Welcome and agenda

All

Postcoordination Implementation Demos

@Anne Randorff Højen 

@Kai Kewley 

• Continuing the frontend demo to show other use cases

• Discussion

Review Feedback for: Postcoordination Guide (Phase 1)

@Anne Randorff Højen 

• Discuss Guide Feedback

• Last chance this week for new feedback on the guide. We will publish the guide at the conference in two weeks.

Agree solution for "Procedure with explicit context" transformation

@Kai Kewley 

Agree when MRCM constraints should be applied. Jeremy is not completely comfortable with using "Procedure with explicit context" when adding context to a procedure.

Topics for Business Meeting

@Kai Kewley 

Please suggest additional topics for the April business meeting.

• We will be publishing the Postcoordination Guide with "Trial Use" status. We are likely to have a wider audience than the normal group so we could spend some time talking through that.

• We have been asked to consider how LOINC codes (and other alternate identifiers) should be used in ECL as part of the Regenstrief collaboration.

• ECL 2.2, allow working between Full SNOMED and GPS (highest and lowest match) syntax sugar to make it easy.
  
  

The items below are currently on hold

URIs for language instances

ECL v2.2 Proposal

Find the leaves of a set of concepts - example use case (find the proximal set in the international core / or in the IPS) - example:

• Example use cases

  • Proximal ancestors in a specific module: (  > |concept|  {{ C moduleId = 1234 }} ) MINUS ( > (> |concept| {{ moduleId = 1234 }}) )

    • X =   "> |concept| {{ C moduleId = 1234 }}"

  • Leaf nodes:   < |concept| MINUS (> (< |concept|))

    • X =  " < |concept| "

  • Removing any redundant concepts (ie subsumes another concept) from a set of concepts

    • ^ |ref set| MINUS (> (^ |ref set|)

      • X = " ^ |ref set| "

Find the root concepts of a set of concepts - example use case (find the proximal set in the international core / or in the IPS) - example:

•  

  • Example use cases for 

    • Root nodes of an extension module:   (< |concept| {{ module = X}}) MINUS (< (< |concept| {{ module = X}}))

      • X =  " < |concept| "

    • Only the 'root' concepts from a set of concepts

      • ^ |ref set| MINUS (< (^ |ref set|)

        • X = " ^ |ref set| "

• X MINUS (> X)

• HOMEWORK - Suggest some syntax for this.

  • leaves(X) - eg

    • leaves (> |concept| {{ C moduleId = 1234 }} )

    • leaves (< |concept|)

    • leaves (^ |refset|)

    • Pros: Easy to read

    • Cons: More consistent with the long form of ECL rather than the short form

  • L(X)    - eg

    • L (> |concept| {{ C moduleId = 1234 }} )

    • L (< |concept|)

    • L (^ |refset|)

    • Pros: Easy to type

    • Cons: L could mean anything? English specific.

  • _ (X)   - eg

    • _ (> |concept| {{ C moduleId = 1234 }} )

    • _ ( < |concept|)

    • _ (^ |refset|)

  • !_

    • Pros: Looks like lowest / floor

    • Cons: No equivalent highest / top symbol

  • !<

    • Pros: Easy to type. Familiar looking syntax.

    • Cons: May be too similar to children of - confusing?

  • !!< (bottom) !!> (top)

    • Pros: Easy to type. Familiar syntax. Different enough from <! and >!

    • Cons: None that I can think of

  • ⌊ X ⌋  (bottom),  ⌈ X ⌉  (top)

    • Pros: Matches existing mathematical syntax for the Floor and Ceiling functions, which have similar meaning.

    • Cons: Could be challenging for some people to type on the keyboard

  • <!!  (bottom),  >!!  (top)

    • Pros: Easy to type. Familiar looking syntax. Won't be mistaken for children of. Both top and bottom can be represented clearly.

    • Cons: Too long? (Makes operators three characters rather than two).

