SNOMED CT Editorial Advisory Group

Page tree

Date: 2022-06-15

Time:

1030 - 1200 PDT

1730 - 1900 UTC

1830 - 2000 BST


Zoom Meeting Details

Topic: SNOMED Editorial Advisory Group Conference Call
Time: Jun 15, 2022 10:30 AM Pacific Time (US and Canada)

Join from PC, Mac, Linux, iOS or Android: 
https://snomed.zoom.us/j/89050682834?pwd=R3dISFhIR0R3NnN2LzNMeGFsT2I5QT09
    Password: 251865


Attendees

Chair:

AG Members

Invitees

Observers

Monica Harry 

Ed Cheetham 

Toni Morrison 

Victor Medina 

Yongsheng Gao 

Marie-Alexandra Lambot 


Apologies:



Meeting Files:



Meeting minutes:

The call recording is located here.


Objectives

  • Obtain consensus on agenda items

Discussion items

ItemDescriptionOwnerNotesAction
1Call to order and role call

Jim Case

This meeting is being recorded to ensure that important discussion points are not missed in the minutes.  The recording will be available to the SNOMED International community.  Joining the meeting by accepting the Zoom prompt declares that you have no objection to your comments being recorded


  • Recording of meeting approved by participants.
2

Conflicts of interest and agenda review

Jim Case

None noted.


4X (person) vs. X of subject (person)Jim Case

A question from a member country on when to use "X (person)" vs. "X of subject (person)" has exposed issues with determination of equivalence in information models that either split the relationship from the condition vs. using a precoordinated Situation concept to represent the SUBJECT RELATIONSHIP CONTEXT.

3/15/2022 - Update

A report on the use of person concepts as values for the SUBJECT RELATIONSHIP CONTEXT attribute is located at:

https://docs.google.com/spreadsheets/d/1LTPSInpRC_HMPniQANM8NL86WCieSAttoPYDS_yxjno/edit#gid=1

  • Are familial relationships Roles or Persons?
  • Given "X of subject" is primarily used as the value for this attribute, should these concepts be moved out from the Person hierarchy into their own "value set"? 
  • Is Person the correct hierarchy for these to be placed?  
  • How do we handle "Fetus of subject" given the sensitivity of some members of having a (person) semantic tag? This is needed when procedures or conditions are performed/recorded in a fetal record as opposed to a maternal record.

Discussion:

Previous discussion has been moved to Confluence discussion page at: X (Person) and X of subject (person)

    • Changing roles in family structure requires a distinction between biological relationships and non-biological familial relationships.
    • Existing "Person" hierarchy is in need of review and updating.
    • Suggestion has been made by multiple people to handle this issue through reorganization of the values for "subject relationship context" (separate hierarchy?)
      • "Role" hierarchy exists.  Values imply a relationship to a physical entity (e.g. 766941000 |Therapeutic role (role)| implies a substance or product)
    • Also has been suggested to create new attributes that allow for more specificity
    • Need to consider that subtypes of "Person" are also used as values for the FINDING INFORMER attribute. 
    • The reference paper provides two mechanisms to represent "fetus"
      • Creation of a robust fetal anatomy hierarchy
      • Use of a SUBJECT RELATIONSHIP CONTEXT value (provided subject relationship context values were moved out of the person hierarchy)

A question was raised as to whether we need the "X of subject (person)" at all since the relation itself defines the relationship.  These concepts originated around 2009, but only a few were present at that time.  It may have originated due to the need to represent the "Fetus of subject".  Suggested that we separate out the difficulties of fetal disorders and findings from the rest of the subject relationships.  

This also raises the question about the degree of pre-coordination there should be in the International release?  

The distinction between X (person) and X of subject (person) is lost when looked at in the context of a medical record.  The element within the record in which a term is used provides the context in most cases.  The most compelling reason to keep the X of subject is related to fetal disorders and procedures, which is something that has been agreed is a separate issue.  (i.e. representing "father of fetus" as opposed to "father of the subject of record")   

The FINDING INFORMER attribute is problematic and needs a comprehensive review for its usefulness.

There are also issues with the relative non-specificity of FSNs in the Situation with explicit hierarchy with regards to the subject relationship.

It is reasonable to assume that the underlying reason for the addition of the X of subject concepts is no longer relevant. 

Decision:

Test the impact of replacing "X of subject" values from the situation concepts.  Inactivate the "X of subject (person)" if no adverse effects.

Review the existing person hierarchy for quality.  Future considerations will be around roles, gender identity and sexual orientation related to person. 

  • Jim Case to implement changes to the Situation concepts.
8

Measurement Findings:

Proposed changes to FSNs

Paul Amos
  • Proposed changes to FSNs: Please read the attached Briefing before the meeting and if possible return your comments for discussion at the meeting.
  • This meeting will address the following issues:
    • Concepts representing "borderline" levels
    • Concepts representing "therapeutic medication levels"
    • Concepts representing "Abnormal" or "outside reference range" levels

Decisions to date:

  1. Following the discovery of a decision made in 2010 that within the context of measurement findings "increased" = "above reference range" and "decreased" = "below reference range" the EAG members agreed to support a change in FSN rather than inactivation.
  2. In the rare event that a user has interpreted and used one of these concepts to represent a relative increase or decrease in measured value; advise use of 442387004 |Increased relative to previous (qualifier value)| and 442474009 |Decreased relative to previous (qualifier value)| with clear FSN e.g. Increased blood glucose level relative to previous level (finding) for implementation at the local level.
  3. There was agreement that the notion of "normal" is context dependent and similarly "within reference range' should be interpreted within the context of the whole clinical record. Therefore, it was agreed that concepts representing "within reference range" will have their FSN/PT updated and a synonym description of "normal" should be retained or added as appropriate.

