HL7 Standards Organisation meeting

Salt Lake City, 1-5 October 2001

Report by Michael Mair

Contents:


back to contents page

Introduction

The March 2001 ISOTC/215 meeting in Seoul Korea was dominated by the reality of HL7 takeover of the global Healthcare Standards arena. However, the US-derived standard is itself incomplete. The urgency perceived by many nations in getting a rudimentary National EHR implementation may force them to adopt an eclectic approach, in the short to medium term. The ISO committee itself is in danger of being trivialised into a media exercise by participation of consumer groups.

top of page

Introduction

The October 2001 meeting of the US-based Health Level Seven standards organisation was held in Salt Lake City. It was held three months after the successful international affiliates meeting in July 2001 in London, UK, and immediately after the results were known for the first HL7 Version 3 ballot. The meeting also took place just 20 days after the tragic attacks on New York's World Trade Centre Twin Towers, and on the Pentagon. There was heightened tension on the flights and in the airports, and there was the certainty of imminent war. There was also increasing uncertainty about the fate of Air New Zealand during the trip, so the small New Zealand contingent consisting of David Fallas (of Sysmex Delphic Medical Systems Ltd) and myself also did not know if we would get back safely.

I was staying in cheap accommodation in the South end of the city, where the circumstances were humble but the company colourful and diverse. There was a two-mile walk into the conference venue at the Sheraton City Hotel. I do not know how many time warps I crossed on that brief journey, but once inside the conference venue, the group there was possessed of a strange intensity and fervor.

I was privileged by a number of contacts at the conference. First, David Fallas shared the experience and could fill me in on technical matters to do with the history and present practice in laboratory IT systems. His perspective goes back to the first implementations of HL7 in New Zealand, which was then and remains the only implementation worldwide of a National Health Unique Identifier, using HL7 coding. We worked together on the advice at the conclusion of this report.

Second, as 'acting chair' of HL7 New Zealand, I got entry to the Technical Steering Committee and HL7 board meetings, both of which were dinners, which also helped my budget. I got to hear the internal debates on important questions such as the relationship with the US Government, the true progress of the Version 3 standard, and the question of its name. I got to assess the nature of the democracy in this organisation, and got a feel for its dynamic.

Thirdly, Klaus Veil Chair of HL7 Australia was a most valuable resource. Klaus had facilitated the acceptance of my 'acting chair' status, which carried a subsidy on the attendance fees as well as the dinners and associated gossip. Without these my attendance would not have had these privileges, and might not have been possible at all. Klaus is a major power in the HL7 organisation. He also chairs the International section of HL7, and recently drove an HL7 'roadshow' in Eastern Europe, which recruited several countries, such as Lithuania, Poland, Czechoslovakia, and perhaps Russia as well, into the HL7 fold. He, with Liora Aschuler (see below under CDA) were two of the three recipients of a special 'Ed Hammond' award, presented by Ed himself, for Volunteer of the Year.

Finally, I had the opportunity to talk with Liora Alschuler, who is co-chair of the Structured Documents group of HL7, the proprietors of the CDA, as well as marketing manager for HL7. I ended up the week with a clear identification of the immediate way forward for New Zealand on the matter of electronic health record standards. Changed global circumstances also change some assumptions about the context for our local drama.

top of page

The Opening Pleneary Session

Gartner's Vision for Healthcare: The Next Ten Years

This was the keynote address delivered by Dave Garets, Group Vice President of Gartner Inc, which is an advisory group for healthcare planners. It was an objective look at global healthcare, and asked the question: 'What will the delivery and financing of healthcare be like in 2010?'

He took account of two sorts of drivers, 'pre-determined' elements, such as an aging population, high numbers of uninsured in the population (45 million in the US), uncontrolled medical costs, globalisation of disease, more legislative oversight on healthcare delivery, and the evolving EHR and Internet/e-Health initiatives. Uncertain drivers included the privacy arguments, technology for structured data standards and security, the uncertain state of the world, and the environmental bio-threat.

