrastructure for Connectathon. Interesting that they use the 80/20 rule... no one could do this with 80/20 rule.
Joey: So, maybe then we map the CIMI models to the lab resource, and see what we learn from doing this mapping?
Stan: Learning experience, but I was hoping - we were supporting that system. Won't know until we have the people. Idea might be - there are no consumers, so don't make models, but I don't think so.
Joey: Generating profiles.
Comment: If we generate, then implementers won't have to.
Galen: Why do this granularity of... Not message specifications, but concrete bits of knowledge. So, produce atomic size bits of knowledge like serum sodium. But FHIR trying to do the same... overlap... could FHIR benefit? I think yes, but no go through huge implementation exercise.
Joey: That's great. So - we could show - is there a map from what we have to do... Lots of people with profiles with extensions... You could add a field not available. If they are going to make an extension - should come from CIMI models. So not lots of extension.
Mark: Also, in FHIR - can register extensions. So may be a review of extensions finds enough... put one thing in one extension... valuable to communities.
Galen: I was trying to say - philosophical affinities between CIMI and FHIR. We are trying to create specific... CIMI is the builder of content... content-expertise... make explicit in CIMI models.
Linda: Hopefully useful conversation. Joey - I agree with you... exploring mapping... working with Graham to participate in future Connectathon.
Galen: Want to explore with FHIR... not application talk to FHIR application. Only - how CIMI content gets into FHIR...
Michael: With FHIR - no general observation in FHIR. I'm convinced that you need to put a bunch of observations into a general observation resource. But Graham thinks no such thing. I think lots of general observations in CIMI should be put into a resource profile.
Linda: Yes - difference in modeling paradigm.
Michael: Yes - Blood Pressure, Body Temperature... should not involve separate resource.
Stan: Concrete actions. Let's share specimen model with Graham and FHIR... Learn something... Trying to create profile in FHIR, to tell if the resource is valid... FHIR tooling and information model are two separate components. We can do two things: One - Sharing specimen model. Two - Make profiles even though not part of Connectathon.
Michael: Yes - and use general observations.
Linda: Yes - discuss with Graham.
Mark: Yes - if don't have generic observation, then everything is an extension, so FHIR is focusing on the concrete, rather than on interoperability...?
Galen: Excellent observation. They are focusing on concrete and no model behind it. All "hand-built"(?). So CIMI could add.
Linda: So - discuss with Graham, Joey?
Joey: Yes - get specimen model to Graham and continue to make profiles. What Michael said - general resource - we could talk to Graham although they don't want to do...
Linda: Yes - this sounds like a good path forward. The other things Stan raised - potentially look at consuming IMH Lab specializations. Building a FHIR profile extension. Also mapping to CDA templates. I have bias to XML schema although these won't capture all the constraints we need (such as terminology value set constraints). Need validation tool for that. In Singapore, we have found XML and excellent way to begin the implementation work. Is this something we want to do by hand or wait till tooling?
Joey: FHIR stuff will consume my time...
Linda: So unless someone else wants...
Stan: Yes - may generate 1 or 2 models by hand to understand... tool. But... for now... [can't hear]... generation of FHIR profiles.
Linda: Joey - is that all on FHIR?
Linda: Next - Virginia's slides - action items. One on wiki... some not relevant... Stephen said NEHTA... v3 constructs. Mark had question... Does modeling process work? How to do sequential or parallel processing of models - look at templates... and here is MTF action items.
Linda (cont'd): First start work on CIMI modeling Style Guide. Not sure where our priorities are - finish models or style guide? In Groningen - discussed existing style guide. Said would separate into 3 guides. Technical Guide, Modeling process/How to/Methodology, and third - modeling Style. Is that good?
Stan: Yes - good division.
Rahil: Does technical guide include terminology binding? We started writing up in style...
Linda: First step is to Draft Table of Contents. Is some overlap. In technical guide - talk about reference model, archetype ref model and the patterns and terminology...
Rahil: Should be a separate section that talks about terminology binding. Calls for a separate section to explain each of them in detail.
Linda: Will have to touch on in all guides, but in one guide more [fully]...
Rahil: Yes - links... would be helpful.
Daniel: We could look into way the IHTSDO has made... parts in multiple documents, but referenced. Lots of overlap, but one source that is referenced from both documents. So I see that... binding... if terminology binding mentioned... I see benefit of collected... Could have both - action terminology binding guide.
Linda: Good point. What is authoring tool document?
Daniel: Not sure, but DITA - high level - complex. But idea of having parts of document referenced is good idea.
Linda: So - linking so we don't repeat the content. Is that it? So write Style Guide in Microsoft Word, but got to be in online version. So don't need to rewrite - hyperlinks. Then print out single pdf.
Daniel: Yes - and if someone has knowledge of documentation systems... CIMI would have to choose. (?) is a headache to set up.
