CIMI MTF Minutes 20130110

Jump to: navigation, search

CIMI Modeling Taskforce - Meeting Minutes

Thursday 10th January 2013 @ 20:00-22:00 UTC


Linda Bird

Galen Mulrooney

Mike Lincoln

Mark Shafarman

Rahil Qamar Siddiqui

Jay Lyle

Joey Coyle

Stan Huff

Dave Carlson

Eithne Keelaghan


  • Weekly News & Updates
  • Planning and Presentation agenda for Scottsdale
  • Re-examine the use case for TERM_MAPPING in the reference model
  • Review of Demographics Reference Model
    • Scope of 'PARTICIPATION.performer' for Lab Results
      • Person
      • Subject of Care
      • Healthcare Provider (Individual & Organisation)
      • Organisation
    • Participation.location - separate cluster
  • Review of Demographics Comparative Analysis spreadsheet
      • openEHR
      • NEHTA (Participation)
      • NHS LRA
      • HL7 v3 RIM
      • FHIM (Federal Health Information Model) Demographics model
      • HL7 FHIR (Organisation, Person, Patient)
      • ISO13606
      • ISO22220
      • MOHH
      • DCM
      • IMH

Weekly News & Updates

  • Scottsdale January 18th - 20th 2013

Detailed Minutes

[Galen asks Mike a question about hosting a server]

Mike Lincoln: Yes - we have a workbench. Depending on requirements, might be happy to share with CIMI. I've talked with Catherine Hong, but...

Galen: Not sure what it would be, exactly.

Linda: We have requirements - Harold developed a set. [Goes to "Terminology_Tooling_Requirements" on the Mayo wiki]. I know Harold is planning to do updates... this is the basic requirements.... terminology ability and ability to reference...

Mike: Let me show to Catherine Hong - nurse informaticist - working with workbench. I think we should be able to help.

Galen: Also, management of CIMI terminology, if we are planning to use the U.S. namespace, then who would be responsible for doing upload of files. I think VA also uploads the U.S. namespace to the workbench - so already is doing this.

Linda: Not sure if we need this...

Galen: I thought Harold said - set them up... directly through the U.S. namespace

Linda: We went to IHTSDO for namespace...

Galen: OK.

Linda: Small crowd today - Let's get started. Draft agenda [on screen] - I moved 'Scottsdale' to the start of meeting. Plan is to discuss weekly news and updates. I was going to talk about reference model if enough people... and then demographics model.

[Notes: openEHR, NHS LRA, HL7v3RIM, FHIM (thanks Galen), the FHIR, ISO 13606, ISO 22220, MOHH, DCM]

Linda: All OK with draft agenda? OK.

Linda (cont'd): Happy New Year. Any updates?

Rahil: I've started work on the spreadsheet. I added participation functions and party... Started on Link relationships... Parent id... preferred terms... the parent ids. Something we did. I left the ones that are ambiguous.

Linda: There might be a parent... that they fall under.

Rahil: Yes - I left the ones that I wasn't sure about.

Linda: Thanks. Also - discussed with Mike Lincoln about IHTSDO workbench. He will talk to Cathryn and Tim Gromvell (sp?) at the VA.

Linda (cont'd): Next - look at Scottsdale planning. On Monday, may have presentation from John Gutai - his terminology tool. Stan - is Tom dialing in for ADL?

Stan: I emailed him. We put him on the agenda without consulting him. This was what he and Ian worked on...

Linda: Dave Carlson and Robert... Dave will present. Then talking about terminology authoring tools and... On Saturday - "relationship of CIMI to International Standards and Products". Then a presentation on the Registry/Repository, and then MTF working sessions. Any comments?

Mark: Question on registry repository? I can give an update on -> Health Ingentuity Exchange (HingX)> HL7 is one of the sponsors.

Linda: Stan?

Stan: That would be useful. We are trying to understand our options... HingX could be one of those. Not too much time for presentation.

Mark: I could give status update.

Stan: Yes. We put registry/repository because do we need a registry or actually need a repository?

Mark: OK - I will get update.

Linda: For Modeling Taskforce part of meeting... Michael has update. CIMI Reference model... also demographics model. Also analysis of existing approaches. We have Lab Results model to review. Any other things to review?

Stan: This is a big group for input. Are there other important issues apart from Modeling Task Force?

Galen: I am curious about status of RFI that went out last year.

