Ok,

Back from London, to watch an amazing game in Euro 2008 just in time, to see Turkish national team creating sort of a miracle . I just have to mention this. I can remember screaming and jumping around, probably breaking some of Zeynep’s stuff, and the rest is still fuzzy…

The main reason I was in London was to join the Snomed conference organized by Tim Benson from Abies .The event was quite dense, with some very interesting  presentations. It was held at the same place with Smart Healthcare 2008 Expo which was a different beast on its own. Tim deserves a big “Well done”, for the content and the environment he provided.  Too bad no one else from Turkey was there.

There are some many things to write, so a few parts will probably follow this. First things first: Snomed CT is a very critical piece of work, for it overlaps with many interesting domains. The conference showed this very clearly, from decision support to user interfaces, you need to make use of it. However, the strongest impression that I got from the speeches I’ve followed is: “we are in trouble”. We are not in trouble because we do not have necessary standards or methods, but we are in trouble because we simply have too many alternatives to do a particular thing given these standards.

Snomed CT is powerful, and that power can lead to (and I’m afraid it is already leading to) situations where we are pushing a particular technology  beyond its core purpose. Golden hammer is nice term for the problem, where a lot of problems which should be handled by other tools are simply regarded as nails.  The right level of Snomed use is a very critical question, and how to modify relevant technologies to catch that level makes this a multidimensional problem, which requires coordination among different standards bodies.

One of the presentations outlined the general view of HL7 V3 for integrating Snomed CT. HL7 V3 have always recognized the need for using existing terminologies, and there are some cases where you have the option of assigning the “modelling” responsibility for a particular medical concept to either Snomed or to the standard that is linking to Snomed.  I’m not sure if this is a clear description  of the problem at hand, let me try to explain it a little bit more: You can go towards an approach where you do most of the domain modelling with a standard like HL7 V3 or OpenEHR, and use a terminology only as an identifier, and a classifier for a particular “thing” contained in a wrapper artifact, like an OpenEHR archetype. When the terminology related standard has more expressive power then a simple list of things, there are now two ways to perform modelling for a particular concept; one in the terminology oriented standard, another in the container.

This kind of confusion is dangerous, and can lead to major problems in interoperability which is the primary reason why we have all these standards! There are also other aspects, like the reflection of these choices to tooling. Tooling is an often neglected (or cared less about then it should be) aspect of this domain, and we need tooling to diffuse our work in standards to actual implementations. We can not expect existing managers, doctors and even patients to emrace the learning curve of a tremendous amount of standards. Tooling is necessary to make a very large user base to adopt the work in more abstract layers, and I am not happy about the state of tooling in this particular field.

A limited set of tools are being developed for critical functionality, but when decisions like using “right amount of Snomed” are to be made, these decisions can dramatically alter tooling implementations. I for one, would be very happy to see a well designed solution in tooling to give the ability to adjust use of terminology related solutions in medical domain modelling, and that’s why Ocean Informatics is worth mentioning here.

They were one of the sponsors of the Snomed Conference, and they were as enthusiastic and helpful as usual. Now their solution for use of terminologies in Archetypes, deserves a whole post on its own, which is probably what I’ll do. The thing to say for now is: Ocean has been working hard! The terminology server was very impressive, especially considering the response time to web service calls. Snomed CT is not  a small piece of data, we are talking about over 350.000 concepts here.Well, I’ll write about other aspects in following posts, but as I’ve said in the beginning: Well done Tim!. I should not miss it next year either.