Linda: Anyone else - experience with documentation system?
Michael: DITA - I don't have much experience, but Dave does.
Linda: OK - build initially and talk to Dave and maybe Harold. So next week - present Table of Contents for Style Guide. Rahil - do you have time to help? Or Daniel? Table of Contents for Terminology Style Guide?
Linda: I think Table of Contents is best place to start. OK - next - finalize reference model. Michael - will you have time to look at examples of what we discussed in Scottsdale?
Michael: Surgery is Monday, so I don't know...
Linda: Yes - maybe in 2 weeks.
Michael: Yes - and I want to create examples... How the model will look like... during recovery.
Linda: Thank you Michael. You will be bored... So - work on party models - the comparison. I haven't finished with analysis -waiting on FHIM.
Galen: You have already mapped the important stuff. The other... I don't have in...
Linda: I did mapping for person... But address?
Galen: There is data-type for address. I don't have... at all, just city, state, country...
Linda: Can you do into...?
Galen: Can have done by next Thursday.
Linda: Any other... We did person.
Galen: If helpful... FHIM is mapped to HL7 v2. So I can give you HL7 v2 to FHIM mapping. Could provide the basis for...
Linda: Could help, but have not done v2... You don't have time this week?
Galen: Between now and end of month, I am slammed. Then I will be unemployed. So definitely get... and demographics done by Thursday.
Rahil: Are you looking at more structured model for address?
Linda: Now - individual data items. All others are based on patterns... Need to do for address. So address type - or generic model - then specify... so can do direct mapping...
Rahil: I ask because there is a lot of work we did to look at 3 types of address. One - (?)... Two - Line 1,2,3,4,5 structure style. 3 - Highly structured with specific... like building#, street# - quite detailed. I don't know if looking at all 3 or...
Linda (cont'd): The different models take different approach to that. NEHTA - highly structured. Address Line 1, then Address Line 2, and allows these to be flexible. Say what line, and what the components are. So we need to support 2 different models for address lines - unstructured address lines, and those where the parts of the address line are modeled.
Rahil: Yes - should be at least 3 variants of an address so can cater to different requirements. I have material on that. Don't know if Galen has or not.
Galen: I don't know if I have that particular... [noise - can't hear - bad connection]
Rahil: Did you want it?
Galen: Yes - would be helpful.
Linda: Send to the group, please.
Linda: So - note - consider generalized model, plus unstructured and highly structured models.
Gerard: About demographics, have you had a chance to look at Mindmap I sent you?
Gerard: Commission is pushing for addresses. I will send to you.
Linda: Next action item was lab results. Want to finish action items. [reads from list] Joey to send Linda Specializations.
Joey: Yes - we don't have these in CML, so I go back to ASN and then read LOINC codes. I can send you ASN... trying to make it prettier.
Stan: Someone from modeling... can they help you?
Joey: Is that OK to do?
Linda: That is the Lab Report specializations...
Joey: The CBC with differential, CBC without differential, and creatinine clearance and glucose tolerance test.
Linda: OK - terminology binding work. Daniel and Rahil - can you talk at Terminology meeting?
Rahil: Yes - I started putting in bindings. There are a lot there already. I need to ask you... I saw a lot of default bindings already.
Linda: Yes - I stuck some in . Needs to be reviewed. We can have a session off-line... whatever suits you best.
Daniel: We can email about this.
Linda: Good - target is for next Tuesday, but if too soon, can go through models... one at a time... then need to assign new patterns... need to agree and have specializations agree with that.
Rahil: The relationships just basically had a "Has" in front of them and repeated the class. Example relationship Has... request. Harold said - better if not a lexical relationship... want more semantic. Now - I put down what we had in LRA. Do we want to use those or come up with our own and ... replace... later...?
Linda: I would like to see the specific bindings we need as specializations... Relationship types... LRA may be high-level, so we may need to identify...
Rahil: Also - the groups that submitted (?)... whether they used... the relationships we have... specific... whether those who contributed... if have useful...
Linda: The only relationship binding... meaning codes... IHTSDO codes... relationships... Anyone aware of other work in this area? No?
Rahil: So - maybe we will be the first if we see specialized...
Linda: Yes - may need generalized... May be a piece of work... to organize...
Linda: So - the last thing is creating the instance examples. Anyone want to follow up? Stan - IMH want to do this?
Stan: Want to get to it in the next month or two, but other priorities now.
Linda: Everyone to contribute to review of models. Example - moving location (?) to participation... But I think we have done as much as... Once finalization of draft models, need review by group... And once we have repositories and... This will be easier. But... editorial board to review models.
Stan: Please repeat what you said?
Linda: Need to develop a process for review. First - modeling task force review... External review by clinical groups. Is that what you thought on how to get to more public...?
