2017-10-03 IA Small Group Call

Owned by Stephanie Klepacki
Last updated: Nov 20, 2017

Date

Attendees

Scribe

Goals

  • Continue work on ANF modeling of CDS KNART terminology artifacts and view a demo of a survey instrument that can be used by participants to indicate how they would model a set of example clinical statements.

Discussion items

TimeItemWhoNotes
10minANF modelingGroup
  • Keith:  Summarized John's feedback on the clinical statement modeling instruction ("ANF modeling") document.  We want everyone's feedback to be addressed.
  • Claude:  Some of what they took from CIMI evolved as they worked through the ANF.  It might be worthwhile to step back a bit and understand the motivation as to why we have that separation and the definitions of topic and context (circumstances).  Keith:  Let's discuss this after looking at the work that Stephanie and Catherine have done regarding the questionnaire/survey.
 25minDemo of ANF Modeling Questionnaire Stephanie & Catherine 
  • Stephanie:  The questionnaire is included in the University of Utah's REDCap survey tool.  She demonstrated the questionnaire that walks a participant through a series of sample clinical statements and gives them choices for how they would model each clinical statement, such as what portion of this statement do you consider to be the topic and which part the circumstances.  Participants can choose from a list of pre-defined responses or enter their own response.  There are multiple ways that the responses can be displayed, such as with a series of drop-down lists or through matrices.  Keith:  Suggested another method, which is to have a participant copy and paste into a text box the portion of the clinical statement that they think best represents the topic and circumstances.
  • Catherine:  The question to her is how do people put boundaries around the topic and circumstances.
  • Keith: For training purposes, he thinks that having the drop-downs and matrices are good.  For the first step, we're trying to figure the rough boundaries, and having a clinical statement with X number of questions will be helpful.  We are testing the reproducibility of the model.  Recommends we proceed with the free text response.  His question is what kind of data can we get out of it?  For example, can the survey tool recognize when two people entered the exact same response?  Catherine:  Yes, and the data is exportable, such as into Excel.
  • Claude:  He really likes the questionnaire form. But, we have to be very careful about the questions we ask.  The goal of knowing if the people read the instructions and correctly applied them is very difficult from whether the prompt we are giving them is clear or not.  For example, the clinical statement of acetaminophen is very vague since it doesn't state whether it's to administer it or to document that it has been administered.  Keith:  We need to agree that one of our goals is that none of our clinical statements are ambiguous.
  • Keith:  His strawman proposal is to use the copy/paste option vs. the menu/matrix and we will have less bias as well.
25minReview of Claude's Proposed Claude
  • Claude:  Displayed a document showing his attempt at addressing some of the questions that still remain.
    • There are three components that make up the patient record, each with their own proposed definitions:
      • Clinical statement class
      • Topic class
      • Circumstances class
  • Keith:  The topic object can be a single statement or have multiple fields within it.  Same for circumstances.  If you want to draw them up for what you think they are but not for domain-specific concerns leaking into them, that is fine.
  • Keith:  Asked Claude to send his document to the group so we can all provide our feedback on it.  Having it in the Glip forum is fine for collecting the feedback.

Action items