ECL v2.1 - Requirement proposals (to be archived)

All

Potential requirements for ECL v2.1 - Discussion and brainstorming

• Daniel's comments

  • An explanation of the results of this ECL: <<385540001 |Olmesartan containing drug| MINUS (<<763158003 |Medicinal product (product)|: { <<127489000 |Active ingredient| = <<412259001 |Olmesartan| })

  • Cardinality of focus concepts is not supported (Example use case: which concepts have 2 or more stated IsAs)

  • (Related) Using 116680003 |Is a (attribute)| in refinement/dotted expressions, e.g. 395026007.116680003, 395026007:[2..*]116680003=*
    Snowstorm and Ontoserver give different results for dotted expressions.
    https://snowstorm-alpha.ihtsdotools.org/snowstorm/snomed-ct/fhir/ValueSet/$expand?_format=json&count=10&url=http:%2F%2Fsnomed.info%2Fsct%3Ffhir_vs%3Decl%2F395026007.116680003
    https://r4.ontoserver.csiro.au/fhir/ValueSet/$expand?_format=json&count=10&url=http:%2F%2Fsnomed.info%2Fsct%3Ffhir_vs%3Decl%2F395026007.116680003

  • We've done history for concepts but not for any other components...

    Realizing the information on inactive concepts is different and more extensive from other components, {{ + HISTORY... }} is (just) sugar and maybe just maybe there might be reasons to sprinkle other components with this sweetness (or not).

    Examples: previously active refset members (and their fields), previously active descriptions (e.g. Refers to association refset), ...

• Context supplements - e.g.

  • << 56265001 |Heart disease|  {{ + CONTEXT }} – This syntax is too general, as there is a risk of including absent finding, not-done procedure and family history

  • << 56265001 |Heart disease| {{ + CONTEXT-DEFAULT }} ? – What would this mean?

    • Brief form:

      • [[@ecl_query]] {{ + Context (Temporal = [[ @temporal_value]]}}

    • Expanded form:

      • [[ @ecl_query ]] OR  (< 243796009 |Situation with explicit context|:
                { ( 246090004 |Associated finding| = ( [[ @ecl_query ]] ) 
                       OR |Associated procedure| = ( [[ @ecl_query ]] )
                 ( |Procedure context| = |Done| OR |Finding context| = |Known present|),
                    |Subject relationship context| = |Subject of record|,
                     |Temporal context| = [[ @temporal_value ]] } )

    • Example 1: << |Heart procedure| {{ + Context (Temporal = *) }}

      • <<  |Heart procedure| OR  (< 243796009 |Situation with explicit context|:
                { 246090004 |Associated finding| = << 56265001 |Heart disease|,
                    |Procedure context| = |Done|,
                    |Subject relationship context| = |Subject of record|,
                     |Temporal context| = * } )

      • Example 2: (<< |Heart disease| OR << |Heart procedure| ) {{ + Context (Temporal = *) }}

        • <<  |Heart procedure| OR  (< 243796009 |Situation with explicit context|:
                  { ( 246090004 |Associated finding| = (<< |Heart disease| OR << |Heart procedure| ) 
                         OR |Associated procedure| = ( << |Heart disease| OR << |Heart procedure| ) )
                      ( |Procedure context| = |Done| OR |Finding context| = |Known present|),
                      |Subject relationship context| = |Subject of record|,
                       |Temporal context| = * } )

    • << 56265001 |Heart disease| {{ + Context (Temporal = *, FindingContext=<<|Known present| }}

      • Will return all types of heart disease, plus concepts like 394886001 |Suspected heart disease (situation)|, and 429007001 |History of cardiac arrest (situation)|

      • Expands to: 

        • << 56265001 |Heart disease| OR
            (< 243796009 |Situation with explicit context|:
                  { 246090004 |Associated finding| = << 56265001 |Heart disease| } )

  • However, you may want to exclude (or include) specific contexts - for example:

    1. To ensure that the finding was about the subject of the record (and not a family history, e.g. to exclude 429959009 |Family history of heart failure (situation)|), you could say:

       • << 56265001 |Heart disease|  {{ + CONTEXT (relationship = 410604004 |Subject of record| }}

    2. To ensure that the finding was 'Known present' (e.g. to exclude 394926003 |Heart disease excluded (situation)|), you could say:

       • << 56265001 |Heart disease|  {{ + CONTEXT (finding_context = << 410515003 |Known present| }}

    3. To ensure that the finding was about the subject of the record AND known present, you could say:

       • << 56265001 |Heart disease|  {{ + CONTEXT (relationship = 410604004 |Subject of record|, 
                                                                                                   finding_context = << 410515003 |Known present| }}

    4. ?? Is there any use case for restricting adding temporal context? (e.g. temporal != << 410513005 |In the past|)

  • Is any more syntactic sugar required? E.g.