His team had considered more that a dozen axes by which the outcome of all this might be projected to the year 2010 but selected just two: 'Accountability for Payment' and 'Standards and Structured Data'. These translated into the polar extremes of Single Purchaser (eg Government) vs Multiple Purchasers (eg individuals) for the 'accountability for payment' axis, and 'Blurry islands of data' vs 'Robust data sharing' for the 'Standards and Structure' data axis. From these, he constructed a simple grid of possible outcomes.


Healthcare 2010 Matrix

Vertical axis: Robust Data Sharing vs Blurry Islands of data

A: Rational Rationing
You get what you want and Big Brother will pay		 B: Free Market
You can get what you can pay for
C: Irrational Rationing
Brute force rationing		 D: Wild Wild West
Healthcare by advertising and promotions

Horizontal axis: Single Purchaser vs Multiple purchasers


It is easy to identify that we are in on this grid. With our public and private sectors, we have competition between models of single purchaser and multiple purchasers, and there is often ideological unease between the two polar extremes. For the Structured Data Standards argument, we are trying to move to robust data sharing, and this is what the Standards debate is about. Having the options set out in diagrammatic form takes some of the ideological heat out of the arguments, since we can coolly consider them as options with implications rather than causes.

It was of interest that the Gartner team was seriously considering that the US might be forced into a 'single payer' option for healthcare delivery. This would depend, they pointed out, on the employment market. Where jobs were at a premium, in a recession, the employers would be tempted to ditch welfare support, which could cost '30% of the cost of manufacturing a new car' (from General Motors).

For New Zealand, Garet's comments under 'irrational rationing' might be recognised as representing the present reality, although with the prioritisation initiative, New Zealand is trying to introduce some method into it. As Garet put it:

'Because the single purchaser has such extensive influence, it is able to dictate many aspects of the provision of healthcare. However, because there is no way to effectively aggregate and analyse data on healthcare, the decisions made by the Government are arbitrary and driven more by political factors such as lobbying and influence groups than by sound scientific analysis.'

The US would appear at present to be in Quadrant D. To migrate to quadrants A from C or B from D will require data standards, and this realisation formed an important context for the discussions in the days ahead, for participants of all ideologies and persuasions.

Stan Huff's Plenary Talk

The outgoing Chair of HL7 was Stan Huff of Intermountain Healthcare in Salt Lake City. This sophisticated organisation has a considerable empire in the Utah area, and highly developed electronic medicine. Nevertheless it was interesting to note what they could not do. In particular, they could not exchange health records with other US-based healthcare organisations.

In his address, which followed immediately on the Gartner's vision for the next ten years, Stan sought an analogy between standards development, and the development of musical notation. In particular, he mentioned Johan Sebastian Bach, whose work can be accessed today because of a standard musical notation. The existence of the notation did not stifle creativity, but rather liberated it.

The ready riposte to that is that music is empty of content, whereas the HL7 standard defines content absolutely with the rigid structure of the RIM (see RIM section below). However, the sentiment was there and expressed, that a 'standard' should in no way constrain or pre-determine what can be communicated.

Stan also commented on the proliferation of XML health record markups, and told us to be beware of the 'XML barbarians' who promise interoperability for healthcare applications using XML, but who are not connected to other standards development processes. The point was well made that adhering to such activities would not solve anything for interoperability long term.

Stan was asked from the floor whether he thought that the electronic health record would just amount to a repository of HL7 messages, and he replied that that it was very close, but that you still needed a standard for the names of problems, signs, and symptoms. Also needed was a way of finding if someone has records even, and standards would be needed for that kind of infrastructure. He said there were sociological problems around understanding access, and that anyone can get in to a record system if they knew an insider.

