Difference between RIM by HL7 v3 and CIM by openEHR

This week I finished the HL7 Fundamentals Course, where v2, v3 and the corresponding Reference Information Model (RIM), from which the Clinical Document Architecture draws its specification and shortly FHIR.

From this course, I learned that some people behind HL7 v3 and thus RIM, were the same people, who also worked on openEHR in the beginning. In addition, several naming concepts such as Participation, Template and so forth can be found in the standard HL7 v3 and the open specification openEHR. Finally, the naming of RIM and Common Information Model (CIM), suggests some deeper similarities.

Consequently, I wanted to ask, in this community, how distinct the RIM and CIM actually are? And where they intersect?

1 Like

Ping @thomas.beale - who may be the only one around way back then?

Thanks, Heather, for guiding me. I ll ping him, but I am not sure what ping means at this discourse channel:

  • Can we ping participants at this discourse Channel by simply adding @heather.leslie for example?
  • Or do you mean I should email him instead?

The HL7v3 RIM was created by Woody Beeler, Gunther Schadow and other key HL7 personalities at the time (just pre 2000). I went to HL7 meetings as an Australian government standards representative during 2000 - 2006 and later as part of the UK gov team.

I remember doing a presentation on archetypes in about 2002. The RIM builders did not want to hear it. I wrote a 25 page technical analysis of HL7v3 in 2004 or so, which showed why it was unlikely to be workable (major modelling problems; etc.) I provided this to the AU and UK government efforts at the time. Neither group could find any technical error, but nevertheless, both proceeded on the v3 pathway, to later regret (particularly in the UK).

openEHR was started by two groups who had worked previously on the GEHR project - the UCL Centre for Health Informatics and Multi-professional Education (CHIME) department led by Prof David Ingram, and Dr Sam Heard and myself from (at that time) Australia. Sam and Dr Dipak Kalra and others in the UK formed the original group who did the earliest architectural conception of openEHR.

We tried to get some collaboration with HL7 back in those days, but unfortunately without success. The adoption of HL7v3 ended up costing the UK, US, Australia and others many billions. It was declared by Grahame Grieve (lead of FHIR) in 2013 to have failed (the blog post is linked here, although it appears the original blog is now unavailable).

Participation was an HL7 thing (it’s part of a conceptually sound pattern of participation of actors / entities in acts, but HL7v3 tried to use it for everything). In openEHR we reused the concept, although today’s version is mainly influenced by Basic Formal Ontology (BFO). ‘Template’ was an early idea among some people attending HL7 including Angelo Rossi Mori (Italy) . There was a Templates SIG and then a WG for some years. This group tried hard to get HL7 to adopt elements of the openEHR domain modelling approach (i.e. archetypes) but in the end was unable to convince others.

I’ve never heard of any ‘CIM’.

Some references:

5 Likes

I do know of CIM meaning Clinical Information Model (conceptual similar to archetypes) as defined in iso13972:2022. They used to be called Detailed Clinical Models in a preceding iso norm. (IIRC)
They are mostly used in the Netherlands to model zibs.
It’s mostly data type definitions (similar to openEHR data types) and modelling strategies (policy), but not information classes(like openEHR RM EHR model), no formal modelling language (other than UML) like ADL and no persistency semantics (contribution etc) and no implementation specifics like rest-api specs.

And there is CIMI a stranded (?) initiative to do modelling jointly.

2 Likes

Thanks a lot to you two, for answering my question. As well as for your patience.

@thomas.beale Thanks a lot for your concise and broad explanation on what happened at these times. Yes, I agree with you, HL7 v3 was not well received, but nevertheless, the idea was quite smart - but this does not cut it. Your answer also allowed me to understand the similarities and differences between HL7 v3 and openEHR, and it was also very interesting to read, what actually happened then. And also thanks for the references, I hope I ll find the time to read them.

@thomas.beale and @joostholslag : In this thread CIM means common information model and is stated as the title of the following specification page for RM Release-1.0.2 (Common Information Model), why I considered to ask what the commonalities between RIM and CIM are.

3 Likes

Got it - we do have a spec called ‘Common’, which is just one of various such specs making up the Reference Model. All those specs have a title of the form ‘xxxx Information Model’. The Common IM (as we usually refer to it) is not the ‘main information model’ as its title might imply, but rather a spec for certain common elements used in the other specs (ironically including the participation pattern, which is the basis of the HL7v3 RIM).

2 Likes