CIMI MTF Minutes 20121101
- 1 CIMI Modelling Taskforce - Meeting Minutes
- 1.1 Attending
- 1.2 Agenda
- 1.3 Weekly News & Updates
- 1.4 Modeling process
- 1.5 Comparative analysis spreadsheet progress
- 1.6 Detailed Minutes
CIMI Modelling Taskforce - Meeting Minutes
Michael van der Zel
Rahil Qamar Siddiqui
- Weekly news & updates
- Overview of HL7 CCD and v3 Lab Results Models (Mark)
- Continuation of Laboratory Results Comparative Analysis and Modeling
Weekly News & Updates
- The Netherlands Meeting - confirmed (Sunday 2nd to Tuesday 4th)
- Other CIMI meetings:
- UML Taskforce met on Monday
- Terminology subgroup met on Tuesday (special guest: Alan Rector)
- Paper presentation by Harold
- Glossary subgroup meet every 2 weeks on Tuesdays
- Meeting next week: CANCELLED
- Analyse clinical models submitted (with value sets)
- Identify maximal set of data elements
- Remove 'out of scope' data elements (Style Guide)
- Select appropriate CIMI Modelling Patterns(Style Guide)
- Define CIMI model structure (Mindmap, ADL, UML)
- Add Terminology bindings
- Meaning (nodes, node relationships)
- Value sets (maximal set from submitted models)
- Add Example Model Data Instances
- Technical Validation
- ADL, UML
- Clinical Validation / Review
- Confirm mappings from submitted models
Comparative analysis spreadsheet progress
Use cases include:
Issues to be followed up on:
- Reporting priority vs Testing priority
- Test Procedure/method included at both panel and test levels.
Weekly News & Updates
Linda: Mark - to present v3 models today. Please let Virginia know if you can attend the Netherlands meeting. The UML Taskforce meets on Mondays. Dave - can you report?
Dave: Tom will work on initial set of ADL. Then build a lab template. Want UML - baseline - then use to work on patterns.
Linda: Terminology subgroup met Tuesday. Harold presented Alan Rector and Rahil's paper. Alan Rector was at the meeting. Next week Rahil will present LRA approach to terminology binding. Glossary subgroup meets every 2 weeks. Will next meet on Tuesday in 1 Â½ weeks.
Linda (cont'd): Another agenda item. Next week I have a conflict. I am landing at Heathrow [and it is] hard to attend the meeting. I suggest that next week's meeting is shorter. Start Â½ hour earlier and end earlier.
Peter: In Heathrow, you need to walk for miles to go through security again. I barely met my flight. You won't make it.
Linda: You're right. New plan. I won't be able to attend. Let's cancel it and make it the week after. It takes 2 hours to get through Heathrow. Any more updates? No? OK - screen to Mark for HL7 submission.
Overview of HL7 CCD and v3 Lab Results Models (Mark)
Mark: I am going to walk through a number of basics of CDA. How... got to CCD and how... got to Consolidated CDA. All is available on HL7 website. With new policies, should not be hard to get to. If not part of [HL7]... I can get you permission. I want to show basic CDA model...
Mark: On left-hand-side, the basic documentation architecture, who wrote it.
Mark: Pieces inside sections can recurse as much as possible. Allows nesting. [Zooms in] Once into body of document, can have as many subsections as want...
Mark (cont'd): See section and body choice. Within subsections can have as many entries... very close to openEHR and EN13606. Tell me when see clinical statement on screen...
Mark: These can be grouped into subsections. We see observation at the top.
Mark: Basic Observation Class... and mood codes... and it has a reference range.
Go down - show bottom half...
Mark (cont'd): See - second box up from bottom is Organizer. If have panels or... can be organized in this. Now - go up... see procedure - this is where procedure-type goes. Up a little on LHS - yellow and green thing for entities enrolled.
Mark: See drug or other Material. See a blue specimen, so model can deal with specimen. The way v3 works - this is a huge model... [and] is tightly constrained.
Mark (cont'd): Switch to text in CDA. This is some of the detail. For example... a derivative of Observation Class. Mood code - Occurrence of event. The instance of an act... The meaning is given by a number of codes for structural codes in HL7... something equivalent to SNOMED code.
...where there are occurrences... says this is WBC or... panel or something.