Linda: And when we have tooling, this will be easier, but until then we will focus on the internal review process ... So once I have changed I will put models on Google docs.
Peter: Are we going to have a group vote on that... model before it is official?
Linda: First Michael will show... before vote. But if unclear with instance and can't reach agreement...
Stan: Even if not have to vote, would be nice to have official vote... like in Pleasanton.
Linda: Yes - and if other parts need reviewed, it would be good for people to say... So first part - demographics. More stable... that is easier for tooling.
Linda (cont'd): Now - prioritize additional models to work on. Immunization is next? A couple of months away from this. Need to start comparative analysis first. So in next month - tie up what we can and look at comparative analysis...
Linda (cont'd): With respect to... comparative analysis... Anneke sent me a template based on spreadsheet. Anneke added value set columns. I will send around - All to fill in their own models and any that they add...
[Shows spreadsheet on screen]
Linda: I have collapsed... I want to process, will need to... Not sure if another area where we can set on line system so people can add. Cool to have website so can add collaboratively.
Rahil: Were we not looking at GitHub to do work collaboratively? I remember that GitHub...
Linda: We were using it. When we started looking at attachments... Mindmaps... Google docs - not have version control... Decided GitHub not have document capability.
Patrick: GitHib is geared toward more text - code... not code with binary attachments. Not at granularity... GitHib not buys us anything... For what we want to do - not help us here.
Rahil: Could use Sharepoint... as a repository for resources... Could be documented as well. And you have user-rights... write access. I don't know if free.
Linda: You do need to buy. Would have to set up for CIMI to do.
Rail: I can ask Nicholas.
Patrick: That is a Microsoft... cross platform. Would this work for all?
Linda: Can you check this, Rahil? Let us know.
Rahil: Yes - if all are OK, then I can ask Nicholas.
Daniel: I have experience... Usually around Linux... I can post documents, but some work better in Microsoft. But no merging capability like you can do with text files in GitHub.
Linda: Now using Google docs, so question is, if value-added on top of that. Rahil - look into?
Rahil: Yes - I'll ask for other suggestions.
Linda: All agree - immunization is next model? Stan - your understanding?
Linda: Anyone remember if we included immunizations in last call on models? I have a number of immunization models. Does anyone remember? If there has not been a call for models, I need to send out now.
Linda: OK - Tooling workgroup with Terminology and Information Tooling subgroups.
Linda: Dave on line? OK - FHIR profile worked - already discussed and collaborative analysis - I will send around. And... especially spreadsheet version, to see what is in Archetype model... and especially on that topic - Joey needs to look at... Firstly - I can show you the Singapore Logical model done as a tooling... captures the archetype model across top. This is based on ISO 13606... We have the name of the reference model attribute - for example 'item' or 'participation'. The Reference Model Attribute and the Reference Model Type helps us to show how the reference model is constrained by the archetype, and further specialized. We have the cardinality... derivation of name... Anything in archetype will be a column... spreadsheet... May be a way for us to explore... Early on in... task force... a way CIMI can... especially in machine-readable way... the node name and how do hierarchy... The relationship to parent... So this is relationship binding and ... binding.. Cardinality definition examples... Another constraint...
Linda: One alternative is to expand to fully align with Archetype model... and to expand...
Linda: What are people's thoughts on how to explore this? This is an example of Heart Rate... Heart Rate present is a data item... I can send this draft out. Any comments? Joey?
Joey: I thought people wanted to table this. I don't want to push things people want to do.
Jay: May be difficult because of [bad connection - noise]... Might be... to make that...
Mark: [can't hear]
Linda: We just have an extra column?
Linda: So show to clinician?
Rahil: Would it not be possible to have archetype model as a pattern... Would that not be easy in generating lots of models?
Linda: Yes - but how far is Dave's tool from operationalizing? We can explore with Dave...
Rahil: Yes - is impressive. Don't know how to put all in spreadsheet. But I agree - tooling is where we want to go.
Phone [Joey?]: A prototype - but he built it in 1 and one half weeks. So it's been another week...
Linda: So - yes - we need to include.
Joey: Still send out spreadsheet?
Linda: Yes - even though out-of-date... [Linda keeps reading] Comparative analysis and... Some work for Stan. Sarah - work on website... Oh - here is Priorities for Model Domains. So if focus on Lab and make sure have... for immunizations, then we can move ahead.
Linda (cont'd): Authoring tool options - we need Harold and Dave. Modeling - Stan? Or Dan? Model Repository...
Linda (cont'd): Any other topics or action items? So - next week - we will intro Table of Contents for Style Guide. Going to look at Party/Demographic Models. Anything else? Now Tuesday - will talk about Terminology Tooling, and if Daniel and Rahil have anything...
Linda: Anything else? We'll finish early.