CIMI MTF Minutes 20141113

Jump to: navigation, search

CIMI Modeling Taskforce – Meeting Minutes

Thursday 13 November 2014 @ 20:00-22:00 UTC


  • Linda Bird
  • Stan Huff
  • Thomas Beale
  • Patrick Langford
  • Joey Coyle
  • Deepak Sharma
  • Gerard Freriks
  • Jay Lyle
  • Robert Hausam
  • Sarah Ryan
  • Eithne Keelaghan

Regrets from Harold and Daniel Karlsson

Action Items

  • ACTION ITEM: Patrick to contact IHTSDO
  • ACTION ITEM: Send an email to HAROLD about generating the BMM re: getting rid of Structure Type in Item Group.
  • ACTION ITEM: Sarah to take the lead in addressing copyright procedures
  • ACTION ITEM: Gerard to provide Sarah with access to ISO documents
  • ACTION ITEM: Gerard to send list of copyright information to Sarah or Stan
  • ACTION ITEM: Stan to annotate the discussion on Pre versus post coordination of method, specimen, and timing in the CIMI preferred lab models so we can remember what was said.

Draft Agenda

Proposed Agenda Items

  • Brief update on AML submission – Harold Solbrig
  • Progress on lab models – Patrick Langford
  • Do we need structure_type in Item_Group?
    • We agreed to remove this from the RM
    • Ask Harold to remove ----------------------------------------> Action Item **************
  • Finalize the essential meta data attributes that need to be present in the models
    • Copyright information
      • Notes out to IHTSDO and LOINC
      • Plan to include as url
    • Import source, etc. for management of lifecycle
    • Acknowledgement of original source
    • Should not include realm specific approval dates, etc. in the data file
    • List of metadata items from the ISO standard
      • Gerard will send the ISO list, and his own list ----> Action item **************
    • Sarah Ryan will lead this group, will include folks from Oxford, also Ian McNichol
  • Pre versus post coordination of method, specimen, and timing in the CIMI preferred lab models?
    • Methods – we prefer post coordination of method
  • What is our approach for indicating which models are the CIMI preferred models?
    • In an independent part of the repository
    • Preferred by who, for what purpose, when
  • How do we indicate which models are part of the same isosemantic family?
  • Language attribute – add it to statement and remove it from the coded_text data type in the reference model?
  • Different coded_text data type for single code versus code expressions?
  • Representation of “leaf” models as model files or as a tuple?
  • Any other business

Detailed Meeting Minutes

Review Agenda

Stan: I have – agenda – brief update on AML from Harold. Might be others who can give more. Patrick on Lab models. Come to closure on structure type element in Item group. Talk about essential meta-data attributes we need. And in lab models, talk about issues of what our preferred style is for pre/post… coordination of models... and timing. And I can lead that discussion. Then we need to say what our approach is to say CIMI –preferred models. And despite all our talking about iso-semantic models, I don’t think we have a way to know which family of iso-semantic models it is... And single code vs coded expressions... And leaf models as model files or as tuples. So a big agenda. Other? Is this the right order?

Tom: Might want to add to agenda…

Stan: Can you tell me now… what I should add.

Tom: You’ve added.. essential meta-data… and the stuff about meta-data representing… importing and tool… At least those 2 things.

Stan: I’ve captured those in my agenda. Anything else to add? [No response] AML Update

Stan: So the brief update – the AML has been submitted. So this is very exciting and I am happy and proud of what has been accomplished there. And a large number people who have worked on it – Harold, Deepak, Tom. Robert Laurio, and some other folks Robert brought in, Jay Lyle, Cential(?).. That is an amazing accomplishment. I’ve talked to other people... I don’t know how long people have said – we need a way in UML of binding coded attributes... and that is one of the primary things the AML does. It does other things... Aligning the object – archetype model. And I am happy and excited about that. So – I’ll’ be glad to buy anyone on that team a dinner the next time I see you. I am really happy about that. Thomas, Deepak – comments?

Tom: I am coming out for dinner!

Stan: Good!

Deepak: There were a lot of us last week working... Harold, Robert… A team effort...I remember thinking… the only way this will be done is if Dave Carlson… and other... top officials.. That is what happened.

Deepak: I am going to AMIA next week – maybe we can meet.

