2017-10-17 IA Small Group Call

Date

Attendees

Not Present

Scribe

Goals

  • Plan for this week to resume work on draft guidelines for ANF modeling of CDS KNART clinical statements and hopefully finalize them while Keith is on travel.

Discussion Items

TimeItemWhoNotes
10minSchedule Additional Calls this WeekGroup
  • Additional calls were scheduled this week, as shown below, so we can work on the document while Keith is on travel, particularly on how to differentiate the topic from circumstances.

    • Wednesday, Oct 18, 1-3 MT
    • Thursday, Oct 19, 1-2 MT
    • Friday, Oct 20, 9-11 MT (extended standing call by 1 hour)
 50minClinical Statements and Topic vs. Circumstances Group 
  • Claude:  In CIMI, a clinical statement (CS) is in the patient record that is generally made by the provider of record but could be from a relative of the patient.  Information is included such as who made the statement, what was recorded, and who was the signer and co-signer.  All the attributes provide metadata on the statement that was made. At some point, you have the question of how to analyze the data contained within the statement.  In CIMI, a CS has a topic and context:
    • Topic:  What the statement is about, generally a procedure (clinical action) or a finding.  The topic can be specialized or not.  
    • Context:  Strongly aligned with situation context in SNOMED. For example, what procedure is it and the context tells if you are planning it, proposing it, or have performed it already.  For findings, you are asserting the presence, absence, or suspected presence.
  • Claude:  In ANF, it has moved away from CIMI because it's not using situation with explicit context and (missed part of what he said but it was regarding results and finding).  We have broken out the what, who, where, when, etc., but we never really fully specified how the split would occur.  We are now fairly different from what CIMI had in mind for the split between topic and context. We need to be fairly specific by what is meant by topic and circumstances and does the dichotomy still make sense.  If yes, what do they mean so we can start thinking about the rules about what goes where.  John:  Agrees that the split between topic and circumstances is our sticking point.  The CIMI split is pretty straightforward and it's basically the SNOMED context.  What we have been discussing is arbitrary, such as splitting the statement "Head CT with Contrast."  It seems like for every procedure, we need to lay out the split.  If we had a URU (i.e., understandable, reproducible, useful) explanation of the split, it would be extremely helpful.  Another example is for a medication order where the dose as part of the topic but other information is included as part of the circumstances.  None of us on this call are proposing the split and really understand it and can explain the distinction. 
  • Claude:  Maybe we can step back.
  • Joey:  Would like to discuss Keith's reluctance to see the difference between questions and answers.  If you go up to any physician and ask what is the systolic BP, they will tell you it is X mm of mercury.  Another example is the hematocrit value.  If you say what is the dot blot hemorrhage, that is something different.  Keith is saying they are the same things.  But, the computer  has to process them differently.  One is an answer to a question and another is a statement of presence or absence.  How do you make a consistent, single model and put them in the same slot without telling the computer the differences?  Patrick:  You could put them in the same slot but you have to tell the computer the differences and if you have to do that, you have to lay out all of those differences.  Joey:  Yes, or put them in two different models.  John:  He and Keith agree on dot blot hemorrhage, presence, and then quantifying it (e.g., patient has 4 of them), and there is more you can say about it as well.  Joey:  He could say the same thing about the human body but he doesn't find it useful.  We need to break them down into individual clinical statements, which is everything you say about dot blot hemorrhage.  Claude:  Encourages John and Joey to come up with cases either for or against for Q&A vs. assertions.  Joey:  He understands what he is saying about the individual things you can say about DBH (missed the rest of what he said).  Claude:  Is there a different way you express a disease process vs. the manifestations of it?  Is there a difference between the way you model either the single thing or two things in terms of terminology-based EL actions?  They are two different media for expressing these statements.
  • The group revised slightly the draft definition of a clinical statement that Stephanie had provided.
  • The group discussed events.  Claude:  Hurricane Irma is an example of an event that has a potential impact on the subject of record but it's not a procedure or a clinical finding.  It's an event that could impact the subject of record.
  • Claude:  We have this artificial separation between topic and circumstances.  The CS is talking about something, such as the patient's systolic BP or an accident the patient suffered.  That is really the intent of the topic.  For ANF, the idea is because you can leverage the terminology more fully than what is being done in CIMI, and Keith mentioned it can include things like quantities and dates, not just concepts, the idea is that topic is the concept I'm going to be talking about in this CS and it could be a post-coordinated expression.  The topic is what my statement is about.  Recommends we ignore circumstances altogether and focus on the notion of what is this CS all about and determine the criteria for the information to be included in the topic.
  • Stephanie:  Described differences between how we handled the states of an action, which is by which clinical statement type it is - either Action Request or Action Performed.  This is different from CIMI context that includes proposed, requested, performed, and not performed.  Comparing to it, we have two that are missing - proposed and not performed. Claude:  Agreed.  You may also have performed but aborted, such as for Head CT with Contrast.  That is why we have the separation of topic and context.  The topic is not just in CIMI as purely the what but it also includes some of what we call the how.  In CIMI as you define the archetypes, the topic clearly should be the what and could be a post-coordinated expression.  Theoretically, it could be any concept in SNOMED.  For ANF, his understanding was that the restrictions in CIMI don't necessarily apply.
  • Claude:  Keith wants to minimize the number of classes for condition and observation measurement (not sure I captured this correctly) and minimize the number of distinctions between proposed and ordered and use one instead - requested - to produce the broadest range of CS that can exist.  The problem is if we generalize too much with too few classes, somewhere you will still need to express the differences.  We have to be able to represent everything with 4 classes. We could go the CIMI route but Keith doesn't want that.  There is a conservation of complexity.  The generic representations need to be fine-tuned to have the expressivity required.  At the very least, the CS has to be about something where you capture what it's about.  If you look at FHIR, they have a field for codes, such as for procedure or a diagnosis like Diabetes Mellitus. Whatever it is, we should call it the topic, and it's really the what.  Regarding the types of action statements and the states, let's defer as it's a complex issue.
  • Sarita:  Regarding the issue of differentiating between topic and circumstances, she understands John's concerns about not being able to make the arbitrary distinction and we need to be able to justify it.  These are questions that have been going on for years but in order for us to complete this document, we need to put it in writing and get it done.  Claude:  Initially, we said the topic is the what and it's the type of procedure and what it targets and everything else about it goes somewhere else.  John:  The challenge is not to justify it but be able to reproducibly do it in practice.  If some system has to decide what is part of the topic vs. circumstances, it is impractical and impossible.  The order as written by the physician is, for example, Head CT with Contrast.  Now, some system has to know that "Head CT" is the topic and "with Contrast" is some other thing.  We can't enumerate every single procedure.  Why do we need to make this distinction?  Claude:  Displayed a FHIR example that for a procedure request, it has a code value (i.e., the topic) and what goes everywhere else, such as body site, performer type, requester, priority, etc.  We are trying to come up with very clear guidelines on what goes in code.  John: Expressed his doubt in being able to do this but he doesn't want to be the one in the way.  Claude: We may be able to come up with something that meets the needs of the scope of the KNARTs.