Stan: The proposal is - work with HL7 to manage our logistics. Covered Friday morning. We think we'll have a proposal about what HL7 will charge. We'll meet - the Interim Executive Committee - meet with Chuck and we'll give you an update. We will also be discussing the RFI, around the registry/repository tooling. We need to make a decision about what we want to do.

Linda: Other modeling topics?

Galen: You mentioned sample models today on this call, but sparse population. Might be better done in Scottsdale.

Linda: I agree. Maybe make a bit of progress today. Other input? OK - we'll work out rough plan for sessions. I don't think enough people to discuss Use-case for Terminology Mapping. Defer decisions till Scottsdale, but looking at Demographics.

Linda: We took openEHR Reference Model Demographics as is.

[Linda shows Diagram]

Linda: If we took the same approach with demographics, to remove any clinical content from the reference model, then we could simplify this model quite a bit - for example, we could move 'capabilities' to the archetypes, rather than hard-code this in the reference model. Also, contact and party... keep this in mind.

Rahil: Is there a definition for capability?

Linda: Yes - part of openEHR.

[shows demographic_im.pdf]


2.2.11 Capability Class


Purpose Capability of a role such as her modifier, ...


Galen: My gut reaction - probably something we would want to archetype. But patient... might want to call out patient as reference model concept, whereas specialties could be done on archetype level...

Linda: Yes...

Rahil: We have this as part of demographics - what CIMI calls demographics... Qualifications or positions that a person has...

Galen: What do you mean by position?

Rahil: Role - what is in someone's contract. If someone holds position of General practice or Gynecological practice or special roles... May have more than one role. What is in contract? If credentials are that or... So - part of demographics, and not part of archetype.

Linda: Yes - but LRA in England - is country-specific. We talked about that with the LRA... countries might want to specialize the demographics models for their own country-specific requirements.

Rahil: As long as can extend... But I don't know what benefit to have in archetype as opposed to... instantiation. Reason?

Linda: Yes - with Reference Model, is hard-coded. Whereas, in archetypes, more ability to change.

Galen: So - things that are always true - put into Reference Model. But if conditionally used...

Linda: Main parts of the Demographics reference model that we need to fix are the Party, because that is what is defined when we have a participation. The Participation has a party and that points to the demographic class 'Party'. Parties can have relationships to other Parties. And we have the actor and the role subclasses - similar to HL7's 'Role' and 'Entity'. An Actor also has specializations 'agent', 'Organisation', 'Group' and 'Person' … and the model includes the Capabilities, or professional qualifications. And party identity... seems to be how the party is referred to. Tends to be the name... Whereas the identifiers of the party are included in the archetype details.

Galen: So - role is a subclass of party. I have never liked the party pattern, but I will do my best to be okay with this. I'm OK with Role being there, but capability should not be in the Reference Model, but in archetype... I thought there would be a contact type... not just validity date that is important to have.

Stan: If we put patient in, would that be a specialization of person?

Galen: No - of role. I think HL7 did well. I don't care where Doctor went to medical school and his taxID, but if he is the one on the table, he is a patient. So depends on role. Making these roles subclass of Parties. My first object-oriented program I wrote I did this, and it causes problem... and if it has a subclass of person, these are mutually exclusive. Interesting that Role is subclass of Party. And I would make patient a subclass of role.

Mark: Also, I might be a consultant, and I might be participating as a emergency doctor in an emergency situation.

Galen: Yes - I wasn't thinking of licensed role, I was thinking of role in act.

Linda: These sound like Participations in the record. The participation type can be “Discharging Doctor” and this participation is performed by a Party.

Mark: So that contains party.

Linda: Allows use of HL7 RIM... Role linked to party... or actor. But actor can be involved without going through role.

Mark: Not sure of that because you always have a role in every participation.

Linda: I talked with Graham about this. FHIR has found that sometimes using roles in a participation seems artificial... and it may be more appropriate for a participation to be directly performed by an entity... because the role is implied by the participation.

Galen: It is usually the participation, not the role. So participation of scoper may not mean a thing. Done on participation - not role.

Linda: My understanding is will always have a participation. But role may be optional. Graham has good example.

Galen: I go from act to role and skip participation.

Mark: If you had entity participating directly, that has interoperability problems. So if we do that, should create role - that is an unspecified role.

Linda: You are right.

Mark: Also with model derivatives like CDA - semantically interoperable.

Galen: Those tend to be hard-coded. If from HL7 - rip out. If to HL7 - add. It's when it is variable, need a spot to add.

Mark: Can't do mapping in computable way, need a human. Nice to be computable.

Linda: We can define the assumed role for each participation.

