Reusability in designing archetypes

I’m really interested in hearing your views on reusing (sub parts) of archetypes when designing archetypes.
I’ve been thinking about this for some time, recently with the design of the advance intervention directive (or whatever it’s called this month :p) but what triggered me right now was this example in the EHR model specs: EHR Information Model

The example shows a single archetype for modelling a glucose tolerance test. The observed blood glucose values are modelled inside the archetype. So when querying blood glucose values you’ll have to query separately for blood glucose values as part of this OGTT and ‘normal’ blood glucose values in Laboratory analyte/result archetypes.
I’m not that interested in debating the example design (it may be outdated or ‘just’ an example).
The example with the advance intervention decisions archetype is that we model interventions as part of the archetype while we also have procedure, medication instruction and service request archetypes that could (partly) be reused for this purpose.
I think the main reason for the current approach is that it’s way easier both at archetype design and template design time.
I think most of the tech needed to reuse more parts when designing archetypes exists. Things like clusters, component archetypes and openEHR terminology.
To me it feels not openEHR like to choose just the easy way. I fully agree on choosing pragmatically for single archetypes. But I’m looking to get some different views on doing this in a more reusable way, mostly for intellectual fun and to further my openEHR and health informatics understanding.

(Relates a bit to one of my previous questions about linking data elements in a care plan

Hi Joost,

That example in the specs is misleading because we would definitely do things as you suggest. The OGTT archetype was deprecated a long time ago, so we would just use the 'standard lab test Observation and related CLUSTER archetypes.

I don’t agree with you here. The advance interventions archetype is essentially a set of recommendations, not the formal orders/requests and procedures that might (or not) arise from these. So for me, this is modelled correctly. We do try to drop aspects into clusters where they might be reusable across different contexts.

The other issue which causes a bit of angst is that we (IMO!) have to balance keeping the set of published archetypes reasonably stable, against updating against arguably more optimal reusable patterns, but which would cause significant braking change.

It takes time for these patterns to become apparent and for our understanding of the overall shapes to emerge.

Paradoxically I might suggest that if anything, it is the technical implementation pressures that lead to compromise . It is easy enough for us to build templates that have a lot of tiny reusable components, but can be quite hard for the tech folks to line them up with the customer requirments.

@heather.leslie probably has more experience of this kind of basic pattern wrangling around e.g smoking and tobacco or physical examination.

I learnt a huge amount from doing the histopathology work Clinical Knowledge Manager (

and a realisation that what can seem like a simple, reusable pattern such as lymph node spread, can get very variable depending on tumour type.

Welcome to the world of the archetype-wrangler!!

1 Like

Agree, Ian.

Clinical medicine is messy.

Simple clinical modelling is messier when we try to represent inherently messy data.

Smart clinical modelling is messiest and hardest; when we try to identify patterns that underpin the messy data, to optimise for reuse.

Sometimes we have to compromise; sometimes we have to be pure in our approach.

Clinical modelling is a science, but also an art.

How to identify the ideal balance between academic purity & pragmatic compromise, while understanding the consequences of each?


Cross post with another example and the usecase for modelling with more reusability: Choice between data element and SLOT? - #26 by joostholslag