Mark (cont'd): Now we are in the Continuity of Care document. This is a set of formal constraints... CDA-2. The clinical statement part for results is greatly constrained. We took larger model and scoped it down - procedure and observations. In CDA model - certain constraints. In the bottom, have the organizer for panels. Off to left we have specimen... at top, have entity Relationship. Going down further...
Peter: Will you send this out?
Mark: OK. Section 3.1.13. You should get this. Template identifier goes along with this. If it is an organizer - going down - bunch of conformist...
[Mark reads screen]: An example - Result organizer shall contain at least one organizer/id... The value for "organizer/code..." Gone from general clinical statement to... constraints.
Says what an instance of this should contain.
Have similar for procedures...
This is an example of level of detail.
Josh: Is there a formalism behind statements like this?
Mark: I can help find the formalism for you. The other thing that makes this easy to implement the ... project has come up with a ...each table-driven way to do this.
Jay: ...tool also does this.
Mark: So we have 2 sources. I want to spend 2 minutes tracing history. Group in U.S. called HITSP. They did CCD specification for [HL7 Clinical Document Architecture - CDA]. What you will see is measure and lab results. A definition of value-set in terms of LOINC. If we scroll down - can see others.
Mark: Extensional and Intentional Value sets are defined... so CCD document can be used to... in a formal way meaningful use...
Mark (cont'd): A number of these things have been put together in the "Consolidated CDA". One of the nice things - they put together a spreadsheet. You'll see a hierarchy, e.g. Allergy Section will contain a... Allergy Observation... See Results, Observations, Constraints, Overview. Might want to start with CDA. Just to give you one other picture from consolidate CDA... This is one of one value sets with the Consolidated CDA and its history, these are things CIMI could look at to see if we have missed things or CIMI has missed... So a lot has been defined... There are lots of formalisms. So before I go to V-3, any other questions?
Peter: You had cardinalities - reading the minutes from the last CIMI meeting - cluster vs. element. It has 0-to-1 or 1-to-1... It seems to me you can't argue if systolic goes with diastolic, but you can represent... different... So, rather than argue about whether element or cluster, can ask if 1-to-1 or...
Comment: The way you make unambiguous, you put a cardinality on it...
Comment: Not the way we do it in EN13606...
Peter: But this makes it subjective.
Comment: One way of defining, and everyone buys into this model. The idea of a cluster is more about containment, a tree cluster.
Mark: Let's do this as a different discussion. I would say, the condition of cardinality is important, in either case.
Linda: Please send reference to value sets you showed.
Mark: Yes. Zip files or... I want to go with HL7 next. This is 2012 Normative edition. So - go down to look at laboratory first.
Linda: This is model Canada is using.
Mark: This is general enough except for details. In laboratory world, we have one of these document-pictures [shows on screen]. Similar to clinical statement except has a lot more detail... and is just for results. You can see: Reference Range, and a precondition, and can see that this can be a component of an encounter or... 2003(?) normative edition.
Mark (cont'd): We're going from picture to message description. At bottom - have 2 rows of icons. So I can go back to diagram, or can get a Table view of same structure. You can see... attributes..... We are at... If scroll down, see specimen, record target, performer, etc. If I click on what is highlighted, will go to Normative edition. So - a lot of click-through abilities.
Mark (cont'd): We were looking at Table view. There are also a spreadsheet view, so this is same information done as a spreadsheet. Each of these can be seen with details arranged horizontally. So that is what I wanted to show you. Also, for XML, there is a schema view of this information. [shows XML] Here is the consumable piece, and so on. So can get to all this level of documentation just from this HTML document. I also have a tutorial for this. If people want to see. Consolidated CDA and V3 - derived from same model. CCD and CDA - more tightly constrained. Mostly thought of as implementation guide in V3. Either way, valuable resources for Cimi.
Ian: Does new CDA... have...?
Mark: Allows you to specify a few more things more completely and with better constraints. Will also allow some additions. You should be able to use CTS2... in either, to create value sets... terminology binding... Harold can tell us about [this].
Rahil: When do we choose when to use V3 model or CDA or CCD - criteria?
Mark: Depends on level of detail needed. CDA level of detail - good for clinical reports, but may not have detail for transaction level. So, for example, if have lab - I need to have updated versions and need to track by patient ID's or... ID's. At document level, that information is not always needed, so has less of this information. On other hand, document level can reference... has an instance identifier that can be tracked. Sometimes useful. If get... about lab being for another patient, then I can delete.
