CIMI MTF Minutes 20140417

CIMI Modeling Taskforce - Meeting Minutes

Thursday 17 Apr 2014 @ 20:00-22:00 UTC

Attendees

• Deepak Sharma
• Stan Huff
• Gerard Freriks
• Harold Solbrig
• Joey Coyle
• Mark Shafarman
• Stephen Hufnagel
• Ian McNicoll
• Dave Carlson
• Jay Lyle
• Patrick Langford
• Eithne Keelaghan

Draft Agenda

1. Next face-to-face meeting in Phoenix May 1-3
2. DRAFT Proposed US-Realm Informatics profile - Stephen Hufnagel
3. Final questions or discussion on options for RM supporting panels
   1. Voting ends on Monday April 21
4. Any other business

Detailed Meeting Minutes

Stan: Will talk about face-to-face in Phoenix. Then time for Steve Hufnagel to talk about DRAFT Proposed US-Realm Informatics. Then talk about Ref Model options we sent out. Voting ends this Monday, and if time permits, Gerard sent out... and we will pursue that subject as well. Other things?

[No Response]

Next face-to-face meeting in Phoenix May 1-3

Stan: OK. Next meeting is Thursday, Friday, Saturday before HL7 meeting in Phoenix. Is in same hotel as other HL7 meetings. This is because the agreement we have with HL7 to collect meeting fees and other moneys. So they have reserved the meeting rooms.

Stan (cont'd): You should have also received meeting agenda. Please review and comment on whether we have too little or too much time for each topic. There are things here... model creation, OMG/AML... CIMI activities with SHN, FHIR... and then membership, policies... etc. We can discuss now, but best if all review the agenda and see if we are talking about appropriate things.

Steve: Should we register before, or can we register when we get there?

Stan: Usually it is best before, but the rooms have already been reserved, so... But - Yes. Pre-registration would help. You should have received email for that.

Steve: Is there time set aside for business discussion in the agenda? I did not see this. Initially it was for Saturday.

Stan: Should have been on this. Initially with Jamie Ferguson... I think this "Adoption and Value" on the agenda is the same subject, but we might want to refocus. Also, the HealthCare [?] subject touches on that. Good point, Steve. I will look at this. Other?

Steve: You have a question mark with Keith's name. Will he be there?

Stan: I haven't been able to talk to him yet. I thought he was going to be at the HL7 meeting.

Steve: He is. He may need to adjust his travels. Part of what I am presenting is Keith's work, the so-called "Legos".

Action Item: Track down whether Keith will attend Phoenix meeting

Stan: OK - We will track that down. Now we'll go to you, Steve. I'll make you presenter.

DRAFT Proposed US-Realm Informatics profile - Stephen Hufnagel

Steve: Are you able to see my screen? (Slides)

Stan: Yes.

[slide #1] Executive Summary

[Slide #2] Agenda (3 Separate Parts/Briefs)

Steve: OK. I have been asked by DOD and... to take the CIMI HL7 work that the (?) have invested in for the last 15 years and try to put into an engineering process. Because currently it is hard for people to imagine how to put it into a (?) process. So - no federal agency up to now has endorsed this. They have asked me to vet this. ...to get everyone's input.

[slide #3] Mission: DoD-and-VA Data Sharing

Steve: So - I was asked - what does this do for the patient. And I added this because we informaticians... tend to get wrapped up with the... So, a patient has an injury in one part of the world, and he gets transported because want quality of care and no duplication of tests and treatment. So the objectives: Get the right information to the right person at the right time. Looking for the "easy button" for the physician... who knows what the system wants...

[Steve reads Slide #3: Fit-for-Purpose Data]

Steve: Fit-for-Purpose: when and where it is needed by clinicians provided by agile, aligned and interoperable systems. Data Quality - ensured by standards-based :Informatics Factory" interoperability specifications, Implementation Guides and test Criteria

Steve: So data quality is ensured by standards.

[Steve reads #3 in slide #3]

Steve: Project Management Accountability and Audi Readiness - using EHR System Functional Model (EHRS-FM) to baseline, roadmap, categorize and configuration manage DoD-VA Data-Sharing architecture.

[Steve reads #4 in slide #3]

Steve: There needs to be a monitoring and reporting dashboard for/by IPO-and-GAO Agencies using HL7 Service Aware Interoperability Framework (SAIF) Enterprise Compliance and Conformance Framework... As we go forward... I will show you how it will be used.