He proved this with a journey through his daughter's health records. She had recently suffered a complication from surgery, and he took us through the charts he had accessed through his special privileges as a Intermountain Healthcare administrator. The demonstration was powerful, firstly because of the power of the data that was available to him, and the sophistication of its presentation. He was able to see graphed lab results, and investigation reports. It was also amazing how open the system was to him, and how easily spoofed were the access control mechanisms. Our HES concept would not have permitted such access without digital signature and authorisation of the role relationship giving 'need to know', which is evidently not in place in Intermountain Healthcare.

Stan also of said that he thought that Natural Language would be the Principle medium for understandability of the EHR until 2007-8. This emphasis on Natural Language echoes comments in a paper from the early days of HL7:

"Pre-defined entry sequences do not take non-standard medical situations into account, yet that is where much of the clinical interest lies. It is in the nature of very sick patients to be sick in unique ways, and it is here that professional judgment and patient Specific intervention is most needed. [Lincoln et al 1993b]

Stan also gave an assurance that Version 2.4 and its analogues were still going to be supported by HL7 for as long as it was needed. However, he pointed out its many deficiencies, such as that no consensus was ever proven for the methods, that there was no formal model, that there was a lack of consistency, that there was no connection to standard terminologies, and that there was no migration path when technology changes. He said there was no need to change technology interfaces if your present applications were meeting the need, but that Version 3 would offer fine-grained interoperability. Version 3 was for the leading/bleeding edge applications, but it was not competing with Version 2*.

Stan Huff's most dramatic contribution from our viewpoint was his demonstration of his clinical concepts, which he had developed with the 3M corporation. He displayed these at the 'decision support' working group meeting organised as part of the UK led EHR 'special interest group'. These entities may be critically important for the evolving global EHR standard.

top of page

The UK Electronic Health Record Special Interest Group Meetings (EHR SIG)

The UK-led meetings were interesting for the completely different atmosphere they engendered from others. In a sense, they represented an attempt to take over the design of the CDA body from the Structured Data group. Clearly there is going to be a development in the way structured data is handled, and the British were jumping in with both feet.

Picture of the Grail from Afar

Announcing the British initiative at the Technical Steering Committee dinner, David Markwell pointed out that since defecting from CEN two years ago, they had been attending the HL7 meetings as observers. Since nothing much had happened that was interesting from their point of view, they had realised that they were going to have to do some work themselves.

He explained that the UK Government had mandated a design for a GP-to-GP facilitated channel, and there had been vendor agreement that the level of structural processing in the model is higher than the CDA.

He said that the level three CDA was not nearly ready, and that the methodology that the EHR SIG would be working on would include interfaces for clinical guidelines and decision support.

He said that they would end up with a design for the level 3 or 4 CDA. He was suggesting that HL7 recognise the Electronic Health Record as a legitimate standards project, and was laying claim to the territory with the Special Interest Group. Otherwise, he said, it would have to be a Working Group that developed the EHR standard, and that was a far longer bureaucratic process, and so he was sure that people would agree to go for the shorter route.

Later, at the first meeting of the EHR SIG, there was unanimous agreement from all present in favour of this development. I said his project was our project, and that we would help in every way we could. German delegates Joachim Dudeck and Bernd Bloebel said similar supportive things, which was remarkable because their own HL7-based exchange format, the Sciphox project, is well advanced. Gary Dickinson from the US was in complete accord, and said that that was where he would want to place his security developments. Bernd Bloebel agreed, which is important because, with Gary Dickinson, he co-chairs the ISOTC/215 access committee. Bernd is active as well in a major European initiative on these matters. We all pledged our allegiance, and our addresses were taken, and we were to be kept abreast of breaking news.

At the decision support presentation given by the same team, I was less sure about the UK-led project, because of the way they saw guideline prompts fitting into a generic EHR structure. They wanted to take the prescriptive standard down to the grain of each decision, where there was a choice between options. I am not comfortable with a generic interface for clinical guidelines entering into the intimacy of each decision in the clinical stream. This is because such a tightly structured clinical interface might impair the multidimensional modelling of the trajectories of natural events that healthcare providers need to work with.

