The definition of “has intended site” is wrong (or doesn’t align with the range).
You don’t need to look far to see:
- 738956005|Oral (intended site)| - doesn’t mean “pertaining to the mouth” oral site.
It means something like “taken by the mouth for systemic effect via entral route”.
<738982001|Oromucosal| - “pertains to the mouth” - but are not subsumed by “Oral”. - Similarly <385287007|Parenteral dose form (intended site)| or <738904002|Cutaneous (intended site)| and 738987007|Transdermal (intended site)| - are not “sites”. The upper part of my arm is a site - to which I might receive an injection or apply a patch (systemic effect) or cream (local).
EDQM might call this “intended site” but that doesn’t mean it’s right. (I assume the definition is lifted from there?)
What want for the drug model is a route of administration. Because that’s what’s clinically relevant.
- FHIR has MedicinalProductDefinition.route **“**The path by which the product is taken into or makes contact with the body”
- SNOMED CT has Procedure.HasRouteOfAdministration for concepts like 698511005|Administration of hormone via oral route|
This hints at what the use case is - closed loop logic between medication instructions (pre or post coordinated) and dispense.
For example if I have a medication instruction “Paracetamol 500 mg p.o.” .
This might be coded in the back end using
777067000|Paracetamol only product|
#500 258684004|mg|
26643006|Oral route|
Possible medications for dispense should then be able to be identified using ECL, but they can’t be because the dose forms use “intended sites”.
We’ve been trying to support this exact use case for years in our dose based prescribing support (particularly for hospitals) - but it just doesn’t work due to the current modelling. This has been a known issue since 2017.
I suspect the motivation for creating “intended sites” is probably similar to that for Unit of Presentation - in that the existing content was a bit messy, and EDQM alignment…
The problem is also that for both areas - extensions are going to need (already are) add further content - so the range for both is never going to be a pure mirror of EDQM - which should never be the goal anyway.