[slide #4] Executive Summary (EHRS-FM)

Steve: 1.1 released in 2009. Version release in 2014. 320 functions. 1000 Conference Criter (CC's).

Steve (cont'd): Record infrastructure - Auditing. There are 24 things that can happen to a record, such as deleted. Destroyed...

[slide #5] Executive Summary (Recommendation)

Steve: US Realm version would have the HITSP (Healthcare Information Technology Standards) version. Mapped/Linked to functional model... The HL7... Legislative mandates, such as CLIA... And take the... and Fast-Health information resources... and make... NIEM and FHIR...

Steve (cont'd): An activity model - 1 or more activities would occur that are dependent on functions. Results in one or more information-exchanges. So would all be defined in this profile. And then separate for DoD... And VA-input care - mainly chronic care for retirees.

Steve (cont'd) So the objective - to harmonize clinical operations... And the systems objective... I am focusing on DoD and VA. But Health and Human Services has CDC and Indian Health and...

[slide #6] Business-Informatics architecture

Steve: Activity model - most applicable to this group. An interoperability specification factory. When I started working with CIMI, I would talk to William Goossen in the Netherlands at CIMI... and they all used terms and I thought they were talking same story, and I have had to take...

Steve (cont'd): In U.S., we have the FHIM that is being developed. And... with this...DCM/CIMI archetype... Data elements... grouped into a module... So in HITSP... a group of 16 data modules... medication list and... Could be grouped into a subset of the 16 for a transfer of care. It was determined to be insufficient. Now we have CCDA... Can have a summary of searchable care... There are 45 models.

Steve (cont'd): You could expect for each of the 45 models, that, for example, for Blood Pressure, might be a composite DCM or archetype that could define... or the metadata... Along came Keith Campbell at the VA who said you must define structure. For data element, should be a semantic assertion. For example: Blood Pressure taken on resting patient on forearm... and also whether True, False or value.

Steve (cont'd): So - whether going to run in Protégé, if say... want to do patient-data-analytics... So if patient had BP, and wanted to run medication... It is critical to know if patient is on a treadmill or chronically ill and bed-ridden. So - semantic assertion... So that is what LEGO-semantics are doing. It uses (IHTSDO?). Use "Lego" because are using primitive concept, and are composing... whereas use... and build information models.

Steve (cont'd): Then have tools. Can take LEGO information or LEGO-info-model and can convert into CDA document or FHIR resource or... So - now have tooling that can go together and support runtime environment. So that will support the Systems Exchange Architecture [see bottom of slide #6].

[slide #7] Interoperability-Specification Factory

Steve: Any questions?

Stan: So, in this architecture, the FHIM [Federal Health Information Model] - it has objects at about the same level as FHIR resources?

Steve: It is a many-to-many relationship. So FHIM - a FHIM class can be described by a FHIR... FHIM is a UML model. FHIR is an XML implementation schema. So in a perfect world, there could be a 1-to-1 correspondence.

Stan: Yes. The worry, we've seen this with FHIR resources, where those are close but not the same.

Steve: I am creating FHIR profile and making them the same. Are creating the FHIR resources and equating them... So this proposal would be that there would be some harmonization... So, if disjoint, we would sit down and sort out and make consistent. So, I want this to go to National Coordinator because they could say "we need to harmonize the FHIM and FHIR to be consistent with how we want..."

Stan: 2 levels of harmonization. The traditional use of FHIR. Take resource of [?]. And that is the [?] for Blood Pressure.

Steve: Another level of harmonization is the data dictionary. We will have to agree on this or will end up shooting each other in the foot.

Stan: I agree. My question is... are you going to take advantage of what CIMI is producing?

Steve: Yes. There would be a harmonization when all the stakeholders... Where this would be a catalyst. Where we will identify what is disjoint. Like what HITSP tried to do. Tried to harmonize differences among groups. This moves U.S. in that direction.

Stan: OK. Is that what Keith Campbell called "SOLOR"?

Steve" Yes - SOLAR. SNOMED, LOINC, and Rx Norm. And we would create... would be used for terminology.

Stan: Yes. And the idea of SOLOR would be a non-overlapping, non-redundant subset of those terminologies that would be used in... clinical data. So only that used in clinical data representation. So would be a subset of those. Not only pouring SNOMED, LOINC and RxNorm into data dictionary. But would be using non-overlapping, non-redundant parts... and support the LEGO modeling aspects of the system. Is that right?

