I’ve had a question from a colleague about scientific basis for modelling styles, which I had to say I don’t know of. The specific example was how to model a diagnosis with qualifiers, for example the diagnosis “renal tubular atrophy” with modifiers “not detected, light, moderate, pronounced, not classified”.
We may have our professional opinions about how this should be modeled for the greatest possible expressivity, scalability and safety, but does anyone know of any published works that discuss this?
I’m not aware of any formal published literature on this, other than endless papers on how ‘ontology’ will magically make this all work.
That specific example is hugely driven by reporting/analytics requirements, and very specific to renal tubular atrophy as a condition. That;s not wrong but is just a very particular perspective. The only paper I know of that comes close to addressing how health information evolves /morphs as it is is workflowed through a system is
If I was investigating tis academically I’d start there and look for citations from more recent work.
Has anyone access to the CIMI documents? While not being strictly scientific there was a real wealth of discussion points regarding modelling styles and approaches. Unfortunately, this seems to be not available anymore on the website.
Good idea. A search for material by ‘Stan Huff’ (CIMI lead) and ‘clinical modelling’ does provide some good stuff
https://academic.oup.com/jamia/article/21/6/1076/786753but also brought up this world-beating analysis…
Stan and his team approached me recently to understand our approach to modelling ‘Anatomical location’ so that they could provide input to the FHIR work. This triggered the Anatomical location thread after I couldn’t explain to them the use of ‘bilateral’ within the current scope.
They had additional advanced use cases including the location of a breast lump in 3D in a patient lying prone or with the breast compressed in a mammogram, especially if they needed to clearly distinguish one particular lump from others. We haven’t modelled this explicitly yet, so something else to consider for future work.
The circle keeps going around and around… Personally I love that we can all learn from each other and that we are willing to drop the egos/pride and are willing to do so. We will all benefit.
@ian.mcnicoll Thanks for the reference to the Huff paper - this is very much what I was looking for . They present some rules, for example “put static knowledge in terminology and instance data in the model”. In case of tubular atrophy, this is instance data (because changing with the patient) and therefore has to be put into the model.