Stan: Thank you – we look forward... A huge milestone. Move forward with implementations… making this a standard... not finished, but an important milestone. I am thankful for the work being done there.

Stan: So – Patrick – some progress to report since Amsterdam?

Progress on lab models – Patrick Langford

Stan: So – Patrick – some progress to report since Amsterdam?

Patrick: Yes- I do. I won’t be showing anything now – my virtual machine crashed… So – there have been a number of things I have been working on with the models. Mainly, bug-fixes I have worked on since Amsterdam. And a few things cropped up – Tom told me about. And also – a long data name and a short data name – and for any of the names that are too long, I use the short name. And if none, then I don’t create a (?LOINC) list for those. So we will have to decide how to handle these. The long are too long – cause problems for many file systems… We have to keep that in mind. The other thing – the UCUM-to-SNOMED mapping. And (Vicky?) got a group at IHC to start working on that. So they have … There are a number of people working on it… Sindaling(?)... and Jay Yeal(?).. They are waiting to hear back from ITSDO about a mapping file for SNOMED... that should provide most of the mappings we need... Otherwise they will have to take a look at this data… There are a number of… similar... Would have to take a look at that rather than use the mapping file.

Stan: When did you talk to them?

Linda: Do you know who they contacted at IHTSDO? Did they go through the email or…

Patrick: They gave me this info this morning? Should I ask them to contact you?

Linda: Normally, if send an email to… will go to content team… I can follow up with David or you can ask them to send to me.

  • ACTION ITEM: Patrick to contact IHTSDO ***

Patrick: OK – I will ask them

Linda: Thanks, Patrick.

Stan: I rummaged around and found a file that is 10 years old at least – has some of the UCUM codes or... codes… I don’t know how many of the mappings were there.

Patrick: Right here, Ian(?) sent me… I gave him the raw data... There are quite a few instance s of same meaning... not UCUM codes… printouts… what people thought the code should be… For example - Milligrams per hour… (?) There are a lot of codes like that. If the LOINC data just had the UCUM codes, this would be a lot easier. If we just had the... but these variations will be a problem. So that needs to be done.

Stan: They are working on it.

Patrick: Yes.

Stan: Anything you need from us?

Patrick: I have what I need to make progress. I will be bugging you, Linda – for IP codes and AT codes. But for now I am doing OK.

Do we need structure_type in Item_Group?

Stan: Good. So, next item – closure about structured-type, an item within group. It seemed that we were tending towards… no one is using structured type in modeling efforts. So my bias would say – take it out and we can bring it back and if we have a good use case we could document what we wanted to use it for. I am open for discussion. Thomas - ?

Thomas: Yes – I probably instigated the actual attribute 10 years ago. We had the idea in openEHR that clinical modelists would want to model a flat list or a tree or a table… ology… or... have tabular data… otology.. that structure type… It turns out that clinical data is unpredictable and everything turns in to a tree. You think you have a table, and there are other bits and pieces. Tabular data is pretty common in medicine. Maybe the thing to think about – the use case... If we were to write an archetype that represented… ophthalmology… or would we want it to have an indicator in the structure that would enable applications to more easily figure out…? They can probably work it out anyhow, because if ophthalmology application, probably knows… Those are probably the only ones from openEHR that have an identifiable structure that don’t get... by the fractal effect... So I would throw it out unless…

Stan: Other?

Linda: I would be happy to get rid of it.

Stan: OK – I think we are unanimous. Let’s get rid of it until… we have to bring it back. So mechanically – how do we get rid of that. Ask Michael?

Linda: At the moment I think that is Harold. Have to ask him if he generates the BMM. And Patrick manufactures that… So send an email to HAROLD about generating the BMM...

  • ACTION ITEM: Send an email to HAROLD about generating the BMM re: getting rid of Structure Type in Item Group ***

Stan: I know Harold has Enterprise Architect.

Patrick: I have it too but not the tool that Michael has for generating the BMM.

Stan: OK – we’ll ask Harold.************************

Finalize the essential meta data attributes that need to be present in the models

Stan: Next item – finalize the essential meta-data attributes that need to be present in the models... a couple of them the copyright information. The... in source... ...know whether this was machine-generated, so have to go back and edit the source... so not get overwritten next time someone generates the source.. And there is the whole issue of what other meta-data attributes we require in the model. We are probably not prepared to go thru a list of things now. But... a list of things being used in openEHR… we have a list from INTERMOUNTAIN… we have a set of... elements we need to include.