Mark: Yes - then can do transformations automatically and save labor.

Linda: Go to analysis spreadsheet - person. At LHS(?) - Lab specific ones. Location: IMH... patient, performing, provider, service in NHS - a place entity... in v3 RIM - the place... Address - IMH...

Galen: In FHIM, the address is a data-type.

Linda: Is that v3 data type?

Galen: A simplified version.

Linda: Available on what you sent me?

Galen: Yes.

Rahil: We have address as a data type as well. Not part of entity classes - part of role.

Linda: And you are using ISO21090 for this?

Rahil: Yes.

Linda: For Electronic Contact, we have compared the models rom openEHR, IMH, EN13606, ISO22220, ISO21090/HL7 v3.

Galen: And the FHIM is schizophrenic. Like HL7. Can say home phone or work phone. But I also will model... telecommunication... is hard-coded. There are two approaches. You can be specific... or you can have a bag of phone numbers or IDs... and have a type-code to tell you what it is. Reason is... not sense for hospital to have a cell phone... So no way to be explicit about what you are expecting, so I model explicitly.

Linda: OK - still allows you to fix for certain fields... to say...

Galen: OK.

Rahil: In LRA - it is part of the demographics model, the ... identity. That has the telecommunication attribute. Uses the... from ISO29090(?).

Linda: I used... LRA Demographics model [shows on screen]. OK?

Rahil: Yes.

Linda: Next is Party. A lot of models that looked at Party. I particularly looked at... and also the v3 RIM. Common attributes between Role and Entity. I think it may make sense to put these common attributes on party instead of both Actor and Roles.

Mark: Is a question of having meta-data so can do interoperability... going from one structure to another... don't want to do by hand.

Linda: Yes... With actor, more specific archetype and models. In terms of generic entity actor - I looked at LRA, HL7 v3 RIM, ISO13606, and openEHR.

Galen: The FHIM does not have an equivalent.

Linda: OK. The Person model - openEHR has person archetype, and models for person name, person- identification ... IMH has individual person - Is that OK Stan?

Stan: Yes.

Linda: HL7 has... ISO 2220... has...DCM has... FHIM (Galen) has... NEHTA... [Linda showing spreadsheet]

Galen: One of them has subject of care. I consider that a role, so is... out.

Linda: Good point. I also have included subject of care, so should remove those attributes from Person.

Rahil: We don't have a separate subject of care role, so is generalized.

Linda: OK. Organization actor... openEHR... FHIR has .... class.... HL7 has ... class. ISO13606 has organization class.

Galen: In FHIM, the organization class is in the provider package. Should probably be in common, but...

Linda: OK. Next is Role. LRA has role class, and v3 RIM. I looked at those for common characteristics for roles. Any others?

Linda (cont'd): For person-role, only LRA has this. In LRA, person-role has specific characteristics for roles played by Persons. I will need to check if this overlap with the Person archetype - however I'll leave this here for the moment.

Linda (cont'd): Subject of care... was going to... specialization of person or a role. In... In v3, subject-of-care is a role. I have added as a Role. Anyone have issues with that?

Rahil: Unless we have certain characteristics of subject of care role, we can achieve with person-role. So... have person-role be a subject-of-care role. If specific attributes that we will require for subject-of-care...

Linda: We have a healthcare provider, which is specific to subject-of-care...

Mark: Participation vs. role. I can be patient participating as emergency tech if I am an emergency tech. If role of defining something as... if mean only participate as subject-of-care. Maybe patient will be part... if family member. Do we have other types of subject-of-care, like family or group, the?

Linda: Different - subject-of-care and patient. FHIR - have been calling it patient, but will change to subject-of-care. And in the analysis, some of characteristics for patient might be better to move to person or general person role. I have done the mapping directly, but we may want to look at it more closely.

Mark: Sometimes a patient is in a record system in an administrative category, "This is my patient". And sometimes is a participant. Need to be clear about which distinctions we need to make and how general.

Linda: I agree and perhaps need someone to do... Do you have definition for these, Mark?

Mark: If you consider subject-of-care - a participation. Patient is always an individual. But you may want to have a group there... do we ever extend to a group and not a single patient? If have a licensed physician who is a patient... have to make sure.

Linda: We did look at subject of care being a group. This may be why FHIR is moving from Patient to Subject-of-care.

Galen: Mark's... If person is licensed physician and you care as... then is a different role. But in a different record...

Mark: I am in violent agreement. At a person level, can have different roles.

