Tuesday, January 14, 2014

The NEHTA Terminology Team Troop Off To Washington In October To Make Sure All Is Well On The Global Stage. Seems There Is The Odd Problem!

This appeared a few days ago.

HTSDO Conference Report October 2013

Created on Thursday, 02 January 2014
The IHTSDO Conference Report October 2013 Meeting has been published.
 The International Health Terminology Standards Development Organisation (IHTSDO) Conference Report provides summary information on the international activities and areas of work as discussed at the IHTSDO October 2013 Working Meeting held in Washington DC, USA. The report includes an update from the Content, Implementation & Innovation, Quality Assurance, and Technical Committees, the Substance Hierarchy Redesign Special Interest Group as well as the General Assembly and Member Forum.
Here is the link:
As you will see from the description some of the content is pretty specialist and heavy going.
What I was somewhat surprised by was the apparent instability of the current manifestations of SNOMED CT - some 12 years after it was finally released.
See here for all the history:
In my innocence I has imagined the basic designs and hierarchies for the terminology were settled and that adjustments were all at the edges.
Not so it would seem.
We read:
1. Regarding the  Substance hierarchy redesign project
“Concerns about the ambiguity of many of the existing concepts were raised. As the Substance hierarchy doesn’t have a model at the moment, this ambiguity is a result of a mixture of poorly-chosen legacy terms and lack of standardised nomenclature for many substances. Completely replacing all concepts would have a large impact on both the modelling of other hierarchies and implementers. A pragmatic approach is required.”  (p20)
2. Regarding Versioning.
“Dion McMurtrie (NEHTA) presented on identifying coding systems and versioning of the coding system in clinical documents. This topic had been raised previously with respect to HL7 messages and never reached resolution at the IHTSDO, with the IHTSDO deferring the issue to HL7. However the committee recognised that the IHTSDO should have a position on versioning SNOMED CT, particularly in relation to extensions, given that they are core to SNOMED CT. Work is to progress to define a draft versioning scheme for SNOMED CT for comment out of the URI specification work.” (p19)
3.  Regarding Anatomy Redesign
  • Many of the critiques of SNOMED CT stem from incorrectly modelled anatomy – Kent Spackman showed an OWL preview of the revised Anatomy model.
  •  The Anatomy redesign requires changes to the SNOMED CT model – addition of new logic: Sufficient definitions, General Concept Inclusions, role chaining and anonymous class axioms (e.g. any part of something that is lateral is also itself lateral).
  •  Other additions include surfaces, “skin of”, “subcutaneous tissue of”, “part of lateral half of”, and “bone tissue of” (p10).
So substance terminology (drugs), anatomy and version control are all under a lot of review!
15 years since all this started it is obviously nowhere near done. I wonder when?
Looks like the complexity of such endeavours has struck again and real sematic interoperability is as far away as ever!
One simple question has to be if it is still a work in progress how are those who are actually implementing affected? It must have some impact.
I hope the team are back and hard at it - seems like there is a fair bit to be done.
David.

15 comments:

Eric Browne said...

David,

This is all very depressing. We don't have an international body dealing effectively with healthcare terminology. Witness last year's agreement between the IHTSDO and Regenstrief that entrenches the organisations' control of their own territory at the expense of improving healthcare terminologies.

But nor do we even have a national body here in Australia that can address the basic interoperability needs of the healthcare community.

The HL7 v2 based standards that have evolved slowly over the years are still inadequate to ensure interoperability of even basic information about patients. No national body is supporting a conformance framework for this basic messaging. There is no adequate suite of test messages that addresses this need. There is no facility for implementers to raise questions and have them addressed. It is left to volunteers like Grahame Grieve ( or sometimes me ) to address ad hoc issues on their blogs.

NEHTA has never made any attempt to adequately meet the needs of implementers. Standards Australia does not have the brief nor the resources to do so. There is no place to go to download the latest authoritative value set for a particular field in a message or clinical document. Their is no decent terminology service to underpin interoperability between systems.

What is worse, is that there is nothing on the horizon to suggest that this parlous state of affairs is likely to change in the near, or even distant future.

Enrico Coiera said...

My view is not that anyone is doing a poor job, or is not up to the task (indeed these folks seem pretty smart and well motivated to me) but that this failure to agree on foundational concepts is fundamentally inherent in this conceptualisation of a standard i.e. this is exactly the outcome you expect if you try to do it his way.

Most of the world long ago adopted the notion of a Platonic Ideal or singular truth, and recognised that all concepts are relative and probabilistic contracts ... not so in health standards.

In contrast Google seems to navigate ontology and terminology pretty well without explicit terminological structures like SNOMED to help it.