Copyright information: Notes out to IHTSDO and LOINC; Plan to include as URL

Stan: So, let's talk about copyright information. So, any models that include SNOMED codes need to have a copyright notice. Thomas said – could we put a URL in there that would point to the copyright notice somewhere else… because is a paragraph of information that is several sentences that we need to have in there. So would bulk things up in there. And if it changes... like with IHTSDO changing its name… we would have to change all with… So there are a lot of advantages... a short statement with a URL so they can see the whole statement... There are a lot of good reasons we should do that. We should make sure IHTSDO agrees. I wrote to...(?) to confirm.. and he forwarded on to someone else to ask if that copyright notice is appropriate. Also – there is probably a need to acknowledge the use of LOINC codes. That is less clear. There is wording in the LOINC license that says you should include it, but their use-case was not this sort of modeling use-case… they were talking about documentation… or re-distributing LOINC – should have copyright notice in the file with the redistribution. So I sent an email to Clem and to Daniel Vreeman asking what we should do in terms of copyright notices. Thomas?

Tom: If the model comes from originally from 13606 or... One assumes that there will be a statement that it will have a copyright notice... Good citizenship and all... A good idea to try and establish the culture… trying to do this… sooner rather than later… We have never been proper about this… Everything is about IP nowadays… As long as it is short statements... legally… avoid a problem of re-(vising?)… 70,000 archetypes just because have a… that has to be changed.

Stan: Yes. OK. So, the more general question – what would be the proper process of… who is the author, what is the creation date… a keyword list… What would we want to have as meta-data... Might have to fall into 2 categories... Might have to be there or not able to maintain... We have the versioning document that you authored, Thomas, but it is available from openEHR… other things we should consider?

Tom: There is an advisor(?) thing... Not sure if we can solve that. The openEHR… spec… pretty generic... being like magpies and… available online in a URL… With putting a few extra things… original... custodian... organizations... the current custodian... license… there is no guarantee that covers everything. People in openEHR… seems to be working... We still have to do our own analysis… mirror-data(?)... The key thing I would say must be remembered… some of it is registry… registry-data… about the artifact, but... about an organization… or organization’s idea or notes... artifact... For example, valid date-range... A bad idea to put inside an archetype... It can change… I would suggest the 2 places that we can most efficiently... is that spec that has the latest ADL model of the meta-data… Can be persuaded… he’s talking about... and get that meta-data list. And have to make sure everyone remembers to separate… and registers… that would be strategy to proceed.

Stan: We need someone to lead this. I will contact William. He said he would… And we have your document, Thomas. We need someone to go through that information and come back with a proposal we can all talk about. Does someone want to volunteer?

Sarah: Stan – this is Sarah – I will put my hand up.

  • ACTION ITEM: Sarah to take the lead in addressing copyright procedures ***

Stan: Great – I will forward you Thomas’ file and Williams’ file...

Sarah: I know the people in Oxford were interested in participating. So I will get a small group together and turn this around...

  • ACTION ITEM: Gerard to provide Sarah with access to ISO documents***

Gerard: I have access to the ISO documents, and I will provide you with them.

Stan: OK.

Tom: It might be worth getting Ian McNicoll... He is deeply involved in all that meta-data… deciding CK and… For example, in the recent version of this model… they came up with... people in US don’t tend to think about… but he would be a good person to contact for that.

Stan: OK.

  • ACTION ITEM: Gerard to send list of copyright information to Sarah or Stan***

Gerard: I will send you my list of things as well.

Stan: OK. Thank you, Sarah. If you have roadblocks, let me know. We can find someone else to help with it.

Sarah: I will do my best and yell if need help.

Stan: I really appreciate it. OK. The next item – Well – is there anything else people want to say about meta-data before next item? [No Response] URL from Thomas Beale:

Pre versus post coordination of method, specimen, and timing in the CIMI preferred lab models? - Methods – we prefer post coordination of method

Stan: So, let me share my screen. OK. So – a little bit of background on this.