Rahil: Also have... say subject-of-care... is singular, similar to patient. When talking about groups...

Comment: Like a cohort?

Rahil: If subject-of-care is a cohort of a few people - the equivalent is the patient-groups. I don't know what the difference is between 'subject-of-care' and 'patient', and between 'subjects-of-care' and 'group'. In LRA, 'subject-of-care' is an individual... and not for a group.

Linda: Other point?

Rahil: Yes - looking at sheet. All the attributes of subject-of-care role? Nationality and citizenship.

Linda: Yes - from...

Rahil: Not be an entity and not a role? Things that may change develop on kind of participation - go into role class. So... capacity will not change whether patient or... but their capacity may change in reference to contact-times.

Linda: Role here is distinct from participation since can have... Also... in intention was... perhaps may be more appropriate to map specialization on person and not role. Since probably race and religion are more generic than role. Stan - any comments on patient in IMH?

Stan: Harder than we thought. I support Galen - should be part of core model. But as soon as got past unique identifier, everywhere we went... Our mistake - implementation details - how find patients, foreign key - further complicated. I would be for minimum set of attributes in core model... and everything else be archetype-able attributes.

Linda: What about patient versus subject-of-care... and a person?

Stan: We always think of patient as person whose record we are putting this into.

Galen: In respect to race and ethnicity - I put on person model. Stay the same regardless of role. Even if I think - don't care what race my doctor is... I keep in model. For next of kin - only need... I tend to put that info in role class. But for person using person class - put DOB, and things applicable to all roles.

Stan: Model of Meaning... Even though ethnicity stays the same, it looks like a Blood Pressure or serum sodium - we have multiple ways to get this - we can get different answers and... each time you ask it is like someone else made the measurement. So we made it archetype-able.

Galen: I'll buy that. But for race and ethnicity, we use these for reportable purposes. Just like gender and biological sex are different, I would have defined an administrative version and one for clinical purposes.

Stan: We get from employer or insurance company or the patient when they show up at the clinic. So if more than one person who can enter the clinical thing... Need to be a referee to decide which one is true... So we treat it as an observation. even though one race and ethnicity, in actual practice, you aren't sure whether this is an omission.

Stan (cont'd): Also wanted to be a search key... but is a different question than how you want to model.

Galen: I hear you. Another thought - we came up with notion in FHIM - temporal data type - person name... is the most current info. We flag those things that might change. And race and ethnicity we flag as temporal. So if... enters, then says you are... and audit trail. One way of handling is to flag it... issue... administrative and... your biologic...

[Jay?]: Like CDC - snapshots... An interface problem... how to present that to someone.

Stan: That is the way we are going... with BP, multiple BP's or multiple hematocrits. If need, then apply logic - take last one or... so to decide which one to take.

Galen: Difference is, though, marking as temporal has most recent in... rather than looking all of them up. Only look at for history reasons.

Stan: An implementation issue. I have race and ethnicity modeled in temporal class... A logical thing.

Stan (cont'd): For our purposes, though, I don't think they are too different. I can ask for race and I don't care how you did it.

Galen: Yes - both valid ways of doing things. I am not implying... but may be inferring that one way is better.

Stan: Yes.

Galen: When show subject-matter expert something, they think they don't have to mention since is right there. But modeler thinks they have it all and go off and build... inadequate. So, I want to take clinical knowledge and not bury somewhere... I want... something front and center before programmer goes off and implements something that is nonsensical.

Stan: Yes...

Galen: Race and ethnicity - would want to record. But blood test - want to make this a lab test or...

Stan: Everyone said - we need date-of-birth, race, ethnicity. But further down list... could not say that. People would say well - what about military? What about birthplace? So could agree on top 3, but... a personal bias about whether you include or not. Because of how set up in U.S., and need to report... but this could change drastically with genetics... it is the gene-type we need... Things that start... firm ground... but eventually get to quicksand where no agreement about particular attributes. If some we can agree and make attributes of class, but the less important.

Galen: The beauty of... eventually we want to be specific about everything so folks have that explicitness. What we are looking for here today - how much is in Reference model and how much is archetyped. The RM needs to be a minimalist as possible. So race and ethnicity - not be in RM. But role - important enough.

Stan: I agree - would like to see patient or subject-of-care as first case attribute.

Linda: We also need to decide whether some properties that have been place in the patient model, like nationality, should be reported on person... and allow them to be recorded irrespective of the role.

Galen: Example: I have EHR system and payroll system that I want to use this for. Contain attributes I may only care about if is patient... maybe race and ethnicity...