At a later meeting on clinical guidelines, there was real uncertainty in their model, which is perhaps OK because it is such a hard area. Their work did show that they are seriously engaged. A representative of 'Prodigy' gave an excellent talk. This is a well-known decision support engine. At present I find it hard to consider how generic decision support could ever be non-intrusive.

Stan Huff addressed this meeting. He introduced clinical objects, which he agreed were clearly modeled on the GEHR 'archetype', or clinical template (the Good Electronic Health Record project). He demonstrated a number of such clinical entities, illustrated with parts of the neurological exam. He said he had already worked up 15-25 thousand of these entities, supported by the 3M company, and they were prepared to donate these to HL7.

There followed a most entertaining verbal dual with Leo Fogarty, who is chair of HL7 UK and an important UK EHR luminary. Stan insisted that his new entities were not 'templates'. Fogarty insisted that they were templates, but - deferring to Stan - called them 'the T word'. Ian Purves, of the UK delegation and co-author of the UK GP-to-GP project, concluded this discussion by suggestion that a 'stake had been set in the ground' with the acceptance by the group of the importance of this work.

Afterwards, I congratulated Stan on this first set of really useful clinical objects. I suggested that a global dictionary be made of them. I would have said 'data dictionary', but did not feel on firm ground with that term in this context. It is interesting that we appear to need them as well as the RIM, and in addition to coding systems such SNOMED and LOINC. Hopefully they will develop into an internationally agreed data dictionary for EHR and we would be well advised to contact Stan Huff to get hold of these entities to use in our HES network.

top of page

The Clinical Data Architecture (CDA)

Is there a technique or deliverable already out there that could fulfill 80% of our requirements, with 20% of the effort of developing it ourselves, and which would nevertheless assure us of global interoperability? The answer might well be yes - the Clinical Data Architecture (CDA).

It appears that the CDA is the only actual useable product from Version 3, but that, as a Gartner publication portrays it, makes it a least moving target for implementers at this time. Potentially, it could be the standard-sized box which, like the original Honda motor car, could come in any colour as long as it is yellow with a black stripe.

The evolutionary architecture of this approach has its geneses in the CAIT paper of around 1993 [Lincoln et al 1993b]. Written at the very beginning of web technology (it is said to be one of the first Internet documents), it builds from the use of 'mark up' to identify and wrap documents for retrieval and processing in compliant systems.

"The vast majority of information consists of text, which is well handled in word processing or memo format, but cannot be fully captured when compressed into data records consisting of limited and pre-defined fields. The pivotal advantage of document processing is its ability to bring computer technologies into the service of human needs under human direction without bending the work-flow and information management to machine requirements."

This emphasis on document management rather than structured data was developed first as the Kona architecture via a phase called the PRA (Person Record Architecture). The CDA is a more global concept because these 'communication artifacts', to use Bob Dolin's expression, can contain anything from images, to structured documents. The CDA level one, which is the only completed and balloted specification so far, specifies only the 'header' of the CDA standard document, but also declares that the attributes in it be derived from the HL7 Reference Information Model.

The level one CDA body is divided into compartments with names such as lists sections and paragraphs. The one they always use for illustration is similar the simple SOAP paradigm familiar to most of us from the teaching of Larry Weed used in medicine [Medical Record that Guide and Teach. Weed L, British Medical Journal, 1969] The question is, 'how do you do structured data in a universal and machine processable way using CDA templates?' The argument is all about the structure of the body of the CDA. The header structure is already agreed, balloted, set in stone (or so I understand).

I encountered Bob Dolin of the Kaiser Permanante Healthcare on the way to one of the dinners, and asked him how was it that the CDA message concept appeared to be such a success, while people found the rest of them so hard (RIM, R MIMS and HMDs, etc) Bob is one of the architects of the CDA. He said that more and more people were saying that, and maybe the methodology did not have to be so 'in your face' all the time. The reason the CDA was ahead was just because some people had worked on it more than the other groups had.