Rahil: So V3 is more detailed model if want entire clinical process in that laboratory... from primary care to... could be captured? Whereas, ...[CDA] is a summary?
Mark: Yes. That is why V3 is more detailed. My Dad was a GP in the States...we're... primary care...
Rahil: When do we use "Consolidated CDA"?
Mark: We use it now. Can be downloaded. Used in at least 2 or 3 ways. We have meaningful use criteria. A number are defined in terms of consolidated CDA. So physician practice groups/hospitals that want to be certified for Meaningful Use 1 and 2... Also useful for sharing data between practices. Hard to say how widely used, but there is HIN [Health Info Networks] that are using this.
Continuation of Laboratory Results Comparative Analysis and Modeling
Linda: Thank you Mark. Let's move on to Laboratory Test Result Model [shown on screen - same as last week].
Linda (cont'd): Last week we considered FHIR, Results4Care, IMH... I need help with EN13606 mapping. Gerard - can you help with this?
Linda: So, last week we looked at first few elements. Test performed. Diagnostic type. Specimen. Priority...
Ian: I would argue. Local diagnostic type. By chemistry results - tend to be done by chemistry service. It is who did it. Immunology test - might be done by immunochemistry. Others - by hematology. So... rather than type of test.
Stan: So - values for this - hematology, chemistry?
Ian: Yes - in most cases - it is desirable. But it is not an absolute. May be performed by different service. Would see as a set of local terms...
Stan: The system could refer to a knowledge source that could have that knowledge about how the test name relates to the diagnostic type. That could be local knowledge outside of the model. And if it is entirely local, then it would be interoperable, with the relationship between test names and diagnostic types stored as local knowledge. So, lab-systems can display a wizard with this information. You use these things the way they are... If they are local - you could add back the derivation rule for the diagnostic type.
Ian: You reassured me with your description... based on local knowledge. I would argue - it should stay in. If not, how are people going to express it?
Stan: You need the functionality, but you do it a different way. I have no problem leaving it in.
Linda: So sometimes derivable and sometimes not.
Ian: Yes... override...
Josh: Local knowledge still interesting to outside organization.
Linda: Yes. Moving on to Test Method. In the NEHTA/openEHR - slot or cluster, whereas IMH - single item/data... It is included in test procedure. To include in CIMI model, we need an example we can base on.
Ian: I think you can assume has a text description. I think Heather and Graham added... I am not as familiar... If you wanted to know what is in these, you can assume it is a bit of text.
Linda: And this warrants a slot?
Ian: [You are] probably asking the wrong person. Ask Graham.
Linda: Let's ask Graham. I am... to just leave as test...
Michael: I looked at our model...
Linda: In the Results4Care model, the Test method is at the individual test level - Whereas, we're now looking at the test method for the overall test group (or panel) level.
Mark: It's an attribute in the... observation specialization and has a data type allowed to repeat...
Stan: I think you are right Linda. At the panel level it would make sense if the Test Method named the instrument used. So if put on Chem20 or ... different for Hemoglobin than Hematocrit than Red count. So instrument could be one thing recorded. Second point - important that it is coded ... enzyme level at room temperature versus at body temperature... needs to be coded to support logic around the use of the value... That might be another difference. At individual analyte level it should be coded - Whereas at the panel level it may be free-text.
Linda: OK. Thanks. Text in openEHR CIMI data type - can be plain text or coded. But you said should always be coded.
Stan: Yes - would want constraint to say coded...
Linda: Put test method at panel level as...
Ian: I agree with Mark - make it multiple if we are going to do this. If Graham can come up with examples to argue for panel then...
Ian: Text v.s. coded-text. I agree with Stan. But with labs, you get what you get. Can always constrain, but... I wondered if coded-text is always enforceable.
Stan: Use-case. I think it is right. True interoperability. Some people won't be able to achieve. But if we want highest level of interoperability... But is legal to fall short of that for other reasons.
Mark: Interesting description. This method code provides... This implicit information does not need to be specified. Difference between susceptibility methods, so method code used as additional qualifier... So they are saying it is often explicit or...
Linda: Thanks. So that was Mindmap. Need to move on to who, what, where... So "what" - Result-group in openEHR... So we have Result-group identifier in Result4Care. Question for Mark - 1-to-many?
Mark: Look at model [CCD model]. You only have one at a time, but identifier may be multiple, if 2 systems are talking together. Not mean that organizer can have...
Comment: A single instance can have multiple IDs.
