Less granular (generalized) clinical drug

SNOMED Groups Drugs Extension User Group
1

Hello DEUSG,

Please use this discussion thread for feedback on an additional ‘BoSS Clinical Drug’ concept potentially being added to the drug model.

Previous comments can be viewed at BoSS Clinical Drug Concept view-only comments

2

DEUSG Summary: General agreement that broader Clinical Drug (CD) concepts are needed. Feedback supports introducing these generalized CDs, for example using Has active ingredient rather than Has precise active ingredient, to support subsumption in national drug extensions.
Next step: DEUSG gathering modeling examples from members to inform decisions.

3

I know we’ve discussed this in the past, and I’m not sure where we are up to with it. So it is written somewhere I thought I’d share my thoughts.

First I’d like to share AMT’s (Australia) experience on this topic, as we grappled with this exact question during the development of AMT v4.

We concluded that attempting to align national extension concepts with international Clinical Drug concepts was not consistently achievable, and more importantly, not worth the cost. The reasons were:

  1. Partial subsumption is worse than no subsumption. When we tested alignment, some AMT concepts would subsume under SI CDs and others would not — due to differences in strength representation, unit of presentation choices, unit normalisation, dose form granularity, and salt/base choices. Implementers found this inconsistency more confusing and disruptive than having no alignment at all.
  2. MP and MPF provide the real integration value. We derive substantial benefit from integration at the MP and MPF levels — therapeutic roles, dispositions, chemical structures, form-based grouping. This is where the analytically useful content lives and where alignment is stable. Alignment is easy and this is where most of the value comes from — good return on investment.
  3. CD is in the zone of jurisdictional variation. The specific strength, salt form, dose form granularity, and regulatory representation of a product varies across countries. It feels like international CD inevitably reflects choices that work for some jurisdictions but not others. Variation is more significant, but the simplest example a product available as 500mg tablets in one country may only be available as 250mg or 1g in another. This creates a sea of variation at the CD level. I’ve witnessed this first-hand classifying dm+d, RxNorm, and AMT together and found levels of genuine product variation (not modelling variation — real product variation) that surprised me.

I’ve said this before, but probably only on calls. I have concerns about the “BoSS Clinical Drug” proposal specifically. While I understand the intent - to provide a less precise CD using the base ingredient - the Basis of Strength Substance has a specific purpose (identifying what substance the strength is measured against). Overloading it with “appropriate clinical abstraction level” introduces semantic creep. The appropriate level of clinical abstraction for an ingredient is a clinical judgement that can vary by ingredient, form, and even jurisdiction. BoSS might appear tempting but is not a reliable proxy for this. In some edge cases the BoSS substance might not be in the product at all - it is just used as a measuring stick.

Furthermore, even if the ingredient abstraction were solved, the other modelling differences (jurisdictional strength and size variation, strength representation, unit of presentation consistency, unit normalisation, rounding rules) still prevent consistent classification between international and national CD concepts.

I do support the idea of creating more general CD’s - something between CD and MPF - but before we design a new layer, I think we need to clearly articulate what use cases CD serves that MPF doesn’t, and what we’d be looking for from this new layer. If we can name the specific use case and show that it requires strength-level specificity from international content - rather than from national product data - then we can design the right level of abstraction for it.

Consider the major use cases for international medication content:

  • Cross-border data exchange works at ingredient level - IPS accepts PhPID Level 1, EU law mandates INN prescribing
  • Clinical decision support (allergy checking, drug-drug interactions, duplicate therapy) operates at ingredient or therapeutic class level
  • Pharmacovigilance aggregates signals by ingredient
  • Dose-range checking needs strength, but this is a local prescribing function using national product codes, not an international interoperability concern

MP and MPF are uncontroversial. Alignment at these levels is straightforward, and the value is clear. At the CD level, variation across jurisdictions in strength, salt form, dose form granularity, and regulatory representation means that international choices will inevitably align with some nations and not others.

For those who see value in a shared layer between MPF and CD - what clinical or analytical question does it answer that MPF cannot? If we can articulate that clearly, we can design something fit for purpose. If we can’t, then perhaps the investment should go into making MP and MPF as useful as possible - including addressing dose form granularity gaps that limit MPF’s utility today.

Happy to provide more detail on AMT’s experience and the specific classification barriers we encountered. But I’d be interested to hear from other’s experiences, and particularly goals which might help the design process.