At a later meeting he that said it would need all of a level three development for the CDA to handle structured data entry with the degree of grain necessary to construct, for example, a National Diabetes Registry. One gathered that this was work in hand. It was here that our input was earnestly sought. One may note that the HL7.org website is also open to suggestions on the CDA.

At the discussion at the Structured Document Group Committee meeting to which Liora Alschuler subsequently invited me, it became evident that this was still at the design phase. We could be immediately using level one CDA documents and make our own templates in their bodies with embedded data objects of our own construction. If we fulfilled our requirements with this technique (and it emphasised that we would be the first nation to do it), this could be offered as a design for the CDA Version 3-4. The idea was that we would have input into the US design as a parallel project to the UK initiative, and that by simply following our own interest, we would end up with exactly what they need, or at least a serious candidate for it.

I discussed this with one of the UK design team over lunch. He appreciated that it was a viewpoint, but that it was important not to divide up the effort at this time, or we might end up with two CDA level three versions. He appealed implicitly for support for the EHR sig within HL7 so that we should not be duplicating each other's efforts. The UK market is important to New Zealand vendors, so we will have to follow what they are doing for this reason as well. We should participate in their process while developing our own.

However, on attending the CDA level one course, I was struck by the adequacy of the present level one CDA for our simple initial needs. Later in conversation with Klaus Veil he said that only small countries such as Australia and New Zealand could run with simple projects, that big and important countries such as the UK would have to run with more ambitious projects. However, Liora pointed out that the USA is going for the CDA.

The CDA appears to be offered as a kind of 'shortcut' - a standard format to aim. Like a container on a cargo ship, it facilitates the transport of whatever you like, structured in any way you like. At present, the structuring of the 'body' of these documents has not yet been agreed on, and New Zealand is given the unique opportunity to design one that could be a serious candidate for a world model. This is because of our ability to gain consensus, but we still have to do it. The difficulty in designing such a thing is to get the ball rolling and to get consensus on a wide range of issues from stakeholders.

I do think the CDA could be developed into the HES. There is an opportunity for specific input from us on the actual template.

top of page

Access Revisited

It has not yet been determined how the interaction of CDAs with digital signatures would occur, or indeed any access control and authentication system for CDAs. Liora Alschuler mentioned that they had put in 'security hooks' to the sections of the CDA level one, so that levels of access could be segregated for particular sections. I described our idea in the ISO access paper of meta-data applied at point of care being used to control access by role, with 'need to know' being the access criterion and of override being subject to scrutiny in each and every occurrence.

She said that it would not be wise to identify levels of access to particular sections, and that external engines should provide those. She thought that access regulations should be changeable, and that could not be done while access role restrictions were built into the data. I said I thought it could still be done, and yet be changeable in its end behaviour, and we left that area unexplored at this stage.

Liora was given access control and signatures as part of her brief for the next iteration of CDA development. Bob Dolin was quite dismissive of the relevance of digital signatures and access control as part of the scope of the CDA. He emphasised that the CDA was a 'communication artifact only', and that he did not want to get into a situation where the CDA was being signed twice. I was left with the impression that the whole question of the status of the techniques of access regulation was left open but the 'hooks' were in there. This needs to be explored further.

Relevant here was a rather dramatic encounter with Bernd Bloebel. Bernd is an academic from the former East German town of Magdebourg, who has been the access person on ISOTC/215 WG4 (Security) committee. We met, and he said that access ideas very similar to the ones we had presented in Dunedin were now being used to control access with a data granularity of the individual data item. He asked me to join him and Gary Dickinson has co-chairing a new ISO/TC215 access committee, and I said I was honoured to accept that challenge, but I might not be able to do the travelling. He said it was the same with him. It would be interesting to see if he does come through with that contact.

