Contents:
- Introduction
- Architecture
- The Meeting
- The Presentations
- The General Data Framework
- Steps Towards a Model of Models
Introduction
The progress and outcomes of the work of the ISO/TC215 health records committee have not heartened some followers of the process. The first meeting of the committee was in Orlando, Florida, USA. The goals set then - interoperability of systems and reduction in redundancies - may have been continued outside ISO, but have certainly not been achieved by the ISO meetings themselves. Instead, we have witnessed the expansion chiefly of the work of the US-based HL7 organisation, and the absorption by this group of many of the functions and arguments of the corresponding CEN251 committee. In Vancouver we learned of the alignment of the Object Management Group (OMG) and the DICOM imaging standard with HL7 Reference Information Model.
We have seen the messaging Work Group Two become effectively an additional forum for the messaging branch of the HL7 organisation. Work Group Three appears to have been simply absorbed into the 'Snomed' classification system under the Chairmanship of Professor Schute from the Mayo Clinic. Bernd Bloebel, the designated ISO/TC215 WG4 leader on 'Access' matters, is also an HL7 committee person, and has worked with CEN. The closing plenary session of the Vancouver meeting had a farcical quality, which was scarcely moderated by the selection of a group to discuss a 'business plan' for the ISO/TC215 committee.
Work Group One is devoted to Health Record Modelling and Co-ordination. It has become a small group mainly composed of people who already have a firm commitment to a domain model, mostly CEN, GEHR, or HL7 RIM. The work of ISO/TC215 WG1 has been problematic, and subject to internal attack. Of the ten or so work items originally identified by the group, only three have been explored at all, the Country Identifier Standards, The Access work item, and the General Domain Model.
The Country Identifier Standards work item was adopted by WG1 at the Vancouver meeting in June 2000, but the area is also well covered by international standards, including ISO ones. Part of my purpose in attending the second Vancouver meeting on GDM was to try and find out what the New Zealand delegation should be doing about the Access work item, which did not receive the hoped-for collaboration from WG4, who have now recommended its cancellation. I had some useful discussions in this area. However, the achievement of a General Domain Model appears to be the sine qua non for WG1, and indeed for the work of the ISO committee, which might otherwise be considered to have become an 'outreach' committee of HL7.
The Australian General Domain Model (GDM) was withdrawn and put back to stage 00 by the Australian delegation at the June 2000 meeting in Vancouver, and many delegates despaired of an ISO GDM being achievable at all. The initiative for the GDM meeting came from Ken Toyoda, the Japanese ISO/TC215 delegation head, who also suggested that it be in Vancouver. The Canadian and Australian groups suggested the approximate timing of the meeting, although as it turned out no representative from Australia actually attended. The meeting had seven delegates from Canada, three from Japan, and one from New Zealand, with Canadian Don Newsham in the chair.
Architecture
A major problem for WG1 had been a conflict over the use of the term 'architecture'. The US delegation head specifically disavowed any concept of 'record architecture' in the Sydney and Berlin meetings. This was an HL7 RIM interdiction, emphatically relayed by the US Head of Delegation. However, the concept of 'record architecture' was so important to both CEN and GEHR that there appeared to be nothing more to say.
The meeting notification from Canada had attached an 'exploratory briefing note', which reintroduced the concept of architecture. The paper introduced itself as a 'terminological cheat sheet' for 'these exploratory discussions on the development of a General Data Framework.' The paper made a distinction between 'Information Models' and 'Frameworks'. The author said that:
"Information models are mostly concerned with describing what is wanted or what you have, often in a visual manner. Frameworks are more commonly used to describe and structure enterprise architectures or other comprehensive domains."The author elaborated further on the concept of 'Framework', and introduced the work of Zachman ( http://www.zifa.com/frmwork2.htm ), suggesting that probably the best known Framework was Zachmans' Information Systems Architecture Framework.
Zachman presents a framework where different kinds of model can be located for an enterprise. Could this be the elusive 'models of models', which is a concept that had come up several times in earlier WG1 meetings. It consists of row and columns, with the categories 'who, what, how, why, where and when' heading the columns, and levels of specificity and perspective making up the rows.
Use of the Zachman framework for this purpose would not constitute a Record Architecture, but perhaps rather a place where record architectures could be located in a larger scheme of things. However, the enterprise concerned needs to be defined, and for the ISO/TC215 committee, this would be 'Global Healthcare'. The briefing paper suggested setting up the Zachman Framework as a 'straw man' model of models for global healthcare.
The Meeting
Vancouver was cold and rainy, a different scenario to the June conference. The small meeting room was completely informal; we wrote our names on blank cards in front of us, and because it was a small group, we were quickly able to establish a working rapport.
The Japanese delegation consisted of Yasuyuki Hirose from Okinawa, Tetsuro Kiyotani, and Kanji Nakai from Hitachi Corp. Prof. Hirose was on a late flight, and in the introductions, the other two Japanese delegates were careful to emphasise that they were contributing as individuals, not representing an official 'Japanese view'.
