Negation Requirements Project Minutes 8 March 2017

Back to Negation Minutes

Minutes

Meeting Information

HL7 PC-CIMI-POC Meeting Minutes Location: PC call line		Date: 2017-03-08 Time: 11:00-12:00 ET
Facilitator	Jay Lyle	Note taker(s)	Jay Lyle
Attendee	Name	Affiliation
y	Jay Lyle	JP Systems
y	Monique von Berkum
y	John Kilbourne
y	Elaine Taylor
	Jaime Jouza
	Zahid Butt
	Andrew Haslam
	Floyd Eisenberg
	Susan Barber
	Larry McKnight
	Galen Mulrooney
	Kurt Allen
	Jim Case
	Juliet Rubini
	Hank Mayers

Agenda

Agenda Topics

1. 2/22 meeting: no quorum
2. Ballot review
   1. Use cases
   2. Classes, coverage of spreadsheet
   3. Need for activity diagrams?
   4. How to make dependency on desiderata explicit (e.g., concrete, explicit)
   5. Cases: nkds, intents
   6. draft of policy?

Minutes

Minutes/Conclusions Reached:

1. Most cases can be handled without logical negation operator. Can we identify cases where this operator is necessary?
   1. Example: search for "no cancer" (to screen for a study, e.g.). If "no breast cancer" is a child of "no cancer," then patients with other cancers who have "no breast cancer" recorded may be inferred.
   2. This seems like a problem with using DL. Three things may be necessary for this:
      1. The element must be modeled as a negation - if not fully defined, at least with explicit negation.
      2. The analyst (or agent) must understand how to classify negated concepts (e.g., which DLs can cope, how to identify where inversion is appropriate, etc.)
      3. There cannot be other implied dimensions of negation that are relevant but not handled.
         1. Implied negation: concepts that look affirmative but imply negation; e.g., "left hemiplegia present" implies "right hemiplegia absent"
            1. Places where it looks implied may not be. E.g., 'oriented' and 'disoriented' are not complete -- neither includes comatose. I.e., 'not oriented' does not imply 'disoriented'
         2. This is a problem endemic to language and reality. It may mean that any uses of negation must be task-specific; one cannot represent information and expect that there is no context in which negation may not become relevant.
   3. Information modeling cannot solve this problem, except to specify condition 1 - that you can't do it without explicit negation.
2. "Spleen present" is asked.
   1. But it is in the context of safety - checking for contraindications.
   2. Is there a criterion we can use for determining when explicit negation modeling is necessary or useful?
      1. "Persistence" of condition doesn't work: absence of spleen persists.
      2. The "Persistence" of interest is the information's usefulness
      3. Criterion: is/would DL be useful?
         1. E.g., "NKDA" works fine: what additional cases do we enable by refining it to "DA - Absent"?
3. Desideratum 2
   1. Unclear. Clarify that this is about the problem domain, not solutions.
4. Desideratum 4
   1. clarity, ok. Recommending 'idempotent' may be 'design'; avoid.
5. Key issues
   1. Children; i.e., inferring subclasses when inference is inverted
   2. Equivalence; i.e., when two expressions mean the same thing. [Not so much: this is hard, but not specific to negation]

Meeting Outcomes

Actions review Use Cases for completeness, classification for accuracy (all)

Next Meeting/Preliminary Agenda Items Continue review of requirements

© 2012 Health Level Seven® International. All rights reserved.