    • {{ + CONTEXT (relationship = self, finding context = present, temporal != past) }}

    • {{ + CONTEXT (self, present, ! past) }}

  • Other ideas? Common profiles?

• --------------------------

• Ability to return attribute types (see proposal below)

  • [ attributes ] << 125605004 |Fracture of bone (disorder)|

  • << 125605004 |Fracture of bone (disorder)| . Attributes

  • << 125605004 |Fracture of bone (disorder)| . (<< 125605004 |Fracture of bone (disorder)| . Attributes )

  • [ attribute, value] << 125605004 |Fracture of bone (disorder)|

• -------------------------

• Reverse membership (see below)

  • Which reference sets "contain" the given concept(s) - e.g. 421235005 |Structure of femur|?

    • 421235005 |Structure of femur| . Refsets

    • 421235005 |Structure of femur|. Refsets [ referencedComponentId ]

    • 421235005 |Structure of femur| . Refsets [ targetComponentId ]

• --------------------------

• Other?

Returning Attributes

@michael lawley

• Currently ECL expressions can match (return) concepts that are either the source or the target of a relationship triple (target is accessed via the 'reverse' notation or 'dot notation', but not the relationship type (ie attribute name) itself. 

For example, I can write: 

<< 404684003|Clinical finding| : 363698007|Finding site| = <<66019005|Limb structure| 

<< 404684003|Clinical finding| . 363698007|Finding site| 

But I can't get all the attribute names that are used by << 404684003|Clinical finding| 

•  

  • Perhaps something like:

    • ? R.type ? (<< 404684003 |Clinical finding|)

  • This could be extended to, for example, return different values - e.g.

    • ? |Simple map refset|.|maptarget| ? (^|Simple map refset| AND < |Fracture|)

Reverse Member Of

@michael lawley

What refsets is a given concept (e.g. 421235005 |Structure of femur|) a member of?

• Possible new notation for this:

  • ^ . 421235005 |Structure of femur|

  • ? X ? 421235005 |Structure of femur| = ^ X

Postcoordination Topics

• Discuss feedback on transformation implementation

  • Resources

    • Expression transformation service generates the classifiable form and the necessary normal form from a close to user form expression

    • Contribute your expression examples and write your feedback for consideration and discussion

  • Recap of SNOMED on FHIR discussions

    • What is the functionality scope of a terminology server that supports postcoordination? For example, does it include:

      • Classifying multiple expressions in a single substrate? What are the use cases for this?

      • Assigning (local) identifiers to expressions? What are the use cases for this?

      • Autogenerating or assigning a term to an expression? What are the use cases for this?

    • Does a terminology server that supports postcoordination, include all the functions of an expression repository?

    • What is the relationship between a terminology server that supports postcoordination, and an expression repository?

  • Outstanding questions

    • What are the pros and cons of extending SCG to allow an expression as the focus of a postcoordinated expression?

      • Note: This was raised in context of a NNF generated over a postcoordinated substrate, where the proximal parent is an expression

    • Example of using expressions in focus concept

      • ( 125605004 |Fracture of bone|:363698007 |finding site| = 84167007 |Foot bone| ) : 
        272741003 |Laterality| = 7771000 |Left|

      • 125605004 |Fracture of bone|:363698007 |finding site| = 84167007 |Foot bone| , 
        272741003 |Laterality| = 7771000 |Left|

    • What is the expected NNF when classifying an expression that is equivalent to a precoordinated concept? For example:

      • Expression that is equivalent to 111273006 |Acute respiratory disease|

      • 64572001 | Disease (disorder) | :
        {263502005 |Clinical course (attribute)| = 424124008 |Sudden onset AND/OR short duration (qualifier value)|}
        {363698007 |Finding site (attribute)| = 89187006 |Airway structure (body structure)|}

      • Options:

        1. 111273006 |Acute respiratory disease| :
           {263502005 |Clinical course| = 424124008 |Sudden onset AND/OR short duration|}
           {363698007 |Finding site| = 89187006 |Airway structure|}