On the Canadian side, there was a mix of skills, with Jane Curry working with Alberta Province Healthcare in Edmonton, and also chairing the international section of HL7. Ron Parker presented the Canadian Data Model and is also part of the 'Siemens' company, which now owns SMS and has major investment in EHR development at this time. Mark Fuller works on health indicator development with the Canadian government, and Don Newsham, after many years of health administration, now works as a consultant. Laura Sato, Nishma Kanji, and Sylvia Kenyon work with the Canadian Institute for Health Information (CIHI).
The meeting consisted of the informal presentations on data modelling activities in New Zealand, Japan, and Canada, and there was an update on 'state of play' in HL7 from Jane Curry. Then we launched at the Zachman framework, to see if we could make it work as a General Domain Model, for all the different directions we were coming from. We may have succeeded, but only time will tell.
The Presentations
Kanji Naku from Japan presented a diagrammatic representation of 'Modelling Co-ordination in the Healthcare Domain'. In it, The proposed ISO GDM is given a very central role, dictating conformance and somehow containing both the HL7 RIM and DICOM models. The 'Dicom' imaging standard is the well established imaging metadata format, which requires the inclusion of a standard set of descriptors with any conformant image.
He also outlined plans to develop a Reference Information Viewpoint Model for Japanese healthcare. This would comprise organisation, roles, responsibilities, and the technical view. This would take the form of a 'virtual hospital' and would made in terms of 'RM-ODP'. This presentation led to some discussion when Ron Parker drew a diagram on one of the many hard-copy paper 'white boards' provided of his conception of the role of GDM in interoperability between jurisdictions. He appeared to see the GDM as a mediator between information systems of each jurisdiction, which would somehow translate between them. It was established that the Japanese view was of a central role for the ISO GDM, and that this was shared.
Next, I presented material from the New Zealand Health Information Service (NZHIS). I listed the present data capture modalities, and introduced the 'Health Intranet' (HI), the new communication environment developed by the NZHIS. I emphasised the Health Event Summary (HES) concept, which is developed in the Australian Health Information Network Australia Report (HINA). I mentioned some plans to develop a national 'access system' along with the HES, using the authentication and encryption resources of the HI, and the National Certifying Authority. I suggested that the 'General Domain Model' might just consist just of lots of HESs and their behaviour.
R. Parker provided an update on the Canadian Conceptual Health Data Model (CHDM). The main components of this model include governance, environment, people, resources, and events. This model is meant to represent fundamental concepts in health and their high-level interrelationships. 'Events' were described as the mechanism by which instances of those entities are created, modified, and removed. Overall, the CHDM is intended to be comprehensive and jurisdiction-neutral. It was described as a 'data capture model'.
Jane Curry then gave a presentation on current developments in the HL7 Reference Information Model (RIM). It had been stated that the RIM was initially too US-focused and insufficiently generic to represent international interests. She mentioned the presentation at the Dunedin Access meeting by Maori healthcare workers, and how the question of cultural difference had been corroborated by Japanese and Korean representatives. Thus, a task force was created to make recommendations on modifying the RIM to be more flexible. She outlined changes to the HL7-RIM, and these included the removal of any reference to billing and scheduling from the RIM, because these contained elements specific to US culture. Version 3 of this model will be released in January 2001. She also confirmed that the United Services Action Model (USAM) was now in use as the way the RIM deals with clinical concepts.
She reckoned that 'a view matches a jurisdiction, which is also the boundary around your perspective.' She felt we were 'charged to come up with something that allows harmonisation across perspectives', and gave as examples the Maori perspective and other non-Western views, including those of North American 'First Nations'. Korean and Chinese traditional medicine, and Ayurvedic medicine are further examples, but the list of non-Western medical practices is very long. We were 'Going to have to figure out how to meet the objective of being sensitive to these perspectives.' She said that the Canadian data model was a Conceptual Health Data Model that enables us to know things we care about, such as people, environment, location, governance. This last includes policy, rules, relationships, and contractual relationships - information that constrains and defines. It was a 'data capture model' affected by events at all levels. Events were the 'mechanism by which instances of those entities are created, and modified.' She said there was a new corporation to develop this, which had money to invest.
The General Data Framework
I would like to set down some fragments from the 'brainstorming' discussion, which then followed.
Jane Curry set the tone of the discussion when she suggested:
'All models are wrong, some are useful.'With that brief statement on the metaphysics of modelling, we were able to begin to explore the needs and value of a general data framework.
Don Newsham asked the group to consider what is needed internationally that is not satisfied. We need to facilitate consistency, compatibility, comprehensiveness of health data collection and support a common understanding of language. We need to improve the quality of data and information to support other ISO work items. We should provide a reference point of individual country models and enable coordination between them. We need a tool for navigation, integration and alignment within and between countries. We need this:
- For cooperative analysis
- For provision of care
- For translation of health information
- For common data access across boundaries
- For cost effective data exchange
- For vendor info systems
- To support patient comprehension of health data
- To support health information and products across boundaries.
