openEHR-clinical Digest, Vol 35, Issue 4
thomas.beale at oceaninformatics.com
Wed Mar 11 10:50:43 EDT 2015
there are quite a number of widely known facts and resources ignored by
the paper. See below.
On 11/03/2015 13:16, WILLIAM R4C wrote:
> Hi all,
> As one of the author's of the criticized paper by Blobel et al, I feel
> some need to react and give you some thoughts:
> - OpenEHR after 20 or more years is still largely under construction.
you're the only person I've ever heard say such a thing. As you can see
from this page
online for a decade, the Release 1.0.2 has been stable since 2008.
Hardly 'under construction' (although specific things are of course
always being worked on). The opposite criticism (not enough recent
releases) would make more sense.
> I have asked many times to get names and locations of reference sites
> where I can see a real world archetypes based system in action. No
Quite the contrary: see this page
easily findable from the home page, also visible for a decade in
different forms. Also the well-known locations of various national CKMs,
containing archetypes used in these systems. (Australia
<http://arketyper.no/ckm/>, Moscow CKM offline due to trade war with
Russia..., Brazil being moved as we speak.)
> - the approach with the archetypes is technology driven:
> implementation specific, not clinically driven. It lacks the basic
> conceptual, logical, implementation perspective of ISO 11179. In
> particular the logical modeling is what Blobel et all discuss.
Quite the contrary - the archetype approach is completely clinically
driven, as everyone knows. Unlike every other effort I know of,
including HL7v3 and FHIR, openEHR archetypes are /only /built by
clinical people, in a separated space (CKM). Technical people don't even
come into it, unless they are also docs or health informatics experts
who have clinical / lab / other relevant expertise. I'm not trying to
criticise HL7 or FHIR here, merely pointing out that openEHR clinical
modelling is literally in a dedicated clinical / health informatics space.
Getting /more /clinician input into FHIR's clinical models and more
technical input into openEHR's was one of the motivations for the recent
HL7+openEHR Adverse Reaction joint review (admittedly well after the
Why an academic paper would report the opposite is a mystery. As I said
to Jan Talmon, statements demonstrably contrary to reality don't help
the reputation of the journal at all.
ISO 11179 is a meta-data and registry standard, it has no clinical
content. Whether it has value for standardising meta-data in registries
is another (implementation) matter, unrelated to the design concepts of
> - use and grounding the Modelling in formal ontologies is lacking in
> any of the Modelling approaches: HL7 templates, HL7 FHIR, OpenEHR
> archetypes, 13606 archetypes, CEMLS, CIMI, DCM in UML. The articles
> discussed that with respect to 3 examples. All modelers have a job to
> do. For justification have a look at semantic health net work.
That is more or less true. Semantic Health Net and also years of IHTSDO
activity shows just how difficult it is to get even the most basic
agreements on how to do this. Even BFO, which is a much needed as an up
upper level underpinning ontology has not yet been released as BFO 2.0,
which is sorely needed.
I would say it is widely recognised that the ontology / model
relationship needs major attention.
> - the GCM model allows a much deeper analysis of domain, modeling and
> implementation eg through domains on z axis, business bottom up and
> top down on y axis, and Reference Model – Open Distributed Processing
> (RM-ODP) system development standard on x axis.
Well, speaking as an engineer and software engineer, I would beg to
differ. RM/ODP isn't a formal approach to domain analysis, it's a system
engineering meta-model - a way of formalising different aspects
(viewpoints) of systems. I have actually talked with people involved in
RM/ODP development at ISO (Kerry Raymond
one such person, but also others) and they agreed that RM/ODP is
specifically weak in representing any domain / semantic viewpoint (it
weakly represents it via the 'enterprise' viewpoint). I actually
produced a health informatics-specific version RM/ODP years ago that was
presented at HL7. It was met with mild interest, and subsequently never
used again. Including by me.
The GCM cube is one of those things that looks nice on paper, and noone
can say it's wrong, but it doesn't provide any useful analytical output.
It's not dissimilar from Zachman, FEAF and other similar enterprise
modelling grids. These are designed to help systems engineers not forget
specific aspects of system design. I have been aware of the GCM cube
since about 2003, and have never found a use for it other than a general
explanatory one in academic papers.
If you are going to claim that GCM should be used to help clinical
domain modelling, you have to say how it is going to do this. The paper
doesn't do that.
> OpenEHR, like many others have not a complete picture.
well, that's true at least. We need to have some work to do tomorrow...
> Of course you may critique a paper exposing this lack. But it feels
> like shooting the messenger(s) instead of listening to the message.
> Guys, you've got work to do.
We always have work to do. But apart from the ontology question it's
probably not where you think it is.
Publishing academic papers that ignore well-known available evidence and
projects, and make numerous assertions unsupported by relevant evidence
doesn't help the common cause I'm afraid.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the openEHR-clinical