I suggest that we place access control very high on our work programme. In particular, we should follow up the idea of using the separate stages of public and private key cryptography as part of the access process for health event summaries. We should follow up as well the idea of using role-for-access meta-data as a search criterion, using role information on the searcher's digital certificate to match that in the data.

top of page

The LOINC / SNOMED tutorial

I attended tutorials on LOINC and SNOMED coding systems Clem McDonald from the Regenstrief institute gave an overview of the Logical Observation Identifier Names and Codes, and the Regenstrief LOINC Mapping Assistant (RELMATM). This latter was a brilliant program for help with using LOINC codes. He said that the relationship between the two coding systems could be summarisedas 'if LOINC is the question, SNOMED is the answer', meaning that LOINC chiefly codes procedures and lab material, and SNOMED does the diagnosis.

Actually, Bob Dolin did not subscribe to that view. He said that it was especially in some of the procedural things that there were not SNOMED codes for that degree of granularity. Nobody challenged the view that SNOMED had won the coding wars by simply incorporating and re-organising certain sections in favour of Read codes. The relationship between LOINC and SNOMED appeared to be incompletely worked out at this stage. I have the LOINC CD and will distribute a copy to any inquirers (for a small fee - $10 to cover handling and postage). It also has the latest release of RELMA on it and manuals for both.

top of page

Impressions of the HL7 organisation

The HL7 website attracted a lot of discussion and adverse comment. With characteristic thoroughness, the administration team had done research into its user friendliness, and it was discussed at length at the TSC dinner. Already the site has changed and is more intuitive, but a downside is that more of the documents now attract some cost to download.

This is a very alert organisation, with many committed and 'feisty' participants. They appeared a little fazed by their success in taking over the world, and were trying to review their structures to cope with the enhanced role. The fervour and enthusiastic engagement of the participants characterised the meeting. Nearly 500 delegates attended lectures, tutorials, and discussions over the five days. This was a democratic gathering to the point of fault.

The Technical Steering Committee meeting was oversubscribed, and the group arguably too large to act as an efficient forum. The present structure of HL7 was debated in open forum, and informally. One view was that a new layer was needed in the hierarchy between board and members to make the meetings easier to manage, and perhaps to protect executives from trivia. This was going to be further discussed, along with the role of the international affiliates, and their voting rights in HL7 US meetings. Also mentioned was whether international affiliates had the same voting rights as US members. One definitely got the sense that it would be discussed when the international affiliates were not there. Klaus Veil was a strong advocate of the 'if it ain't broke, don't fix it' view. There was a raw democracy about the proceedings which, while perhaps leading to inefficient process, did ensure that the standards were developed democratically and in a fashion responsive to user needs.

The currently active models are the HL7 Reference Information Model (RIM) and the HL7 Version 3 Meta-model for the Message Development Framework. Just how capable HL7 is at the fine grain is displayed in the HL7 Meta-model that is downloadable for free as a PDF file (from www.HL7.org). One has to be quite well ahead in understanding object-oriented modeling techniques to even understand what much of this material is about, let alone contribute to the argument.

The HL7 Version 3 ballot result might be disappointing to Version 3 purists. At the technical steering committee dinner, Woody Beeler showed an insider's view of some wind-tunneled chasms in the desert near his home, and pointed out that from within that gigantic tunnel system you could literally sometimes not see the light at the end of the tunnel. He said that the view from inside Version 3 was like that at present. The truth is that the ballot did not go very well for Version 3. There was only about a 60% endorsement from the ballot for the data types and RIM structure. As Bob Dolin suggested it, many people had just not understood the material presented to them.

Beeler and others argued that the best plan was educate people into these concepts, but at the same time an uneasy sub-realisation that the whole scheme had not gained the understanding and support that it should have as a new version of a widely applied standard.