        2. 50043002 |Disorder of respiratory system (disorder)| +
           2704003 |Acute disease (disorder)| :
           {263502005 |Clinical course| = 424124008 |Sudden onset AND/OR short duration|}
           {363698007 |Finding site| = 89187006 |Airway structure|}

        3. Other?

  • Recap of internal discussions with Content Team

    • Inter-attribute dependencies

    • Grouping rules

Dynamic Templates

• Continue discussion on dynamic templates

  • Inter-attribute dependencies

    • Acute/Chronic and Inflammation - Adding a clinical course requires specializing the inflammation morphology 

      • E.g. |Pyelonephritis| : |Clinical course| = |Chronic|
        should be
        |Pyelonephritis| : |Clinical course| = |Chronic|, |Associated morphology| = |Chronic inflammation|

      • E.g. |Pyelonephritis| : |Clinical course| = |Sudden onset AND/OR short duration|
        should be
        |Pyelonephritis| : {|Clinical course| = |Sudden onset AND/OR short duration||, |Associated morphology| = |Acute inflammation|

    • Infectious Causative Agents - Adding a |causative agent| = |Domain Bacteria| or |Virus| requires adding a |Pathological process| = |Infectious process|

      • E.g. |Nephritis|: |Causative agent| = |Domain bacteria|
        should be
        |Nephritis|: |Causative agent| = |Domain bacteria|, |Pathological process| = |Infectious process|

    • Congenital and Acquired - Adding an |Occurrence| of |Congenital| to a focus concept with an abnormal morphology, requires adding a |Pathological process| of |Pathological development process|

      • E.g. |Koilonychia|: |Occurrence| = |Congenital|
        should be
        |Koilonychia|: |Occurrence| = |Congenital|, |Pathological process| = |Pathological developmental process|

    • Situations with Explicit Context 

    1. if the procedure context = |Planned|, then the temporal context should be << |Current of specified time|

       1. If the procedure context = |In progress|, then the temporal context should be << |Current|

       2. If the procedure context = |Performed| or |Done|, then the temporal context should be << |Current or past (actual)|

    • Note: for this use case (of |Procedure with explicit context|) perhaps we just recommend (or require) that the full role group is spelled out.

    • Next steps

      • Representation of the content rules

        • Who creates the complete list of rules and how?

          • What formalism?

          • Determine which are mandatory and which are optional

        • Implementation of content rules - e.g.

          • Guided data entry by pre-populating role groups in expression template based on definition of focus concepts (for design-time use, such as mapping)

          • Mandatory content rules could be added to transform process

Postcoordination Use Case Examples

All

Example 1 - Dentistry / Odontogram

• Requires an expression template to create expressions.

• Resulting expression still requires a transformation to make it classifiable

Example 2 - Terminology binding

• Uses a fixed expression template to combine codes entered into separate fields

• The procedure+laterality example still requires a transformation to make it classifiable

Example 3 - Mapping

• Design-time activity

• Map targets may not be able to be fully represented using concept model attributes

• In many cases, an extension (with primitive concepts) should be recommended where there are gaps in the mapping

• There may be some cases in which postcoordination is helpful (e.g. LOINC to SNOMED CT map)

Example 4 - Natural Language Processing

• Usually run-time activity.

• May require manual confirmation of coding suggestions (unless low clinical risk, eg for suggesting relevant patient records for manual review)

Postcoordination Guidance

@Anne Randorff Højen , @Kai Kewley

Practical Guide to Postcoordination

• Proposal - Use syntax (i.e. braces) to distinguish refinement vs new role group

  • There should be a syntactic distinction between refinement and constructive addition (ie adding a new role group). That is:

    1. 83152002 |Oophorectomy| :  405815000 |Procedure device|  =  122456005 |Laser device|

       • is classified as (i.e. the refinement is added to the role groups in the definition of the focus concept(s)):

         • 83152002 |Oophorectomy| :  
           { 260686004 |Method| = 129304002 |Excision - action|,    405813007 |Procedure site - Direct| = 15497006 |Ovarian structure|,   405815000 |Procedure device|  =  122456005 |Laser device| }