Stan: I had a grad student who compared… Regenstrief and Mountain Health and… how consistent are people for using the same LOINC code for the same analysis. How people mismatch between the local codes and the LOINC codes. One of the problems that came up – you have options about whether you pre or post-coordinate… One example on the screen – these are… messages. This is detecting Hepatitis-C virus IgG antibody. In first example, you have… has no method coded.. method part is missing? Then in … there is a place where you can say the method or procedure used… you can put that it was done by EIA. Another part--- the method is pre-coordinated into the code. And so... this is a situation that occurs for lots of things. The newer a test is, the more often the people care about the method that people use. Early in the life cycle of test… the different methods end up with different sensitivity and… So some that are old measurements… serum sodium and potassium – methods have been standardized over a wide range… but antigens and antibodies and other kids of substances are pretty new. So in terms of modeling… the questions I would raise is… we would have different models... one with pre-coordinated and one with… But which would we have for CIMI- preferred. So that is my intro – I would be interested in thoughts and opinions on which of these 2 things should be CIMI-preferred.

Gerard: The rule we have for IHTSDO-SNOMED can be applied for LOINC as well.

Linda: I don't think there is any specific rule regarding that… because both… to SNOMED. Previously we said that we wanted the most post-coordinated possible without pulling apart parts that are normally pre-coordinated. Sort of fuzzy… So based on sentiment I think the... approach is consistent with that… But...

Stan:: Yes- the post-coordinated... up here (on screen) is what we...

Linda: Yes.

Stan: And I think that Gerard was saying the same thing as you, Linda.

Gerard: Yes.

Linda: Perfect.

Tom: CIMI in 5 years time... Have best practice use-type codes... is the desired best practice already known out there and it is CIMI’s responsibility to find that… or should we come up with best practices?

Stan: It is inconsistent. And the inconsistency is generally driven by the frequency of occurrence. So if important to know the difference between EIA and Latex agglutination and 90% of world are doing it by EIA, then for that particular test… to use pre-coordinated rather that post-coordinated. Most of these – there would be a clear approach, but not consistent about pre and post… mostly whether people think it is important… people make a practical choice. So that is one aspect of it. To be clear... One of those places where it is clear that choosing a single preference... you will cause work for someone. If you choose a peer-philosophical approach, you will be asking almost everyone to convert to how they do it… for consistency… even though most of world is doing it a standard way. Another comment… my thoughts about the preferred method – have evolved a little bit. There is value in saying the “CIMI Preferred approach”. But I have come to the realization that people may listen, but we don’t have the authority to say… for things sent to government billing office… we don’t have the authority to dictate to them…

Stan (cont'd): Oops – I am going to get caught up in an update…

Stan (cont'd): So – what I am thinking... I think we want to say a CIMI-preferred method, and we want to nudge people and use that as a way to encourage people to do it in one way, but I think the reality is that people who have the authority which, in the case of... quality in US or patient data for claims... the government or professional body will say their preferred way… So want to capture the CIMI preferred way, but also want to capture the CSF-preferred way, the NHS-preferred way, the… So I rambled on. Additional thoughts?

Gerard: The same way with… will have those problems. But those are… issues… As long as our preferred way can be converted into their preferred way – then no problem. And doing it this way makes it possible that they… to our message.

Tom: I’ve had.. for some time… The idea of profiles, badly overused work – inside archetypes that would point to alternative elements and name… to say which ones were... You can imagine doing it for some… you could potentially imagine doing it for different types of code… Now I... if the idea has legs and I have no technical material… conceptually rather than… The way to go is... So somehow be able to refer to them... So if, in the future, can see that a movement towards... and get rid of the other one… So in future… potentially filter them and…

Stan: Yes.

Tom: Rather than try to be definitive…

Stan: This relates to 2 other things. One, if I understood questions from Michael van der Zel in Amsterdam, this top approach v.s. the second approach – these would classically be… members of the same isosemantic family… would want to know that… model that post coordinated is similar to ... And Michael said – well, it is one thing to know if part of the same family… And I would like to have... If I had the pre-coordinated way, I could make post-coordinated for... So if I had the transformation… based on the model mappings, could transform anything that conforms to that style. I think, Gerard, you sere saying the same thing. Recognize that these 2 are the same thing… so if you store in your DataBase one way, you know how to get from you logical... to the other..

Gerard: The Semantic HealthNet project using … Is the kind of language that…

Stan: So – that is probably someplace where we need to… I am not as familiar with that as I should be… So, rather than create a new way...