Ron Parker suggested that a key part is defining what are the things we need to measure, what are the vocabularies; can they be mapped? Information might be considered 'data in context', but what are the assumptions?
Yasu Hirose pointed out some of the difficulties. For example, some services that might be called 'Medical' in Japan, might be called 'Welfare' in Canada. Yasu took this further, asking 'Why do we want compare?'
Tetsu replied that it is necessary to compare for research, and also it is 'nice to have a translation device.' Kanji reiterated that we need a framework of models.
Ron Parker said that people were mixing anyhow, and there was a need to exchange information between different vendor systems, e.g. HBOC, Cerner, SMS, Oasis. Vendors would be motivated to build their systems to support interoperability. Whereas traditionally a vendor such as SMS had been trying to find people internationally to shoehorn into their systems, that would change now that it had been taken over by Siemens. He pointed out that there was a demand now for interoperability standards, e.g. from the nursing community, and that we appeared to be developing a 'tool set'.
I had proposed the 'Health Event Summary' concept as a fundamental building block of the 'General Domain Model', and the Canadian Data Model also emphasised 'events' and 'data capture' as the fundamental process which transforms the other 'entities we care about'. Could not then a concept of the 'encounter' as encapsulated in 'events' become a standard record building block, reflected in the GDM? These parse or segment reality into 'take units', or defined periods of time in which data capture occurs. The transition between the Assessment and the Plan in those entities is the motor that drives the entire Enterprise.
Jane Curry agreed. She asked 'where is the edge of the health system?' The 'encounter' touches the real world on the ground, and can be extended beyond the clinical encounter to all the kinds of events that support healthcare, including credentialling encounters.
Yasu suggested that 'how the brain works' should be a determinant of the General Domain Model. I agreed, and said that the transition between the 'Assessment' and the 'Plan' in Laurie Weed's "SOAP" model was a physiological reality. Between the 'assessment' and the 'plan', between the 'ideo' and the 'motor', falls a silence 'as between two waves of the sea'. This is indeed 'how the brain works'. As Mark Fuller noted, the conversation at this point touched on the metaphysical.
Steps Towards a Model of Models
By Russell's paradox, a model of models should be impossible. It should be impossible for a set to contain itself, by definition, and certainly not one containing such diverse practices and behaviours as the Global Healthcare Domain. However, the Zachman framework is not 'the containing set' from which all other healthcare frameworks can be derived. Instead, it appears to be a method, a way of approaching phenomena that can be applied at any level and on any scale. In this sense it is 'fractal', i.e. the same structure repeats itself at all levels of granularity. Why should a concept so general be useful?
I think the answer is the same for healthcare as for its utility in other arenas. It is designed for humbler enterprises than modelling global healthcare. It can model the assembly of aeroplanes, or any other complex engineering project. It is designed to do the job of modelling a 'domain', but does not specify the domain. As applied to healthcare, it might constitute the infrastructure for the 'global plan' that Peter Tresseder mentioned in his opening address to ISO/TC215.
It is harder to comment on the slotting in of entities within the diagram itself, and it is hoped that a future version of the minutes might include a guide for how to decide where a healthcare element fits. At the time, the slots in the framework were filled in by a kind of magic like spontaneous combustion. I am excited by the concept of the Zachman Framework. I would see the US RIM as an 'ontology', the HL7 Personal Record Architecture (PRA) as an implementation template, each locatable on the grid, along with other models like the CEN extended architecture, and the GEHR archetypes. I want to know where 'event summaries' fit. Perhaps, like exotic lepidoptera, we can pin out the characteristics of all these cultural products so as to render them interoperable, in time.
At the end of the meeting, we were each asked to say what we thought should happen in ISO/TC215 WG1. This was an opportunity for a personal 'wish list'. I pointed out that effectively TC215 had become an outreach for HL7, a forum where people could contribute to HL7 discretely. I thought that the GDM proposal was useful, a place to put everything and thus show up redundancies and overlap in modelling activities. I liked the fact that it was 'fractal' and could interact with its environment on so many levels at once.
I said that it might not be appropriate to engage third-world countries in the development of an ISO healthcare standard, since about half of the world had no healthcare to speak of anyway. I suggested that it was the developed nations who had the sheer mobility, networking, and finance to evolve it. I suggested that if our constituency included that other half of the world's population, it might be better to develop something that they could customise for their own cultural and economic environments and then make it available. Perhaps through authenticated channels, there would follow verified decision support. This might further help fulfil Peter Tresseder's agenda of a 'plan for global healthcare'.
The GDM model may stimulate debate and allow at last some progress on the role of ISO/TC215, and the development of an international healthcare communications standard. It was a privilege to be part of the group that has initiated this work.
Mike Mair
New Zealand delegate
20 December 2000