Despite using the same underlying tool, not all CKMs are equal. Rather than ‘one CKM to rule them all’, each CKM exists for a specific reason – for example intent, geographical domain, servicing a common community. This allows countries, organisations or programs to operate autonomously from each other when they need to. All CKMs operate independently of each other, with their own Editors and Administrators.
CKMs support some baby steps towards federation but this is hugely complex and a problem not yet able to be resolved. The tension can only be managed by the administrators governing each CKM, and the risk is significant divergence or conflicts between models.
So, when searching for archetypes to use in data sets it is important to know where to look, the focus and context of each archetype, and therefore have insight into the pros and cons of each model. Each CKM may have different ways of governing its archetypes and so choosing the best archetype for a given purpose is not necessarily straightforward. This doesn’t mean that you shouldn’t try to identify useful archetypes in various CKMs, you just need to understand any potential consequence of your choice.
There are currently five public CKMs:
1. The openEHR International CKM is commonly described as the ‘source of truth’ for most openEHR modellers. It is ambitious, and by no means perfect, but the goal is to create a coherent health data ecosystem of shared clinical models, avoiding overlap and minimising gaps, reusable across multiple clinical scenarios – effectively establishing a universal health language. Supporting semantic interoperability is the cornerstone of the openEHR CKM and this intent is one of the key differentiators from some other CKMs.
In this context, the underlying philosophy of archetype design is a proactive approach to ‘deconstructing clinical knowledge’, rather than simply replicating existing content found in clinical systems, forms, messages or data sets. It is primarily focused on standardising current clinical knowledge and practice, aspiring to digitally represent all the things that clinicians and domain experts know and do as well as integrating relevant and sensible existing content. The resulting tension can sometimes be difficult to balance – current content vs anticipating future ‘best practice’ - this is the ‘art’ of clinical modelling. All published archetypes and those currently under review have a design intent of inclusivity of all relevant data points, always aiming towards (even if never achieving) a maximal data set and universal use case.
Any draft archetype in a project is a potential candidate for publication and reuse, otherwise it is rejected or kept separate in an incubator. @siljelb and I share that philosophy and protect the CKM library quite fiercely to establish its credibility as a high-quality resource. It's not perfect but an evolving work in progress. We meet weekly to discuss issues and actively manage the content.
2. The Nasjonal IKT CKM mirrors the great majority of archetypes in the International CKM but local Norwegian governance ensures the representation of all archetypes in Norwegian Bokmål and inclusion of archetypes to support local requirements.
The Norwegian modellers work in parallel with the International CKM modellers, running synchronised archetype reviews and publishing/updating archetypes simultaneously. The majority of the archetypes are identical to the archetypes in the International CKM.
3. The HiGHmed CKM references many of the International CKM archetypes, that is, it makes the international archetypes available in their archetype library as read-only. HiGHmed modellers actively contribute, and keep up-to-date, German translations to the International CKM. In this way, HiGHmed modellers have easy access to the International CKM archetypes in German. The HiGHmed governance is focused on the coordination of the modelling effort within the HiGHmed consortium, with processes that appear to mirror much of the international/Norwegian approach. Identification of HiGHmed-created archetypes that are candidates for inclusion in the International CKM has been limited by a lack of openEHR resources, especially the need for translation from German.
4. The Apperta CKM also references (as read-only) many, if not most, of the International CKM archetypes. It does not appear to have a documented governance philosophy or approach but the content design and governance approach of the Apperta archetypes appear to reflect an alternative philosophy where modelling is often closely aligned to existing content and directly mirroring organisational, regional or national UK requirements, rather than aiming towards international interoperability.
To date, only 30 archetypes from amongst the 842 UK archetypes in Projects have been referenced back (as read-only) to the International CKM, and of these 30, almost all of them are scores or scales, for which there is usually little debate about clinical content or divergence in the modelling approach.
The Apperta CKM has only 7 archetypes in incubators, which implies that there is quite a different approach to the use of Projects and little use of Incubators as a ‘sandpit’ space for immature archetypes.
5. The Slovenian CKM is a bit of an enigma and probably best not to use as a resource. It went live in 2013, has 37 registered users. In December 2020, 10 archetypes were added, to support COVID work, and prior to that the most recent changes appear to have occurred in early 2018. My understanding is that these were the archetypes that were used in the first implementation of the Better EHR around 2013.
(If any of this information about the different CKMs is inaccurate, please let me know.)
In addition, there is a Chinese archetype library, the Healthcare Modelling Collaboration, and I believe there is also one in Chile, containing .json ‘archetypes’.
In reality, anyone can build an archetype for any purpose. But is it a good one? It depends on your intent. At one end of the spectrum, the content of an archetype can be really rough, quick and dirty but if it is only used in your system then it really doesn’t matter. On the other hand, if you want to design an archetype for use within the coherent ecosystem, especially where you need broad interoperability, you need to understand how to design the content of that archetypes to optimise the potential for reuse and semantic interoperability. Considerations include a choice of class, the scope and focus of the content, the underlying modelling pattern, and how to document it clearly.
In practice, we have archetypes in the wild that are at all positions on that spectrum. That is fine, as long as we understand the consequences of our design choice. Designing for local use limits future interoperability; designing for maximal reuse can take more time to investigate and document. However, one thing that you can be sure of, if you use an archetype that is either a draft or has not been designed with interoperability as the goal, then there is a real risk of building up technical debt and effectively you are contributing to the data silo problem again. Simply using an openEHR archetype in a clinical system does not provide semantic interoperability ‘out of the box’; using shared and published international archetypes gives you the best opportunity to achieve degrees of interoperability and break down the data silos.