Steve: Yes. I concur. That is why I asked if Keith will be there at the meeting. And the good thing is all the tools are open-source.

Dave Carlson: Clarification. The tools for this ... on the GitHub... are open source. I have been with... group to integrate SOLOR... Eclipse... So SOLOR - loaded into the SNOMED terminology... So can create a pre-coordinated concept. Can browse Reference Sets... So blurs MDST... trying to make a... to support...

Steve: Want to get to a... Black magic. From LEGOS to FHIM with MDHT. Because like you said... LEGOS is in workbench and MDHT is independent of that.

Dave: I am working on making them not independent... so can define...

Steve: Yes. And Jay told me there is a different binding in FHIM and...

Dave: Yes. FHIM binding has used... MDHT 2.0 next generation that integrates the CTS2 constructs and...

Jay: So the FHIM implementation binding produces CDA but can produce other as well?

Dave: Yes...

Jay: Good. Thanks.

Dave: It was designed generic.

Jay: So not part of CDA-spec?

Dave: Yes.

Gerard's work with Irish Government

Gerard: I finished a report for the Irish government about things that are compatible, but there are differences. Too much to discuss now, but... is a US-approach... Leaving out standards... and... We will need more primitives. Not now but after your presentation, we must find... Ireland, Spain and (?)... And archetypes play a central role in what you do. Are not defining a data dictionary or... You have to [?] more than data dictionaries allow you to.

Steve: That is the purpose of this presentation. So we don't go off without considering other viewpoints.

Gerard: So I must hear full presentation.

Steve: Other questions? My goal is to take all the work that CIMI and HL7 have been doing and how we can use in our environment. So now I will show you...

[slide #7] Interoperability-Specification Factory

Steve: Trying to take informatics... How can be used by implementers who don't want to be informaticians...

[Steve reviewed slide #7]

Gerard: A selection of CD archetypes and templates... Represent the government and... leave to industry how they implement.... You go much further into engineering...

Steve: There is the issue of "What is interoperability"? You have to have structure, semantics... And in the US, the government is trying to experiment where you can standardize at the lower level... So you can harmonize. Clinician says "you give me semi-archetypes..." What will I do with those? Can purchase from vendors and plug and operate... That is why HL7 is so excited about FHIR. It is an attempt to ignore all above and implement... Clinical concept... interoperability.

Gerard: In Ireland, they use... and set some support are provided with semantic interoperability. Whatever they do in own realm is up to them. When want to exchange with government, is 13606... We are aware - HL7, CDA, FHIR... But it is up to the vendor...

Steve: I think our case is similar on commercial side. Federal agencies want to share information amongst themselves. But also, as in Ireland, can... But if use tools and artefacts, will be cheaper and...

Stan: These two things have more in common than conflict. If you have a shared common logical model with appropriate... That will create opportunity for interoperability... So is the... people to create open source... But you can see where something happens... where semantic web becomes more popular... But using RDF and triple for the actual implementation ... The core is having shared logical models. The basis for interoperability... Get a lot from those. And if want to exchange with Ireland, then someone can make a widget to... I may be oversimplifying... Once have common logical model and terminology, then the transformation of images is a do-able task.

Steve: Yes. Keith is... take model and develop deterministic way... the CDA or FHIR...

Dave: The FHIR work is just started.

Steve: Yes - ongoing.

Stan: So Steve - CCDA vs FHIR. You could use FHIR in service area, and CDA a more traditional...?

Steve: That is what DOD/VA is... They have... Initially only did C[?]32, now CCDA... For internal - using Data Management Services, based on FHIR resources - are easier for processing JAVA where... need parsing and parsing... performance trade-off...

Stan: My experience - we have been comfortable with CDA or C-CDA when display the document to clinician. But for discrete data... Are much more oriented to... architecture because CDA is clunky...

Stan (cont'd): My other question - Important term binding... terminology... syntax in CIMI. Mentioned LEGO and... Also IHTSDO... Binding syntax... Dave Markwell and Linda Bird developing... Do you see this all coming together? Would be best if... had...