Question: So anything that has an identifier can have multiple identifiers?
Mark: Probably example is patient-identifier.
Linda: So is 1-to-many. In NEHTA... So we need to make it 0-to-many...
Stan: May be OK, but... one of the primary uses of field [is] if I get an update... So if there isn't an identifier, implies some other mechanism to do updates. I think we should require it. So Use-case where not... is legal, but... How can it work if no identifier?
Mark: Yes - I agree with Stan. Patient safety, especially labs. Maybe work-arounds, but interoperability...
Linda: Depends on if we are considering model submitted.
Linda: Lab Report is the resource. The Result Group is an element inside the LabReport, and this Result Group does not have an identifier...
Ian: Under request-detail, there is 'Laboratory test result identifier'... in NEHTA model, this equates to accession#.
Linda: But this is not mandatory in FHIR or HL7.
Ian: Neither is it in the NEHTA model.
Linda: Yes - for some reason I had the protocol, but Heather explained not...
Linda: Yes - we discussed... But this is for Result-group or Organizer.
Stan: I don't know if we can require... I think we need to... achieve highest level of interoperability. It is allowed, but not [required]. I think it needs to be these.
Linda: Similar to what we discussed... cardinality... one to support legacy system and one to support the 'goal state interoperability'...
Josh: Elements that are required, may not be present...
Ian: I asked you to put back in. Result-group identifier is for... result. Under that is subgroup. That is what we don't have. In this case... Receiver-order identifier... Same in FHIR. I think I was mixed up order and test results identifier.
Linda: Talked to Heather last week - misplaced. You're right - identifier... So whole test is uniquely identified. So FHIR's Result-group is...
Mark: 2 questions: Someone looking at test results. But if I get new lab result, it is a patient-safety [issue] that I apply it to the right thing. Even if I say - throwing away and replacing, I still have to go back since decisions might have been made. So need to identify... we can talk about different ways of achieving this, but... patient safety.
Ian: Many other places when we will use models... in messaging...
Josh: Takes detailed guide...
Mark: Not arguing. But it is a requirement for certain Use-cases.
Ian: Mark - can you clarify - do you have identifiers at subgroup-level? Or just... Abstract subgrouping... Not clear whether identified in HL7.
Mark: Would be an implementation guide - whether... have unique identifier. So we need... Use-cases. Is a Use-case issue that you have to say...
Linda: All organizers can optionally have identifiers in CCD. Must have identifiers at each level... I read organizers will contain organizer ID.
Mark: CCD is a document-architecture. If I have update, I need to know how to update.
Linda: So I understand, but we have multiple Use-cases. Do we need to just support a single use-case, or can we be use-case independent and then constrain to a specific use case?
Stan: Yes - this is an important issue. Do we want to support use cases in transactional environments - where we want integrity - or what Ian talked of, that is, use-cases where they don't care about the identifier because they do not receive updates. So, we can retain what is common between these two use cases in the base model, and then define template models for each of these 2 variations to the model. 1 with identifier mandatory, and the other with optional identifiers for read-only organizations.
Linda: I think this is consistent with approach. Use-case independent with model and then...
Stan: I am not sure... What we are doing here... End up with archetype and template on that.
Linda: Yes - archetype and constraints. So Use-case independent model... We can start with a cardinality of 0-to-many and then constraint this for specific use cases.
Rahil: Yes - this is exactly like LRA... equivalent domain. When we come down to Use-case constraint... when we mandate identifier.
Gerard: I support... being generic. In 13606, reference model - must be generic - generally as applicable as possible.
Stan: There is a principle here - when we want Use-case independent, the model has the elements needed by any Use-case... and then model... we constrain.
Mark: Consistent with... CDA core-model - general... apply... I think we are all in agreement.
Linda: Next - Result-group name. In FHIR - 0-to-1. In NEHTA is text - 0-to-1. In Results4Care - question mark. [shows slide of nfu] Michael - do you know?
Michael: I think might be results...
Linda: In HL7, I assume organizer-code. Is that right, Mark?
Mark: Yes - that would be organizaer-code.
Linda: Most are 0-to-1. Are all comfortable?
Stan: I can't see [he is on the phone]. Can you give an example?
Linda: Complete Blood Count - Result-group name. A text field that we can constrain to be coded.
Stan: Would not have to have a name? Is that what is implied?
Ian: In openEHR this has a cardinality of 1.
