2016-04-18--20 - OBSERVABLE Meeting

1

Welcome and apologies

@Daniel Karlsson

• Remember to record!

2

Conflicts of any kind

@Daniel Karlsson

3

Previous minutes & action items

@Daniel Karlsson

(4)

Intro. to observables - IF NEEDED

@Daniel Karlsson

5.1

SNOMED CT - LOINC cooperation

• Presentation of LOINC alpha release phase 3*

• Expectations and use cases for the LOINC-Term SNOMED CT expressions*

• Discussion and resolution of issues(*)

• Discussion of LOINC panels as orderables and deployment in SNOMED CT

@Farzaneh Ashrafi

See Farzaneh's slide set. Alpha release phase 3 feedback due by May 31st 2016.

Questions:

Jim Campbell. Are the terms there also?
Scott Campbell:Terms are integrated in the release.
JC> Numbers? FA About 14,000 terms representing about 70% of lab terms.

LOINC Mapping Issues Slides

Mapping X+Y concepts.
DK> Subsumption is affected whenever modeling multiple components.
DK> These are disjunctions not conjunctions.
JC> Is it true there are individual as well as panel for each of the components?
DV> Yes

JC> We can manage it better if we create an enumeration - a series of observations. e.g. in a separate ref set.
DV>  Not all of these are panels, some are tests that cannot distinguish between the components with a Positive/Negative result.

DK> This is the same problem as Total Amino acid and more specific amino acids issue, because we don't have disjunction in SNOMED. We could push the problem to substances.

FA> Electrolytes as substances e.g. Sodium is a cation, Chloride is an anion. Model to the electrolyte or to the parent substance?
MO / RD others>  It appears correct to model to the more specific substance i.e. the electrolyte.

5.2

Cancer check list*

• Identify any critical issues

• Inform IPaLM/Observables group on model deployment in anatomic pathology

@Jim Campbell

@Scott Campbell

See Jim/Scott Campbell's slide set.

Goal: Create a useful semantic layer and code the cancer check sheets to share between entities/nations.

Jane Millar> Timeline on the observables.

JC> 2 weeks per check sheet.

Lazlo Igali> Model must be fit for use first. Nebraska work is useful as a trial.

KC> Need to derive constraints from the model.

MK> So to do this the observables model need to be more mature,

JC> Observables model is abstract and difficult to apply. We would like to get as many groups as possible to review our work & remove bias.

MO> Does IHTSDO have a mandate to do this work or does LOINC.

JC>How can we publish merged content between LOINC and SNOMED.

DV> This work would help...a number of these check lists have been modeled already in LOINC.

We could and should think about making the interchange bi-directional.

Raj Dash> IHE licencing of SNOMED CT is a problem in non member countries.

Jane Millar> IHTSDO and IHE are in discussions about use of SNOMED as a value set. They are talking about 300 codes across all profiles. Confine to use cases

 Raj> I will take this back to IHE and get use cases.

JC> If NLM agree we will publish the terminology with bindings to the check sheet.

Jane Millar> You need to be specific, what is the international requirement, use case etc...

5.3

Molecular pathology*

• Present current UNMC approach with specific, deployed examples

• Discuss the model as deployed and attempt to reach a consensus for the approach

@Jim Campbell

@Scott Campbell

See Jim Campbell's slide set.

Jim Campbell proposed a relationship with NCBI  www.ncbi.nlm.nih .gov to include genes in SNOMED CT.

Jim's proposal requires concrete domains to determine the start and stop nucleotide position on the gene.

Jim Case> Perhaps refer to HCNC 1097, and just use the gene name.

Jim Campbell> Raj Dash suggested this also

JCA> Sounds like it is a resource not the concept model.

Jeremy Rogers> Agreed that it makes more sense to point to a link to the definition, but that makes it primitive because you can't reason over a link. We don't want to mimic GO. Is this an extension to SNOMED? I understand there is a clinical requirement to refer to genes. Can HGNC publish their own extension to SNOMED?

Paul Amos> This could use a pattern where  Cancer Institute could do the authoring for us.

Jim Campbell> Save it all and release as an extension. So we maintain a ref set for all the genes that we name and a link to the HGNC.

MO> Custom ref set?

JR> Simple refset with a link....

JCB> I would have thought we also take some information form the HGNC website and include this.

Bruce Goldberg> Any thought been given to the lengthy proteins encoded by the gene sequences.

JCB> That discussion is at the edge of the table and we have enough to deal with.

SCB> We have used the presence of a protein as a proxy in staining.

KinWah> Work going on in LOINC could be used.

SCB> We will put this model into production in Nebraska. The use case is oncology reporting.

Jim Campbell and Scott Campbell presented gene nucleotide sequence changes models with examples and a number of changes were suggested.

JCB> We will model these genes as primitives, and build a ref set in the Nebraska Lexicon.

JCB> Talk to Daniel Vreeman regarding LOINC and genes.

JCB> Talk to Yong about use of partOf in the model

JCB> Daniel to follow up on use of HAS VALUE in the BRCA1 gene mutation example.

hierarchy.

.

6

Functioning

@Penni Hernandez

See Penni's slide set.

Problem with concepts like "does not, dependent"; "able to" and does.

Keith Campbell> We should remove all of the negation like qualifiers. e.g. inability

Daniel Vreeman> A lot of these appear in health assessments and I think volition is a great word for having the ability to do something but may not want to do the test.

Penni Hernandez> We don't know what "Does not X" means so we shouldn't define it.

Paul Amos> "With assistance" needs to be recorded

DK> We need to get rid of the ones that combines level of ability with other things.

Penni> Agree.

Paul> Need to consult the CTV3 people who added the concepts.

PH> Action move foraward with changes to the qualifier hierarchy and that will make the bindings more consistent around some of the functioning concepts

7

Observables Inception-elaboration document

@Daniel Karlsson

DK> Issues with component attribute being used in two places procedure and the new observables model.

3 alternatives (See slides)

Farzaneh presented some examples. An interesting one was bacterial id, which was adjusted to remove the component and added attribute inherent location.

  Observables should not be attributes - Jim Case has allowed the creation of qualifiers for appearance and color

No Decision on component, awaiting next  call: Potentially Role group observables or remove the role groups in the evaluation procedures.

8

Feedback and deployment of technology preview as SNOMED CT ontology and map refset; RF2 extension datasets

@Jim Campbell

9

New content in the era of the Regenstrief agreement

@Jim Campbell

10

Bullet points

@Daniel Karlsson

11

AOB

@Daniel Karlsson

* = joint Monday+Wednesday session