Steve: Yes. I am seeing this as a harmonization umbrella using the functional model as categorization. So lab - 1 set of things. Immunization - another set. So, using factory analogy - seeing those things that come together. So you may not care how you create a CIMI archetype or clinical model, but you are concerned with how going to put together for clinical model. So this is an attempt to find disconnects and then work with groups to... So, off of National Coordinator... to take responsibility to have stakeholders harmonize and... So the answer is yes, and I think with binding, everything I look at different approaches. Software engineer who don't know about Healthcare.... HITSP... Implementers said - too complicated. We want to come in here and... do our job. Graham Grieve wants to foster things that are easier... and this binding...

Gerard: Yes. Binding is very important. But other... Dealing with governance... European commission... A process for governance... being set up. Also... would be nice to see where... harmonize...

Steve: Now in U.S. - no point of contact for you to talk to. So - that is purpose of this proposal.

Gerard: Nice. Also, a few days ago... Mayo and CEM and ISO and... people versed in coding and CDA and... They meet for 1 year and discuss codes and standards. Would try to come up with recommendations for U.S. and European... and... for standard organizations... And Harold will be with us.

Steve: I think we are all struggling with this. There is a problem. We need some leadership. Stan has been trying to do this... To harmonize... and put under the banner of informatics.

Gerard: Good chances to find common ground.

Steve: Yes.

Stan: This is a very useful characterization... An attempt to fit all pieces together. I have to look at more. But I like taking all... putting together... Take advantage of different areas... Second point - given the work with FHIM, the work we are doing with AML - I think would have relevance. Is a UML stereotype and profiles... Make a (lawless?) implementation of... I don't think you have it here. Not mentioned ADL... Are you aware?

Steve: Yes. And Dave Carlson on call, and ... Lawson... I am having difficulty having people understand this complexity. This is a hard sell. They want a simplified view of world that works. They see this as an ivory tower view.. So I have this... factory process... Can put all together... ADL workbench and IHTSDO and... So when needed, can output... for CDA guides and services that meet need of implementers.

Stan: This is good for this audience and CIMI audience, but...

Steve: ....understand this picture.

[slide #6]

Steve: But in dealing with others... But on program manager level, I would stop with this picture... Traceability and accountability at top... We can give to implementers as implementer guide so can hand to... who don't know about Healthcare... I created the second figure first and I got stares, so went back...

Stan: I wonder if you could take time at CIMI meeting in Phoenix to present this.

Steve: Sure. I want to get feedback so they can know whether...

Stan: OK. So I will re-work the Phoenix agenda...

Steve: Yes, and should tie this and Keith's discussion together.

Stan: Yes. I will talk with Keith.

Steve: I can finish the last 2 slides.

[slide #8] Knowledge Management Environment

Steve: I am calling this a "Knowledge Management Environment". All have to be tied to what I showed previously. So, I am suggesting they be put together... So, want to be able to do business and monitoring and reporting... Can assure people that architecture does what legislature and funding intended.

[slide #9]

Steve: So can look at from ... dimension and... dimension and... And the conceptual and logic and implementable views. And you can then say - we have the appropriate... tools... And have the... profiles... Might have the services specified...And the low-level schemas and...

[slide #10]

[slide #11]

Steve: I am suggesting... harmonization... Traceability...

Stan: I think this is very useful. I need to look at the details. But as a way to show how things should be put together... I appreciate you showing this. Questions?

Gerard: I would like to contact Steve and share what I did in Ireland.

Steve: That would be wonderful. I may plagiarize your ideas if OK.

Gerard: I must discuss with those in Ireland. I don't expect problems, but...

Steve: I am in similar situation. I have been chastised. This is not an official approach, not endorsed. That is why I put Red on all slides.

Gerard: I may be in a similar situation.

Stan: Other questions?

Ian: Thanks, Steve. That was interesting. EHR-world - opened up... at the core, clinical, health-environment. How do you get rapid clinical oversight? I don't think...

Steve: I am addressing that by using the word "factory". If we can create less of a black box... it becomes an engineering process. HITSP failed. US government spent millions on it. When implementation came in to create, they said they did not want to spend 6 months trying to understand HITSP document.

Ian: I am meeting with... in Australia now... creating documents... and what implementers need... Methodology... Open project... have small domains. Small concept...vital signs, or... and have... who decide whether artifacts are good to go or not. A new way of working.

Steve: I am trying to suggest that way of working. Now it is the "Wild West". EHR market - fragmented.

Ian: In Australia... we are seeing standards... Maintenance is the problem.

Steve: I think that is the importance of selling the factory concept so maintenance is do-able. Need to crank out... and not spend millions each time we have to change the CDA.

