Pattern for modelling of scores (and PEWS)

Hi Mikeal,

I’m not sure if it exactly the same issue but we have been modelling the very new UK PEWS2 score (re-branded to SPOT)
and had to deal with similar issues.

It is not (quite) up on the Apperta CKM yet but here is the current archetype.

https://tools.openehr.org/designer/#/viewer/shared/Pz9zaGFyZWRJZD0xJDdkMGNiNGVlODNlZDQwNjFhZDQ1NzdlYzY2YmVmMzcz

@AlanF was working with this.

As an even more complex example of a similar issue we also had a real ‘wrangle’ with Uk Sepsis scoring which as well as having different age ranges, has a set of different ‘settings’ hospital, community, untrained staff etc. but with a fair degree of cross-over of the different datapoints and valuesets.

https://tools.openehr.org/designer/#/viewer/shared/Pz9zaGFyZWRJZD0xJGNjOWI5Yjc1OGFiNDRiYjZhN2Q1Y2EwMjUwYTQ0YTNl

was where we got to as the optimal approach though it does still look messier than we would like.

Hi Ian,

Thank you for your answer! It seems like you have used another modelling approach than I have expected for our Swedish PEWS. Do you know where I can find the specifications of the UK PEWS2 score the UK archetype is based on so I can compare the different archetype requirements before I dig down in the different modelling approaches?

Regards
Mikael

Hi Ian,

Ok. It seems like your UK PEWS are not really similar to our Swedish PEWS. Our Swedish PEWS seems to be suitable to model (in one way or another) in a similar pattern as the archetypes Paediatric Early Warning Score (PEWS) and PEWS - simple variables. However I assume that was not possible for you UK PEWS since you modelled it as one single observation archetype using a slightly different modelling pattern.

Regrads
Mikael

I think it is open discussion TBH. We had started on the PEWS2 work a while ago and were not aware of the recent pattern suggestion that you pointed to . We will have a look and see if it might work. Our archetypes are not published yet,m so there is time!!

@David-Jobling @AlanF ??

1 Like

Hi Mikael,

You can see that I’m the author of these new archetypes, currently located in a public incubator as we consider the feasibility of the new pattern.

As there is no standard model for the component variables that make up PEWS, in fact there is probably at least 30 separate approaches amplified by the variations required for different age groups, this archetype has been designed to provide a framework in which one specific PEWS representation can be selected and added to the ‘PEWS components’ SLOT added as CLUSTERs in a template.

So we have developed a single OBSERVATION that will hold the score and method/model used - https://ckm.openehr.org/ckm/archetypes/1013.1.5149 - and a SLOT to carry one of the many possible age/region-specific component CLUSTERs such as the original specific parameters here - https://ckm.openehr.org/ckm/archetypes/1013.1.5148 or a generic framework -https://ckm.openehr.org/ckm/archetypes/1013.1.5167.

PEWS seems to be a very messy concept, used inconsistently between organisations/countries but consistently within an organisation (I assume) or at least for a given patient during a single episode in a single institution. Hence the focus on a pattern that can identify the OBSERVATION easily, but use method/model or the included CLUSTER to identify the specific content. Queries over different method/models or CLUSTERs for content are clearly to be avoided, and not suitable for comparison or research.

It seems that we have a choice between many, many CLUSTERs or many, many OBSERVATIONs. With these archetypes, I’m suggesting the single OBSERVATION framework and many CLUSTER option. We can potentially use the generic framework CLUSTER as a pattern for others… maybe, TBD :upside_down_face:.

The Swedish use case seems to dive into more explicit detail for each age group, beyond the detail in the generic framework CLUSTER. I’d suggest modelling the Swedish each age group in a similar way, but representing the detail accurately, as a group of consistent but separate CLUSTERs, clearly identified and grouped by archetype ID and concept name.

2 Likes

Hi Heather,

Yes, I saw that you are the author of the archetypes and that was one of the reasons I started up the discussion in this thread. :slight_smile:

It seems like you are proposing the modelling pattern we already have sketched and I am currently implementing a draft of. I hope to be able to complete the draft modelling and English translation during this week, so I can share the archetypes with you for discussion.

Regards
Mikael

1 Like