Meeting Introduction

@Peter Williams

5

Start Zoom Recording, review of agenda: MAG Questions in flight

Membership update

October Business Meeting - 2 x 1hr meetings at 20:00 UTC on Monday 5 October and Thursday 8 October.

Concrete Domains 

@Peter Williams

40

Concrete Domains Update on Consultation

SNOMED International Proposal for Representing Concrete Domains in RF2

Socialise Plans and recent clarifications:

• Hash and Double Quote symbols to be used in RF2 as well as MRCM

• Double quotes to be escaped with backslash if they are used within the string

• "Decimal values must have a decimal point" (ML) so we can infer that number without a decimal point are integers.

• Internal Development Analysis

• Proposed Timeline. GL Early

Discuss

• Scale and precision

  • The # symbol in the file doesn't allow the distinction to be made between integers and decimals; this can only be determined by parsing the MRCM. Can we be explicit and say that, for example, 2 === 2.000? From Toni:  "Repeating decimals are rounded to three decimal places (with 5 and above rounded up and 4 and below rounded down)”. Suggestion that whatever is decided is applied to all of SNOMED.

• Language encoding of String (@Yongsheng Gao is this like human language - how do we indicate if the text is English or French - or do we just need to say that it's UTF-8 as per all description terms and remember to escape any double quotes ?). Suggestion that the (human) language is derived from characters but this is not safe eg café. Suggestion that we could extend the string format further, but really if we need to be translating some text we should be using proper Descriptions. Remember (RC) that this value is part of the definition of the concept. Titles of books , for example, are not part of the definition of the book. Drug company names should be concepts, but (LB) some companies register different trade names in different countries. A branded drug is partly defined by the brand name.

Do we have a recommendation for how to handle the file if implementers should choose to combine the inferred relationship file and the concrete value file?

Answer: SCTIDs could be represented in the same column without a prefix. The data type of the field in that case would be String.

Precision: DM - AMT picked a number, See https://www.healthterminologies.gov.au/library/DH_3243_2020_SNOMED-CT-AU_Australian-Technical-Implementation-Guide_v2.5.pdf?_filename=DH_3243_2020_SNOMED-CT-AU_Australian-Technical-Implementation-Guide_v2.5.pdf 8.4.4.1

Stefan question: What about authority-defined disease concepts e.g. "Hypertyreoidism as defined by the Japanese Association of Endocrinology"

ML: String language may be relevant for attributes with characteristicTypeId = 900000000000227009 |Additional relationship (core metadata concept)| (so not part of the ontological definition)

.

Allergy/adverse reaction concept model by substance and/or product

@Yongsheng Gao

30

Allergy Adverse reaction to substances products v3.1.pptx

Allergy Adverse reaction to substances products v2.pptx

Note that whatever pattern is selected should be able to be applied to any condition caused by a product / substance eg poisoning caused by X.

Use of the MOVED TO historical association

@Peter Williams

10

Further support for adding the module id to the international edition when we move any concept to some downstream module. DM: An edition URI would be more helpful in this case.

This is a solution to a different problem from that of not having a replacement concept, so use of PossEquivTo could be progressed separately from the evolution of the MovedTo association.

ML: "poss_equiv is problematic because it may not be 'equivalent' and you may only be able to reference a parent Ie was-a"

Dialect Aliases

@Peter Williams

10

GR: Isn't this more generally a question of refset metadata?

Could we solve this more generally with a CCS or CSS refset ie name/value pairs of which alias could be one particular flavor?

AOB

To discuss next time: Semantic Tags should be part of official metadata so we know what is legal. Also hierarchy.

Anatomy Revision Proposal

@Yongsheng Gao

Review the proposal for revision of bones and joints of the limbs and girdles

Degree of Synonym / New Description Types

@Yongsheng Gao

Annotations for concepts or descriptions

FSNs greater than 255 characters eg for vaccine concepts. Steer from SI is that we don't want to be changing RF2 description limits for the sake of a single use case.

Basic Formal Ontology

@stefan.schulz (Unlicensed)

@Jim Case

Requirement to demonstrate benefit to implementers / end users if we are to fund a project on this

Review and feedback for the document at https://docs.google.com/document/d/1HcBj5bVIg8lB_uyORZU9A_FWKFsw0sxmB6Xg4UYKygk/edit#heading=h.izs8q5fkm2

MAG discussion here: Basic Formal Ontology Discussion

Metadata

Reference set metadata See Agenda item 17 and consider in wider context of machine readable metadata item 11. Traction has been lost, is a sub-group needed to progress?

Historical ECL

@Peter Williams

@Jeremy Rogers (Unlicensed)

Historical Analytics

Negative Delta

@Peter Williams

@Andrew Atkinson

Requirements and guidance for standalone SNOMED CT OWL documents

@Former user (Deleted)

1. Inclusion of copyright text

2. Ontology version

3. Annotations including description types

4. Language codes for annotations

Versioning of Templates

@Guillermo Reynoso

From the EAG: Templates are becoming first class citizens and would be useful to other members. They should be given unique identifiers, packaged and versioned like any other RF2 component.

Not to suggest this is a suffiently robust solution, but SI templates do take an integer version number which is present both in the JSON representation which is deployed to Production and on our confluence pages. For example see Wound morphology

Modularizing the core and/or Migrating content to a dependant module

@Peter Williams

@Jeremy Rogers (Unlicensed)

Check for EAG discussion on same.

@michael lawley is the previous work you did on module composition (as opposed to dependency) available?

Use cases:

1. Return of content to the UK Edition,

2. Separation of clinical core from "Community Content"

Content would need to be inactivated in the International Edition and then the same identifiers would be re-activated in the next UK Edition using their own module. FYI @Jeremy Rogers (Unlicensed)

The apparent "break in time" can be resolved by detection of the "Concept Moved Elsewhere" inactivation indicator and the historical association to the concept representing the receiving namespace.

Actually, it's easier to move to another edition than it is to move within the same release. Would we end up with an active concept in one module, and an active inactivation indicator in the other?

The receiving module should inactivate both the inactivation indicator and the historical association - agree?

The inactivation and reactivation in a new module within a single package is going to be very noisy. Can we make a clean cut and just restate the new modules?

Previous work on Module Composition - discussion including link to my proposal https://snomed.atlassian.net/wiki/x/eIL8Bw