Meeting minutes
Enhancing our FHIR RDF ontology
Ken's draft: https://
(Ken discusses the above document)
dbooth: https://
… I think the best way we can proceed is to not try to tackle the problem top down (eating the whole elephant), but bottom-up, incrementally adding value to our existing FHIR ont.
… Could you write up an initial use case and propose a specific ontology addition that we could consider?
ken: Yes.
… Want to address the eligibility of individuals with housing instability. Eric Jahn is very involved in this.
… There are also other use cases, not sure other people would be interested.
dbooth: Please propose them and we'll see.
ken: Natural Language Processing (NLP), an ont to define the meaning of a phrase or word might fit well.
dbooth: How would that relate to LLM successes?
ken: Each LLM has its own premises and data. The value of a shared model to train LLMs.
ken: Someone doing cancer analysis might use parts of this, but might also need a different view.
erich: What's the end game? More than an OWL ont?
ken: Yes. Company is very focused on data models. My background comes from business of modeling,
eric: What's the end game?
ken: Want to be able to say that this data element in this model is equiv to that data element in that data model.
… Different gurus have different interpretations.
… No single system has all of the data about the pt. Need to exchange data with different systems
ken: Also may want references to other models
… Should also ref schema.org
dbooth: Don't be shy about referencing other things, such as schema.org. We can partition the ont if desired, to be modular.
ken: There are things in the CMS model that are not in the FHIR model.
… What's the core starting point?
dbooth: From a practical perspective, we need to view our existing FHIR ont as our core starting point.
ken: It's very focused on clinical model. Doesn't support some of the things that are needed in this housing eligibility model, for example.
ericp: You mentioned schema.org, Eric Jahn's work. FHIR is very structural, not very ontological.
ken: FHIR is essential in clinical domain, but in social domains like housing, something called FHIR doesn't necessarily help.
ericp: We'll have use cases that w prob have subsumption, and will want clever ont, which we won't get from FHIR directly. Nothing that says medication order is related to procedure order.
… Might be able to extract some interesting stuff from v3.
rob: Yes, there's some stuff in there.
… I think a proc order is related to a med order. They follow the same pattern.
ericp: Good point. We don't capture that in the ont.
… Maybe we could extract more ont relationships from the bindings.
… The inferential value we get might help meet these use cases.
ACTION: dbooth to create an issue that describes where semantic connections are defined in FHIR
rob: There's a place in the structure def for those things
dbooth: issue created: w3c/
ADJOURNED