Linda: That is also consistent with the expression templates that we were discussing in Amsterdam. And that is what the templates were for. Expression templates could be used for exactly that.

Stan: OK – so for this, I heard, for CIMI-preferred model we would prefer the post-coordinated one, which is similar to before… is that OK with folks?

Linda: Yes.

Gererd: Yes.

Stan: OK.

Stan: I don’t think I put this on the agenda, but it is applicable as well. Another kind of heterogeneity... A problem you have in LOINC….

Linda: Excuse me Stan. Do you want to show your screen?

Stan: Oh yes... OK – so cocaine in blood

Stan: So, in lab – you do the test. Sometimes one is reported. And sometimes both are reported. [It is] an absolute concentration that comes out of machine. With drugs – it is typically – law enforcement or other… say whether this is a positive test – would be considered illegal or improper. They interpret the number and say whether it is positive or negative. Sometimes is in a black box… and they don’t report the number – just whether it is positive or negative. In LOINC you have the 724-045(?) code – is blood quantitative (qualitative?)… and the 026… is cocaine abstract concentration – instead of quantitative, it is ordinal. And the interpretation comes out as the value. So if you look at this middle line – in HL7 – you have a place for the numeric(?) question and a place for… This is proper. But what happens is... LOINC – we did not think about it enough, maybe. So others – make this ordinal code and report positive. So positive is… interpretation and in other – considered part of test. And the other thing – people use the quantitative code and don’t send numeric value… just the interpretation… put where it goes… no numeric value. So – following what we’ve done before – we would say, this middle one is the preferred… and we could make models that would say… like this… and the proposal would be… You never… we would discourage people from using ordinal code… we would use quantitative… so if no number sent, we would still send positive as interpretation. Make sense? So what I am proposing… This style... [on screen] would be the preferred…

Gerard: I agree, but would also add the exclusion/inclusion criteria. So I add why it was considered positive.

Thomas: The value of that probably true of the whole Netherlands or…

Gerard: I don’t know...

Stan: The reason is – they change over time. Alcohol level. The level of alcohol that is improper is the same for a long time in a given state… but varies from state to state... And the other thing – it changes over time.

Stan: So we would probably include in the... data.

Tom: In openEHR, the quantity has... ranges and the... range... and people say that should be… and lots of tests… whatever is true for that patient at that time is whatever is true… So any time I have said… gets used all the time for that reason.

Tom: I think the interpretation should not get mixed up with the observation.

Stan: Yes- I think we are trying to deprecate the use of this style, which is not good. Other thoughts or comments?

Linda: I agree.

  • ACTION ITEM: Stan to annotate the discussion above so we can remember what was said***

Stan: OK. After call I will annotate that so I can remember what we said. OK. It is the same issues and probably has the same answer as other things….

Stan: For instance, heavy metals, there are LOINC codes… 967-5 has a specimen of xxx – it is in the LOINC file. A key to say that this is a code to be used for post-coordination. The Cadmium will be found in another part of the model/message. And you have things that say – here is Cadmium in hair or blood or… And then you have a confounding where they say – tissue – a catchall for you finding Cadmium in liver or… tissue.

Stan: So, the very specific tissue would also be said in some other part of the message. So – this is following the same pattern as method. You end up with LOINC codes like this because people measure CADNIUM in hair a lot… but they recognize that you can measure it in other tissue… but not happen enough… cadmium in liver or… so they use more generic… has a lot of similarity to the method. This is one of those things – the proper modeling style would prompt us to say – let’s post-coordinate that, but you will find that in people’s databases, probably more than 90% of data will be in one of these specific forms… You are asking people to convert in 90% of data from EHR to… I think that is the best to do, but I want to point out the implications… the grumbling that might go on… the reality in modeling that in a working system. Comments?

Gerard: This is exactly the same example as you have before. Different now because you have location as one of variables… The problem will be for implementers. Not for clinicians. Should not ever be exposed to numbers... should see on the screen what they want to see… So those numbers are there for implementers... for people who want to do research.