Ian: Need clinicians involved in the process.

Steve: Stan said it. If clinician involved on logical level, and you have a tool that can create what implementers can use, then... good...

Gerard: Yes. Has been presented in Ireland.

Steve: And we have great confidence in Dave Carlson to do this in the U.S.

Dave: But no support...

Steve: So I have a part 2 and part 3. Was painful to get down to 10 slides. So if you want the whole deck, I have those.

Gerard: I will share mine with you.

Final questions or discussion on options for RM supporting panels

Stan: OK. Agenda - final question and discussion on modeling of panels and... I sent out package of 3 options with deadline for EOB [end-of-business] on Monday for people to vote. Any questions?

Harold: Does option #3 include changes in definition of Entry as defined now?

[Stan shows slide] Proposed RM for compound and Indivisible Entries

Stan: We still have entry. Would have to choose if indivisible or compound entry. And indivisible Entry is what we would call statement and... An indivisible entry consists of items - either clusters or elements. And compound entry consists of indivisible entries and... Then can be elements and clusters...

Harold: The definition of entry includes "The entry is the minimal unit of information that a query can return". Do you still view it as indivisible? Or is your intent to move it down to indivisible entry?

Stan: I am thinking that the indivisible entry is the query-safe... item. And if compound statement, would still be indivisible.

Harold: Causes me trouble. Two kinds of compound entries. One where combination of heart rate and subject modifier - fetus. And if it takes heart rate from fetus, then you violated one of the requirements. But if say compound entry is indivisible, then...

Stan: Yes. I am with you. You can construct... The assumption should be, if I specified as items in the entry, and then BP and HR... Yes. Absolutely could not retrieve Heart Rate out of here, and you would not know if baby or mother.

Harold: So - leaving the... of Heart Rate alone?

Stan: Yes. Probably have to be stated another way. This thing conceivable has many entries in it. What you are saying - does reside here, but... applies to things when there are multiple... rather than...

Harold: Yes - would like to be clear. 2 issues. Panel issue - retrieved separately. And fetus Heart Rate issue. Want to tack "fetus" onto it, and... And I don't think can do consistently with this. Either can take compound entry and...

Steve: We have not differentiated the scoping of our meta-data. Are you saying - does meta-data apply to whole... model, or... And that is why Keith's approach would resolve that problem.

Ian: May resolve problem, but creates a huge workload... and lots of governance.

Steve: Like putting a set of parentheses around... to associate meta-data with a set of elements. Not sorted out yet.

Harold: Another way... Careful with term 'metadata'. Stan uses qualifier and modifier...

Steve: "Context data".

Harold: I usually associate context data with [?]. But... I just wanted to know. I think there is more to do.

[slide showing Option 6 - Compound and Indiv...]

Stan: So - have CBC. This is a modifier element specified at entry level.

Harold: A modifier?

Stan: Yes. Says mother or baby. A patient-identifier. We are showing, if I retrieve a Hematocrit out of here, in the retrieval-query, I have to return the query-modifier with the results.

Harold: I don't think this sticks with current definition of entry.

Ian: Entry is a contract that says... those have to be resolved. Query-safe. Have to be resolved at run-time. Maybe 13606 have different rules, but as long as agree that modifiers are...

Stan: If define... result without... is like a partial indivisible entry. The real indivisible entry is that partial thing plus... partial thing. Plus what it is inheriting at panel level is what it is inheriting...

Harold: I can see what you are trying to do, but I think we need some precise verbiage to make sure we are consistent. So not saying one thing to....

Stan: That is the implementation to make sure it is...

Harold: I would be happy to do this with you. I am concerned without some rigor here... I would be happy to talk with you tomorrow. I am still a little worried. Fairly hard... Looks like a topic - puts fear in hearts of... called "context conduction"...

Stan: It is, but... a difference... If this is a modifier, it modifies everything inside it.

Harold: Date?

Stan: Is a qualifier.

Harold: So can retrieve without date?

Stan: Non-sensible, but...

Ian: When have mixed subjects, if indivisible entry was mixed, so let's say Hematocrit...

Stan: 2 styles. Can say... or say subject is at this level, and when retrieve, need to say fetal or mom. So, 2 styles. If... then can't be at this level. Must be... or is unsafe. Or get into complicated rules... get local or... Else you get into context conduction.

Stan: I need to go. We could set up another time. Or could do by email.

[end-of-meeting]