HL7 Watch: Is HL7v3 Overrated by being rated at all?

I eagerly await new posts in Barry Smith's HL7 Watch blog. They're not all that frequent, but when they appear they are bombshells (at least in the world of healthcare ontologies). This week Barry posted two new blog entries reporting on the existing deployments of Oracle's Healthcare Transaction Base - or the lack thereof. Look! The Emperor has no clothes! ;-)

On Tuesday, Barry took apart Oracle's pronouncements about sites where its HL7v3-based Healthcare Transaction Base is deployed. They named three. Barry called the first, Byrraju Foundation, and Barry reports

...I am told that there is no V3 application running in India today and that the Byrraju Foundation is presently not using any telemedicine application that utilizes HL7.

Hmmm. Of course, the implementation could be small enough that it was unnoticeable to his contacts.  If his contacts are not in positions in their organization where they would know about such an implementation, they may have overlooked something huge, but Barry is pretty well-connected worldwide with the kind of people in whose bailiwick this would lie, so I doubt it. (Barry leads the Ontology Dissemination activities of the National Center for Biomedical Ontology and co-directs the National Center for Ontological Research.) This is the only one of the three that is supposed to be live already.

He moves on to Stockholm County in Sweden, where his Swedish colleagues report:

[the first attempt, in 2005]  was abandoned, ...partly because of poor performance (the new application performed significantly less well than the system it was designed to replace, even though it was being run on considerably more expensive hardware), and partly because of a lack of fault tolerance, which made it inadequate as a mechanism for integrating legacy systems marked by a high degree of variation in data quality.

They are now deploying a much smaller pilot that is targeted at handling referrals only. If moving the goal posts leads to victory, we're running at 50% so far, but if we assume the original requirements still stand we are at zero for two so far. Barry couldn't find any information on the third report, an implementation in Louisiana, so we'll assume Oracle is right about that and we're standing at one for three.

My experience with HL7v3 leads me to believe that it has some major problems stemming, oddly enough, from well-intentioned attempts to do things right. HL7 decided that most of the issues with v2.x arose because its design was driven directly from implementation-level requirements and did not capture key aspects of the domain correctly.  They decided to develop a comprehensive information model and base v3 on that model.  The domain is both broad and deep, though, and a comprehensive information model has proved difficult to achieve.

I gotta go for now but hopefully I will find more time to write after the semester is over, in another week or so. HL7v3 deserves a closer look. I'll try to write up my own experiences with it soon.

I wasn't focused on issues related to HL7 back then, but I'm told this effort has been going on since the late 1990's.