This was reflected in the plaintive quality to the debate about a new name for HL7 Version 3. There was a thought that the name 'Version 3' had a negative affective quality, because of the long development process. I venture to suggest that changing the name may not help by itself.

The question remains, is the methodology wrong or over complex, or is it just not yet adequately explained to the HL7 rank and file members?

David Fallas has contributed the following position statement on this to the report:

"One unfortunate side-effect of the HL7 v3 ballot documentation is that it has provided critics with evidence that v3 and the RIM can be dismissed as 'rocket science', too complex to actually implement in the real world. At the October 2000 HL7 meetings in St Louis, a time-line for v3 and the RIM was laid out, with the promise that a ballot would be published in preparation for a v3 release during 2001. Obviously this has not happened, with the majority of ballot returns being negative. The pressure to get the ballot material out by the end of July 2001 must have been enormous, and the result was what we saw - premature, with major deficiencies. One might say, a 'warts and all' demonstration of democracy at its best.

Dealing with the detail of ballot returns at the committee level, however, clearly showed that it was the poor quality of the documentation, which was causing consternation, not the underlying strategy of developing a messaging framework based on a comprehensive model. It is to the modeling strategy that we in New Zealand must look if we seek to improve the ability of our software systems to communicate effectively and to play a more meaningful role in supporting healthcare endeavors in this country.

The model of course goes much further than providing a foundation for messaging development. The HL7 RIM has evolved to its current state as a result of countless thousands of hours of refinement. Over the past two years, various committees have continually challenged its ability to address real life use cases, and in doing so have enormously reduced its complexity. The result is that the model can demonstrate its ability to not only support robust message development, but to also provide profound insights into healthcare application software design.

The clear message for those of us involved in this branch of the software industry is that we must lift our game if we are going to build software that will actually serve the needs of our customers. The HL7 RIM may well be imperfect, but to ignore what it begs to teach us is folly indeed. To paraphrase Stan Huff - we will descend into a marketplace infested with an array of quick-fix, '...pssst, wanna buy a watch?..' software products."

top of page

Open Source - Klaus Veil's Presentation

At the HL7 board, Klaus raised the whole matter of 'open source' standards, and how HL7 was both positioned and perceived in that debate. One advocacy that the New Zealand delegation had made to the ISO committee was that any global standard for should be 'open source', ie free to the user, and with the software code or specification being in the public domain.

Klaus had attended a meeting about open source in London, and reported the hostility and suspicion of that audience toward HL7, because it was perceived as not being open source in its products and policies. He charted some options for HL7 in this regard. There is a problem with revenue generation for standards development, and the chief sources of revenue to the organisation are sales of documents and standards, and membership subscriptions.

The other board members appeared unwilling to condone or even discuss 'open source' as an option for HL7. The reasoning was that the specifications for the standards were in the public domain anyway. Version 2.* of the HL7 standard, the one which is in wide use, had been offered to the ISO committee as a global standard. Many of the delegates held positions in companies supplying software for healthcare, and those are for the most part anything but open source. Klaus put the matter on the table, and thus on the long-term agenda of HL7, and he was encouraged to continue trying, but the members at that meeting did not engage in the open source argument further.

Later we had an interesting discussion of the 'open source' concept with Liora Alschuler and Klaus. The W3C consortium itself is heavily backed by leading technology companies, and the resulting web standards are themselves influenced by this financial input. The example XML (the Extensible Markup Language) was cited as an 'open' development that has born fruit. The way it happened was from the efforts of individuals, and the trend of the argument was that it was more the input and commitment from individuals that makes the difference rather than whether a standard fulfilled the criteria of 'open source'.

