It all is about classifying.
It is all about proper definitions we all share and use.
I believe that when we all interpret the definitions in our own way in our own data bases all is working nicely.
The moment we start to exchange this data we will discover that we are not interoperable,
The moment we start to re-use data in clinical decision support service we will discover that all systems that worked so nicely, have become problems to connect.
As EN13606 we work at full semantic interoperability as much as possible.
So we have to define many things, we use to produce archetypes, properly.
Don’t we all have an obligation to make semantic interoperability possible?
Hello eveybody!
Looking at this discussion, the volume it is taking weighted against the visible progress I ask myself if this is an efficient way to reach harmonised conclusions.
In our experience doing the national EHR standards and the lab report for Austria we noticed:
Once you reach a sufficient level of detail it is not possible to reach generic (one fits all) solutions within available timeframes.
Data models and their descriptions only started to work after all involved had agreed on a specific, crisp and clear definition of the usecase.
No information model / description did work for all thinkable usecases.
We had to tailor information models to specific usecases.
Then bridging from one use case into the other becomes a challenge, as subtle differences in the information will become an issue.
In order to bridge information from one specific use case into the other we typically need to define another specific usecase.
I do not see the specific usecases you are referring to. I therefore can not contribute to or understand this discussion. Maybe I missed something. Sorry!
I actually did think Stefan was making an important point about
clinical modelling. If I have understood Stefan's comments correctly ,
he is stating that in his experience interoperability is rarely
possible without local negotiation, compromise and detailed
understanding of the use case and shared requirements. That would also
be my view, which is why I believe that while good definition and
ontological analysis is helpful, it will never be possible to 'clean
room' interoperability without directly involving the end-use
stakeholders: Interoperability is ultimately negotiated not designed.
The value of the archetype approach is in making this negotiation
progressively possible without asking end-users to become embroiled in
complex technical discussions. Complex and detailed clinical
discussions will still be required but that is only to be expected.
Interoperability is not only about techie, is about people. Peopleware is the most disregarded but the main issue, because our judgment is always affected by catathymic feelings.
I agree with Ian et al. that Archetypes, as a methodology, is not enough to enable interoperability by themselves. It is mostly dependent on the content of those Archetypes that will matter most - and the two(or multi)-level modelling approach just facilitates this by providing clinician-friendly but computable representation of health information. Since we don't use a holistic RIM from which all artefacts are derived from there's a huge degree of freedom for creating archetypes (including which data structures to choose from). Therefore we need firm control of content - which requires human agreement.
There are two types of use for archetypes:
-local, ad-hoc in geographical sense and time
-general in geography and time
The first type will serve local arrangements when, for instance, one needs to integrate systems in one specific hospital with those of a specific other one.
The latter type when we need to express data using archetypes in systems that so to speak have to be designed, yet and used by healthcare providers that have to be born, yet and used in an other healthcare domain, an other languages an other culture.
In the present day reality we are content when we can achieve the first type of use in a situation with partial semantic interoperability.
But we have to prepare for the latter type used in a situation with full semantic interoperability.
It is clear that Stefan points ate the first type of use and that Koray refers more to the latter.
In both cases we need human agreements.
In the first type they can be rather loose, flexible, relatively unstable, and ad-hoc. This is typically the situation with archetypes defined in the 'old-days', the days of learning, in many regional or national projects.
In the latter type they have to be strict, well defined, well managed, well maintained, well owned, well quality assured and very stable and used in a large worldwide community
We must start to prepare for this next phase of the use of archetypes as carriers of many human agreements on the topic of full, safe, semantic interoperability.
Hello!
I agree that we are now in the middle of the "local arrangements" mode. Let us record that "local" means >=10 million patients, 300 Hospitals, 15000 resident doctors in some "local" projects on this planet.
In my opinion arrangements for ANY type of interoperability project in medicine "have to be strict, well defined, well managed, well maintained, well owned, well quality assured". Otherwise they will fail. "loose, flexible, unstable, and ad-hoc" solutions will not survive and scale in any setting. Nobody today will happily accept an unsafe solution.
George Mac Ginnis of NHS has provided a very beautiful slide on when it makes sense to use standards in an interoperability project, see http://sph.continuaalliance.org/docs/SPHBFSMacGinnis.pdf,
slide number 9.
He uses the scales of lifetime, size of patient base and number of services. Maybe that can help to better describe the scenarios we are seeing.
What Gerard calls "local" seems to map to what George calls "Requires operational workarounds". Gerards second type might fit to "Flexible and scalable service" in George's slide and definitely is where we are going for.
I agree that we still have some work ahead until we reach the "very stable and used in a large worldwide community" phase supporting "multiple cultures and languages". In that space the timeframe of a few generations as Gerard suggest seems reasonable.
By all means, the situation as is today and that was in the past, are necessary steps to get to the state of Nirvana some day, wirth full semantic interoperability some time in the future.
Until then we see diverging ways to model archetypes, bind to coding systems and ontologies.