Stan: The same happens with cell counts – still specimens. Body fluids where you can measure erythrocyte counts, and… CSF... They have pre-coordinated the concept... and the more general… You could have wound drainage... could measure it... and that is what the xxx is used for, so you can say. But again, the most general approach is to say – let’s post-coordinate that. But the argument will be – it is not the end-user clinicians who will worry. It is the software engineers who do the surfaces... and return the... in the general form... who have to write code...

Tom: You can imagine... There will be... will potentially be of some value that you could set the binding of the code to the… tissue or body fluid or whatever… You could imagine we could do it if could express conditional bindings… Coded x is bound to this one… but have a rule… path y... implies… Binding here is code 1-2-3… Nice in theory, but then you have to nail down the values… Maybe gastric fluid... Would have to guarantee there is only one or small value set that… would allow you to write that conditional rule. Now – you can imagine going down this path if the ability to do this would mean that CIMI archetypes could pop out exactly the right code… But I don’t know but I think the technical way to do it is conditional bindings.

Linda: With respect to that, I think that will become important to do conditional bindings that… to have a template for the lab test itself to define how it may be pre-coordinated in terms of the method... and in terms of the SNOMED expressions… And the rule – if the specimen is defined in there, they must be… in the pre-coordinated code. But the other thing – because we are looking at these LOINC codes one at a time and defining different rules... what that means overall… are we... redundantly repeating the method... when reported… so could use… but then redundantly use… in the method. Another alternative... Stan : I personally don’t like that as well. But I understand the practical part of why that would be useful.

Linda: Yes, but having the template of the… if you want to say something more specific in the specimen field… if you want to say something that is…

Stan: We probably have to make some actual examples of these things… using ADL… to make sure we are communicating… Some of this you can do in other parts of the model where you say the tissue type... rather than having some conditional binding expression... you could do with the.. To talk about that, we probably have to have some actual models… and describe what we want to do with it.

Linda: Yes – it occurred to me you are assuming there is a separate model for each of these... Yes.

Stan: Yes. What I suspect would actually happen is... literally – every system I know about has erythrocytes and blood – the normal red cell count is represented by… So, any given system might have 3-4... and all of the rest they do as pre(?)coordination because... We would have the general post-coordinated model, and we would have pre-coordinated model… we would know they were part of the same iso-semantic family… And some would say, well for our cancer… we are using… but we know how to get from… and we made it computationally possible for people to use either one and they can be consistent… In some sense this is an argument… authoritative... body as opposed to the level of modeling geeks. So – in terms of a CIMI-preferred way – in defending the choice – I am still in favor of post-coordinated choice.

Gerard: You can add a second... in place somewhere... is the pre-coordinated one... the specific CIMI-model.

Stan: We – being Joy and others – have thought a lot about that. We can come back... to talk about how we see the… between pre and post-coordination... How we can represent the… and have an expression… show exactly how you can get from the pre-coordinated to the post-coordinated. I think there are ways, to do it, but it is new – different from how people are doing it in their systems. Stan (cont'd): OK. I think that covers the items that I was talking about in this pre- vs post-coordination. I will take some more notes about what we said were after the call. But my general statement is that – there is a consensus – the CIMI preferred method is the post-coordinated approach. But we would have specific models for the pre-coordinated style. We will talk about that. Any further discussion on that? [No response] What is our approach for indicating which models are the CIMI preferred models?

Stan: OK . So that leads into – what approach... is the CIMI preferred… And... an associative table that is outside of the models themselves. And it would basically say – this model is preferred by this organization where CIMI is one of these organizations... and it would say the date and time that we stated this preference. And could say – this is Australia NEHTA-preferred strategy and the 130606-preferred strategy… That is what I am thinking… would do that as an associative table… it is an independent act separate from the model itself. 2 things – want to know preferred by whom, and then when expressed… and maybe something else - how more preferred... so history… thoughts about that? Have you done anything like that in openEHR or 31606?

Gerard: There is a document describing it.

Linda: I think… would be repository metadata... in the repository and displayed in... interface.

Tom: In openEHR – the more people try to systemize it... and people who work with clinical data say... it is hard… Competing models – how do we mark…. The question is—preferred to whom... and to what purpose… and the question becomes a lot grayer… because some preferred in one specialty and other specialty… Probably more clinically driven rather than IT-driven. Probably need mode /informatics… the best-practice sort-of idea. Or are we back to an ONC-preferred approach… a Kaiser-preferred approach… Get the requirements sorted out better if seriously going to try and do it...