Discussion:

It is agreed that the replacement of FSNs is less destructive than inactivation and replacement of concepts.  This is only relevant where current content is modeled with above and below reference range.  Concepts that are inherently ambiguous are still used in clinical recording.  Adding forced meaning to these might make them less useful.  Suggested to identify these concepts using a refset that they are problematic from an interpretation point of view. 

This project was an effort to reduce the ambiguity of existing concepts to ensure clarity of meaning.  Retention of ambiguous content may provide clinical recording simplicity but may be in conflict with precision.  There is a conflict between the utterances used in clinical recording and the desire to provide structured analytical data.  The issue may be exposed when looking at new technologies that go from speech to text to coding...

Suggested that the ambiguity can be resolved by using other "imprecise" terms such as "increased" for "above reference range" and "increasing" for "increased relative to prior measurement". Alternative description types may assist in representing broader or ambiguous descriptions.  Another suggestion is to allow for these imprecise concepts, but mark them as such within the terminology.


Decision:

Topic not completed.  Carried over to next meeting.


10AOBEAG



11Next meetingEAG

TBD



















2 Comments

  1. Ed Cheetham

    Dear Jim, Paul, EAG


    Regarding item 8:

    Many thanks for posting the evolving briefing note and minutes for the last meeting. Since the 'SNOMED CT Laboratory Findings' document continues to contain the 'do you accept the proposal...' consultation boxes I shall assume that there is still room for discussion.

    If so, a question: have the developers considered approaches beyond deciding between renaming/modelling and inactivation (decision 1 in item 8 above)?

    In particular, whilst an SI goal continues to eject traces of vagueness and ambiguity from its content, the reality of language continues to indicate that some forms of ambiguity are unavoidable, even desirable. Indeeed this 2011 paper concludes that there is "...strong evidence for the view that ambiguity exists for reasons of communicative efficiency...". If so, efforts to remove all traces of ambiguity from SNOMED CT may reduce its ability to contribute to technical solutions that interact with language more directly (e.g. by participating in NLP pipelines).

    Instead it may be preferable to reframe the 'problematic' examples discussed in the briefing note as merely patterns and phrases that clinicians will continue find useful and efficient but which carry interpretation risks. As decision 3 in item 8 above (and the paper) note, the ultimate interpretation of 'normal' or 'raised/high' or 'decreased/low' depends on surrounding or contextual information. A valid approach for SNOMED CT may therefore be - rather than to reject such content or impose specific meaning on existing phrases - to retain this content, but highlight it as requiring extra care for purposes of interpretation.  I would guess that for a lot of (certainly English language) speech→text→code workflows (such as dictation), phrases such as 'low haemoglobin' will be uttered more frequently than 'haemoglobin below the reference range' (it's simply 'easier' and more 'efficient' to say). By retaining a code with a term that can match 'low haemoglobin' (and recognising its incomplete semantics) SNOMED CT can still play an active part in making such a phrase machine-readable (in the spirit of 'something is better than nothing') - but also be able to provide an machine-readable caveat that actionable interpretation requires particular care and additional information.

    Assuming a domain specified by << 118245000, the statement that '...within the context of measurement findings "increased" = "above reference range" and "decreased" = "below reference range"...' does not hold. I agree it is overwhelmingly the case, but there really are exceptions. In particular, I would be very surprised if a patient with long-standing asthma, speaking about their peak flow being 'increased' (37494006) or 'decreased' (7206008) is doing so in terms of a reference range. Their 'increased peak flow' may be an improvement on their previous week, but may still be well below their reference value. Imposing a 'relative to reference value' interpretation does not seem appropriate.

    The domain of interest may, in fact, be better identified by considering a combination of the language used or the modelling applied. Many of the troublesome 'words' (mostly adjectives) can be found in the 250 or so concepts beneath 260245000 |Finding value (qualifier value)| (of which only ~1/3 have been used in modelling). Is it not worth considering a (highly leveraged) attempt to distinguish those 'words' (including a non-English analysis) which pose the greater interpretation risks and then identify a corresponding set of findings where they are found (or used in modelling) where particular care should be taken. Identifying such a set would make no statement about the rest of the content but may give users helpful warning about its stated members. Many of the set members would be found beneath 118245000, but I'd wager that many would helpfully be found elsewhere too.

    The updated paper invokes "patient safety reasons" for some of the proposed changes - I would prefer it if these reasons were properly spelled out, the hazards identified and risk control options explored. Simply attempting to remove all traces of 'ambiguous' content, or fixing the values as 'above or below reference ranges' may not turn out to be the optimum strategy once such an analysis is done. Certainly any inactivation-based solutions (there are some in the paper but I'm not sure if they have now been replaced by the minutes/this agenda) are of little value if SNOMED CT supports post-coordinated representations of exactly the same 'ambiguous' ideas: if 'low' is ambiguous in a finding term, then 62482003 |Low (qualifier value)| is ambiguous in a compositional expression, and users should also be prevented from making it the target of any 'has_interpretation' relationships.

    Kind regards

    Ed

  2. James R. Campbell

    Paul, I concluded when you introduced the discussion and called on me that I had been reviewing and responding to the wrong briefing document.  I was confused by the number of attachments and assumed that the 'Briefing not for April' was most current.  Can you please verify which document was your reference in the agenda?

    Jim Campbell