However, it remains a problem for HL7 in the wider global community. The situation is similar to that of the SNOMED/LOINC coding standard, which is well on the way to being the only global coding system, but which is not free to the user. Hopefully, a little further down the track, HL7 and SNOMED standards will be made available by some mechanism to unresourced users internationally. It is an unresolved contradiction in HL7s position, trying to be global and yet not quite free. The modest charges for HL7 products may reflect the reality of standards development work (despite the reliance on 'volunteers'), but it sits uneasily with many in the open source community.

top of page

Where We Stand - The Window for the Text

The level three CDA might act as a currency for healthcare. The clinical stories, in all their multiplicity, could continue to flow through the modeling and messaging system, but conveyed in standard-sized trucks.

Although the EHR may end up at one level as a succession of messages, the inclusion of free text as the 'lead story' ensures that the window is there for subtle Natural Language processes. As Liora Alschuler emphasised, the code is not the text. The machine acts a window for the story, and collects some of the debris (the data), but does not impede or direct it in any way, except by explicit intent. Healthcare professionals interact with the stories, including surgically, trying to determine outcomes.

The British interpretation of the decision support engine might intervene directly in clinical reasoning, and therefore be a filter for the story, rather than a window. The story would get through in other ways, but this might make it dysfunctional. I am not sure that it is the role of a standards body to concern itself at all with decision support or clinical guidelines. Any perception by the professional community that the standards work will be used as a new form of constraint on provider treatment options would make it hard to implement.

This also may be partly why 'decision support' as a concept is so hard to get going in electronic medicine. It is the same thing as clinical guidelines, as far as the practical and technical problems of implementation are concerned. However, decision support is orientated to maximise options, where clinical guidelines often try to constrain them.

I doubt if support for either is required at this stage of the evolution of a communication and data standard for healthcare. Instead, I support the view that clinical reality consists of a succession of 'grasps', which are both motor and perceptual. This concept is similar to the RIM 'ACT' class, which incorporates both the 'act' of observing and the observation itself as an instance of a class. It is also in accord with the way the brain processes the experiential stream. [Mair, 'The Multioodal Melody of the Text'; abstract found at www.semioticon.com, Virtual symposia, 'On the Multimodality of Human Communication: Theories, Problems and Applications'. (Program)]

The clinical stream can be adequately conveyed with a 'primal' chunking into 'grasp' units. They can be very diverse in their manifestations, and can included entities such as 'states of affairs', Californian style 'Encounters', or surgical sessions with a single client, etc. They have in common that they are 'hands-on' space-time events, with clear temporal boundaries. The time between meeting and parting is variously structured by the behaviour of the actors and their machine agents. Tacked end to end, summaries of these health events will make a simple basic electronic health record.

I suspect that clinical guidelines and access to decision support will be the property of particular applications, whereas the 'standard' is about how we transmit and store medical data in a generic way. However, it is interesting that so many people want to worry about the plumbing of decision support, before we have even built the data houses for the plumbers to work in.

top of page

The Global Context for our National Drama

The future of New Zealand Healthcare will be constrained by New Zealand's economic trajectory, in a fast-changing global context. For New Zealand, as for other nations, the absence of a supra-national plan or concept of a world order at this time makes it hard to know what to aim for, in terms of the economic and demographic environment in one years time, let alone five years or longer.

The implementation of electronic medicine has got rather bogged down in many countries for the lack of a workable universal way of doing health records that would make this data machine processable.

We are in a position to actually plan the harvest of some of that data, and if we succeed, we get to design how to do it. If we could manage health 'on its data', by machine processing, this would allow us to free up more of the resources of the health sector for managing disease.

In working on the HES, or CDA, or at least in the area of healthcare standards, we might help to provide one of the currencies which will facilitated the 'computation' between different ways of life and the resources they use. This mutual adjustment must occur over the next 20 years or so, for survival.

Such global contextual dreaming might be important for our motivation (the view from our 'vantage point'), but it does not itself do the job of implementing the HES.

top of page

Game Plan

David Fallas and I worked out the direction that we thought a New Zealand Health Event Summary project might take, and we offer these thoughts in conclusion of this report.

We note that Stephen Chu's project is already well established.

Mike Mair
Timaru
27 October 2001