Stan: Yes. That was my impression. You have sort of done this by saying...

Tom: Not dictators... What happens is... you get… Always integration and transformation… You could pick anything and say “We prefer that”... the openEHR archetype… one is as good as another… because all of the exterior are going to have to transform to another… having one target is better than… If you… data… you showed them 2 different versions that adverse reaction… they’d say – just give me one. I know that won’t always work… but that is the... in openEHR community… reasonably diverse openEHR environment. I don’t hear… got to solve this preferred…

Stan: IT has come up a lot in IMH. Especially when get outside of lab data. There are much-preferred models... people putting the data in. SA a general rule – those putting data in – prefer pre-coordinate... And those doing analytics – prefer post-coordinated.

Tom: Yes – analytics want… they are going to cross-cut… that is data-warehouse… so you could ask – what is the data-warehouse... going to look like… is all of that... I would want to be clear before spending an effort to make it work – to make sure we are solving a problem that is really there. I don’t think the war between primary and secondary data… is ever going to be won.

Stan: I don’t think I understand. I am trying to solve the exact question I asked about lab data – pre and post-coordinated. My lab student went out and looked at what people using in their Databases. Sometimes pre and sometimes post. We want to state the single preferred way. That is the use case. I don’t think it is ambiguous.

Tom: I can say there are alternatives all over the place. Pre and post coordinate. The only question I have is – are we in the business of saying what is preferred, or are we in the business of saying... So that is where I said the tagging(?) system… preferred for who and where...

Stan: That part is not hard – to say an organization name and some purpose and coding which model they prefer.

Tom: There are quite a lot of alternative inside a given model… probably sit inside the same archetype… for a given… or maybe you are thinking of having separate archetypes… does not seem to be always between 2 or 3 different models. If someone wants a horse and someone a zebra, similar animal underneath...

Stan: I would have to understand the examples… But the red shape – pre-coordinated, and hexagon – the post-coordinated...

Stan: So, do we want to say… the alternative is where you translate codes, but you … to the application either pre or post coordinate data… but I think this… is not possible… You could pre or post coordinate specimen or category of collection periods… whether is 24 hours or 6, our collection of urine... But I am not understanding- this is what I am trying to get to... preferred to what… a model of... so I can make this truly interoperable system that does not know about…

Tom: That implies having one model, unless you are putting the... mapping stuff into the model as well… the transformation layer as well,

Stan: Or as I understand historically – openEHR said – this is the only model we are concerned about – the model we are going to use for exchange. That is your own business… and transformation… That is why we have one model... to create interoperability. But CIMI – is trying to make it easier for people to make the machinery to do the transformation… so I can create... and share.. I can create automated machinery to do the... not only the data I want to produce as the common one, but also formally have… and I can give them help in instance translators.

Tom: Are those shapes at bottom – are they also archetypes?

Stan: Yes – non-preferred archetypes.

Tom: So – adapter archetypes for the CDA or whatever. If take adverse-reaction type... done with FHIR group… there is quite a lot that are coded… for any data... pre or post-coordinate… originally cancer diagnosis… then you go look at actual CDA…

Tom: The pre and post for that CDA could be just for that hospital.

Stan: I am not going out looking for them and no commitment that we would create these comprehensively. Only create these if found people who found value in… What we care about foremost is creating the models that we want to be the bases for interoperability. And creating the… and you would only make these models on demand and to help people implement…

Tom: So - on bottom row…

Tom: So … would be templates, not just archetype… going to be a closer artifact... In that translator in node... super-smart… would know the… I don’t know. We try to solve that by creating templates... any template that… we try to make that as much as possible. You might say – if only preferred form archetypes available to make… [I missed some]

Stan: If we don’t have value, we won’t make these other guys... the hexagon… but where it has value – where a whole part of the world is doing it a certain way… where you can get to one or two of the… common or preferred way that allows interoperability... will get people to use those models… that is my idea... theory...

Tom: Linda’s expression(?)… slides in Amsterdam… the blue archetypes… someone to make… Well we can… solve those… and between conditional rules or Linda's thing can solve… have not done enough… archetypes… I am wondering if is an on-demand thing… when someone says… can anyone in microbiology say…

Stan: ...

[Meeting Ended Abruptly Due to Technical Problems]