Stan: I would think it is required.
Mark: I checked with CCD - it is required. May be selected from CPT4?
Linda: I am looking at the NEHTA model. Where do I look for result-group-name?
Ian: Scroll up. Top level test name - that is the one that is mandatory. So we are down in Result-group...
Linda: So Blood count - would be at this level?
Ian: Result-group would be... It is there,
Linda: Yes - result-group-name.
Ian: I think idea of group-results is exotic. Key thing is test-name.
Linda: in NEHTA model - refers to result-group having a name or code. Then CCD model - Mark?
Mark: The organizer code... will contain exactly 1 organizer code.
Linda: Michael? Result-group name?
Michael: Cardinality 0-to-1. I am surprised.
Ian: We need to take out protocol...
Linda: I don't understand.
Ian: Refers to whole test. So if CBC ordered - comes with request ID and given ID back. Receiver-order-identifier - used to identify the test.
Linda: Placer-order-id and filler-order-id - we will define in the Test Order model. But this lab-test-result-identifier we thought was given after test performed, so mapping might be correct. The whole entry.
Ian: Yes - to whole ordered test - not individual.
Linda: ID - mandatory in CCD, optional in Results4Care. In NEHTA - my memory is - they did use a generic model without constraining it. Stephen - are you online? [no answer] Ian - do you know?
Ian: I am pretty sure not compulsory. Go straight down to Results level. I think Use-case is for microbiology tests. Reason for high mid-range sub-grouping.
Linda: OK - need for Use-case agnostic level.
Stephen: Linda - I am here. Essentially, what Ian said is correct. Results-groups are for grouping only. Have to check with Graham to confirm. The second one - about the test method. The intent of test method, cluster - it's a whole lot of details we need to... Did not get around to defining in detail. We need to do this, so I cannot say now.
Ian: In developed one with... from New Zealand, but have something you could use.
Stephen: Please send to me.
Linda: And send to group please. So result-group identifier, Result-group name. Next is Result-item = the node that points to cluster slot...
Stephen: One test method to confirm with Graham - not necessary now.
Ian: I can give example of what Graham was thinking.
Linda: Good - next meeting. OK. Results4Care - the Lab test (is the container). HL7 CCD - an organizer-component, where the component is an observation.
Mark: I'll verify.
Linda: So, Result Item leads to an Observation result Slot. The next node in the test result section is Results-subgroup - allows you to organize the results into a hierarchy. In HL7 CCD, the model is one of the most sophisticated, as it allows you to define hierarchies to any level.
Mark: Yes - and requirements same at each level.
Linda: NEHTA - Result Group... Result-group can also have specimens - 0-to-many. I found in NEHTA model - under Result-group. Result-group has 0-to-many specimens. I don't recall this being in other models at organizer level.
Ian: I think it does. Pretty sure as far back as HL7 v2... Blood test - 1 specimen. In histopathology test - multiple specimens. Same might be applied to micro.
Mark: CCD - If need to deal with cluster... specimen observation cluster... points to any number of specimens... (?) can be organized within a grouping.
Linda: Need to incorporate an analysis of the HL7 v3 models. But it's good to know other HL7 models are compatible with NEHTA. So all comfortable? Next - Interpretation. This is in the Singapore model. All others - not at Result-group level.
Ian: Probably need for histo-pathology, but not put at this level. ...microscopic interpretation. I would leave out of base model.
Linda: Question on Exclusion Criteria. When is it used often enough for base model and when... for specific Use-cases?
Ian: I think that is when you let clinicians figure out. It is arbitrary. So - this is source of truth.
Linda: So is... agreement to... interpretation... Mark?
Mark: We have interpretation-code, but... at discrete observation level... is normal. They would have individual tests.
Ian: That is at... test level. Certainly multiple statements of interpretation. But I can't see at grouping level.
Linda: Look at last two - sequence...
Stephen: I found some examples where groups get... semantically... For example, test-result names = sputum microscopic culture and sensitivity... and group named to Staph aureus and other micro-organisms with colony and whether sensitive or resistant. So group might get renamed. But CBC... group not get renamed.
Linda: Thanks. Last two - sequence - Singapore and HL7 has a sequence# - right Mark?
Linda: 0-to-1. And last one is Status. NEHTA... HL7 has Status code on each of organizer... Running out of time. We'll go through this during the next meeting. I am not available for next week. Will send out spreadsheet. We will meet in 2 weeks.