resolving TBD LOINCs
- Improving value sets within the IG
- Patrick: TBD codes need LOINC concepts (to be created)
- red-highlighted codes are no longer used in our IG
- functional annotation will be replaced by molecular consequence
- several new concepts still require definitions, draft defs added for molecular consequence, callable, and not-callable
- variant inheritance: def contains a link to SO hierarchy, should the def contain a binding to a value set?
- Jamie: the spreadsheet needs to be developed fully before LOINC can create any codes for us
- Liz: mode of inheritance - current def in spreadsheet is not correct, Liz will fix
- Patrick: will transpose content to LOINC's template
- Patrick: content review - are we missing anything?
- Liz: how does associated cancer differ from associated phenotype?
- Jamie: assoc cancer is intended to recap a statement from a knowledge base, not used for diagnostic statement, intent was ... <moved too fast, Bob couldn't keep up>
- May: are all of these intended to be codeable concepts?
- Jamie: examples from Tempus reports, medications used in other cancer types/indications/variations, need a way to express those even when they don't match the patient's specific cancer or var profile (e.g., approved therapies for a different indication)
- Patrick: assoc pheno and assoc cancer could be merged, but contextual indication is different
- Jamie: need to determine whether to model this as structured/computable vs. text blob
- Liz: what relationship/statement are we trying to capture?
- Kevin: need to differentiate between patient's diagnosis from what the drug treats
- May: this references a knowledge artifact, may appear to be prescriptive when it isn't
- Jamie: could use 1 associated pheno field and move complexity to the interpretation section of the report
- Liz: knowledge bases like ClinVar should hold the evidence, be used/referenced in overall report
- Jamie: remove associated cancer, update associated phenotype def to pertain more broadly
- intent is to match a var observed in a patient to something in a knowledge base
- Patrick: need to capture "given the observed vars, patient is likely to develop this cancer type"
- Jamie: that would be a diagnostic implication, not medication implication
- Kevin R-P: must ensure add to docs to make it clear that cancer should be expressed as phenotype
- Jamie: differentiate risk-of from context (see FHIR-26945)
- susceptibility is from germline test, different from genomic profile from a tumor
- Kevin P: need a way to codify risk
- Jamie: can leave some of that up to the code system
- Liz: is region coverage (% value) already in LOINC?
- Liz: exact start-end 81254-5
- Patrick: that def is out-dated, references only start position, need to capture both start and end
- 81302-2: looks good
- need to review existing LOINC defs
- diagnostic-implication / therapeutic-implication: Do we need codes?
- Will be in STU2 (voted on by group as a consolidation of multiple profiles)
- For new loincs, try to phrase as a question, send a link to the definition, send an example report, determine the data type
- Jamie: variant implications (AA change, functional effect, pathogenicity) and diagnostic implications
- should those statements be associated with variant or diagnostic?
- Bob F: this is where the SEPIO framework could help, see VA work on statement modeling, need a balance between modeling as discrete components and capturing text blob
- <reviewing concept defs in Sequence Ontology>
- Joel S: need to be able to associate multiple vars with a single outcome/effect/risk, too simplistic to tie implication to a single var
- Kevin P: we can associate to a group of vars
Today 2:01 PM
would "risk-of" fall in prognostic implication?
My goto article for those PGx terms: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5253119/
That is a very good question Liz ... also what we will need to be able to support is polygenic risk scores.... which by definition are risks
Additional IG content/documentation
Jamie looking for "volunteers" to go through recent slide presentations and create some introduction material to help promote use of the IG.
Multiple examples could be walked through in the IG or in an external resource
May: mCode created a searchable KB in Confluence to help implementers, will provide demo if needed
Arthur: interns could increase bandwidth to get work done on the IG - can we get some?