Alan Rector on SNOMED, HL7 and Quality Assurance

In an excellent keynote lecture at the recent Semantic Mining Conference on SNOMED CT, Alan Rector makes a number of points of direct relevance to HL7 and its continuing problems. As he puts it, 'Unless we can formalise the mutual constraints ... HL7 v3 + SNOMED = Chaos'. 'The documentation is beyond human capacity ... to write or to understand'.

Rector puts forward a series of proposals concerning Quality Assurance for the new SNOMED Standards Development Organization, from which the HL7 organization, too, could benefit immensely. What is needed is:

– Member of board for Quality Assurance - sole responsibility
– Senior member of operations unit for QA- sole responsibility
• Changes in process
– Fully open and honest QA process
– Public statement on problems and public health warning until fixed
– Public commitment to specific quality criteria and success in specific applications.
– Engagement with providers, vendors on key criteria for QA
• Major technical effort
– Set up an independent quality assessment unit
• The developers can’t lead it because they can’t see it!
– Invest in crash effort at improvement of quality
• Initial goal of say 25K guaranteed QAed & reliability assessed codes

Quality, as Rector points out, 'is best addressed by openness.' One might add (drawing on the experiences in England): HL7 v3 should refrain from promoting itself as an ISO standard until its own QA problems have been fixed.

AMIA Panel on the Future of HL7

This panel will take place at 3.30-5 p.m. on Tuesday November 14, 2006 as part of the Annual Symposium of the American Medical Informatics Association in Washington DC. Panelists are:

Barry Smith (organizer)
Werner Ceusters
Christopher G. Chute
Charlie Mead
Don E. Detmer (moderator)

Further information is available at: http://ontology.buffalo.edu/HL7/panel.htm
Audiotape of the presentations is here: http://ontology.buffalo.edu/HL7/AMIA_06_audio
Slides for Smith's presentation are here; those for Ceusters' presentation here.
Aftermath 1
The presentations from the four panel members generated a (sometimes heated) discussion. Several speakers pointed to existing implementations of the RIM (contradicting one claim in Smith's presentation, to the effect that there are no implementations of the RIM). In discussions subsequent to the public part of the meeting, however, it proved that in several of the mentioned cases it was not the RIM itself which had been implemented, but rather some version of the RIM which had been fixed up to make it implementable.
Aftermath 2
It is interesting that both of our interlocutors accepted that there are problems moving from design to implementation of HL7 V3. Yet still they see no problems in the fact that HL7 V3 has been declared an ISO standard.
Aftermath 3
Both of our interlocutors insisted that they want critics to join the HL7 movement, rather then throwing potshots at the RIM from the outside; otherwise, they say, HL7 will not pay attention to what we have to say. But some within HL7, including our friend Gunther Schadow, are already paying attention. And surely this is right: if problems are found in an endeavor of such importance for the future of healthcare and of healthcare information technology, then it should be incidental who points out these problems. What is important is that they are fixed.
Aftermath 4
We brought arguments to the effect that important aspects of HL7 RIM can be supported by no known reasoning systems; other aspects are incompatible with the reasoning systems underlying important terminology initiatives such as SNOMED CT and NCIT. In response, our interlocutors insisted that HL7 RIM it is just a messaging standard; thus it does not matter if it does not support reasoning. But does this jibe well with other claims being made on behalf of the RIM, including claims made at this very panel?