Linda: Maybe if we can't decide... we should provide them in both entity and role.

Jay: "Provide" - what do you mean, in model or archetype?

Linda: In archetype.

Stan: Not obvious - implications of model... if I choose to make attributes of entity or role. Not clear.

Linda: Having them on role allows difference for different roles.

Galen: ...Requirements for placing in right spot... reference model or archetype. At first... race and ethnicity... on the person. But not sure if on Reference Model. So, #1, things that are intrinsic to person - we want to model on person... not put in Reference model, but in archetypes.

Linda: First in Role or actor entity?

Galen: I would argue - Language, Race, Nationality - belong on person. Disability is tricky. Organ donor type - not sure.

Mark: These are questions that in v3 - some are clinical observations about a person. If not, then sometimes grouped under lab test and sometimes person, and won't have 1 place to go. Also, some have more than 1 disability... so, on attributes level... things that identify the person...

Galen: Identify the person, but... Agree?

Mark: That is an edge case - not all will ever agree. But they need meta-data, so can be exchanged.

Linda: We need to get down to specifics. Maybe some edge-cases, such as confidentiality... maybe at subject-of-care level there are some attributes specific to that, such as diet, which we would not move to person... whether we put attributes on person or role, we have...

[Linda shows spreadsheet]...

Linda: Moving on to next... provider. Can be individual or organization. The IMH model for provider may be either an individual or an organization. And others do the same.

Stan: IMH - the patient contact is like an encounter, not a person-contact.

Galen: The verb, not the noun.

Linda: Oh.

Stan: We have provider back in... but I don't think we have any yet.

FHIM: In FHIM - have healthcare provider. I generally point to healthcare provider when... But when... point to an individual provider.

Linda: In IMH, you have a provider class.

Stan: As you can see, it is a minimal amount of stuff.

Linda: I thought we...

Joey: Can you show your model again? This model is broken. Last item - an error with the old...

Linda: I noticed that...

Joey: We have a new version, if you go to...

Linda: Please email URL.

Joey: OK - it is in test mode.

Linda: Available?

Joey: Yes.

Linda: Provider was to cover individuals and organizations?

Joey: Don't know that answer.

Linda: [Looks up on CEM Browser]

Joey: Not valid - can't have items and data. So that is a broken model. So each of those will have depths to it... If click information node on provider-type.

Linda: OK - very useful. Thanks, Joey.

Linda (cont'd): OK - for generic role of healthcare provider. I looked at IMH, the green EHR model. Rahil - the LRA model... had specialty code. Others had under provider. Is this what you only provide for individual providers?

Rahil: Yes - domain-level provider. We have them categorized as... identified healthcare individuals or healthcare provider who is a person... We have these catalogues... these hold all the identifiers...

Rahil (cont'd): Let's say there is a class for that role... We have... extensive domain for participation... I can provide to you. The main specialty code has to do with individuals. In NHS - the specialty under which the person has participated in that record... Distinguished from their contractual role... specific rule... maybe not be a straight mapping. I think in your mapping to the provider... domain model... I will provide.

Linda: Thank you. So looking at Individual Healthcare Provider...

[shows spreadsheet]

Stan: I don't know.

Linda: ... In the v3 RIM... a notion of employee... different from Healthcare Provider... Mark?

Mark: In RIM - same kind of distinctions. The employee is related to a particular organization and has certain privileges. The employee may have additional role of Healthcare provider. What generally will be... is the certification that the person has... often in the States, doctor works in hospital but not employed by hospital.

Linda: OK....

Rahil: We have... Not sure if was direct match, but... The code attribute with type CE. The spreadsheet... extreme Left hand corner...

Linda: I did not use column A, I used...

Rahil: Yes - not employee in published model.

Linda: ...not yet looked at NEHTA and... For healthcare provider organization... openEHR... needs to inherit the attributes of an organization.

Linda (cont'd): So that is an overview.

Galen: If you send me link to spreadsheet, I can do mapping for you.

Linda: I sent link to Google drive...

Galen: Not using Mayo wiki?

Linda: I am not good at linking whole documents to the wiki... Harold is linking... I can send you Link.

Galen: I don't have Google drive installed.

Linda: Just follow Link... should wrap up meeting... Thanks Galen. Other comments?

Joey: When looking at models on website - can see... when click there... open it up... under the key... Need to find something that is a value set - will get all the values. Not complete - just examples.

Linda: Thanks everyone. Safe travels to Scottsdale.

[end of meeting]