Here's my heretic take on the standards process ....

http://coiera.com/2013/11/01/are-standards-necessary/

an excerpt

"It should come as no surprise then that standardisation is widely resisted, except perhaps by standards makers. Even then they tend to aggregate in competing tribes pushing one version of a standard over another. ...

In IT, standards committees sit for years arguing over what the ‘right’ standard is, only to find that once published, there are competing standards in the marketplace, and that technology vendors resist because of the cost of upgrading their systems to meet the new standard. Pragmatic experience in healthcare indicates standards can stifle local innovation and expertise[4]. In resource-constrained settings, trying to become standards compliant simply moves crucial resources away from front-line service provision."

(I will now remove to a quiet place and put on a hard hat for all the bricks that fly back.)

Grahame Grieve said...

I think that Snomed-CT has a very long way to go before it is the answer, but I also don't think that google have demonstrated that you don't need explicit terminological structures. Perhaps they've shown that for speculative search on a massive source, it works ok, but driving google search to find what you want can be a real art-form.

I disagree with Enrico that standards are the problem here, at least in principle. Standards can either foster or stifle innovation - it depends how they approach the problem.

Some standards are very actively resisted, while others get enthusiastic uptake. The core of the problem is this: "trying to become standards compliant simply moves crucial resources away from front-line service provision" - in principle, you would think that the point of standards is to improve front-line service provision...

Eric Browne said...

"trying to become standards compliant simply moves crucial resources away from front-line service provision".
This could be said of almost any investment in healthcare. The investment becomes worthwhile if the benefits outweigh the costs. My complaint is that there is little recognition of either the cost or the benefit in many cases. If we don't have implementations that can demonstrate these, then how can e-health policy be guided ? We don't just need evidence-based healthcare, we need evidence-based e-health to support it.
Perhaps if each and every healthcare system vendor were to invest the billions that Google has in its search engine infrastructure, we would be able to share healthcare information safely, efficiently and effectively.
But firstly, I don't think we could ever afford it, and secondly, I would not like to trust my life to it.

krowwolf said...

The Dictionary and Thesaurus are not "done" either.

> So substance terminology (drugs), anatomy and version control are all under a lot of review!

Drugs - there are new ones all the time, and withdrawn ones

Substances/Anatomy - there isn't a one-best way to model things. The use and requirements of SNOMED has evolved and thus the upper level modelling needs to change. It is taking time because we need to get it right to avoid future changes. Anatomy is especially hard - the FMA has it's own problems, dealing with surfaces, regions, connections, networks is fraught.

Versions - this one is a mis-understanding. The technical issues for version control were fixed years ago with the RF2 file format (essentially, allowing for national extensions). The actual issue being reported here is about clarifying a set of canonical machine-readable names for versions.

To Enrico, if a concept-based approach was value-less then Google would not be investing heavily in the Google Knowledge Graph.

Regarding the benefits or otherwise of standards one of the major challenges with this space is that the benefits are often accrued downstream from the costs. Essentially, system boundaries create an opportunity to optimise for local maxima rather than a global maximum.

Dr David More MB PhD FACHI said...

The Dictionary and Thesaurus are not "done" either.

We all know that - but the structure and mechanisms to modify are!

> So substance terminology (drugs), anatomy and version control are all under a lot of review!

Drugs - there are new ones all the time, and withdrawn ones.

So hardly news - its the framework not the items that are still not sorted

Substances/Anatomy - there isn't a one-best way to model things. The use and requirements of SNOMED has evolved and thus the upper level modelling needs to change. It is taking time because we need to get it right to avoid future changes. Anatomy is especially hard - the FMA has it's own problems, dealing with surfaces, regions, connections, networks is fraught.

Versions - this one is a mis-understanding. The technical issues for version control were fixed years ago with the RF2 file format (essentially, allowing for national extensions). The actual issue being reported here is about clarifying a set of canonical machine-readable names for versions.

Sorry the issue is CDA documents managing different versions. I believe it is still an issue

As a note there are some very smart people commenting above and they don't buy your apologia.

It needs more work at a fundamental level or a rethink!

David.

Andrew McIntyre said...

SNOMED-CT may not have been perfect but 5 years ago I was told that the drug hierarchy was not good enough and yet we still see no replacement. While not perfect the existing SNOMED-CT would - with some investment make a significant difference. In a commercial world you would maintain the existing version while working on a better replacement but with the involvement of a public service mentality we have no maintenance of the existing terminology, a useless mostly disconnected brain dead AMT and as a result a lack of new terms in the existing SNOMED-CT with no replacement in sight.

If these people were doctors then patients would die of starvation while they kept them nil by mouth for months debating which operation to do!

Existing standards, (while neglected because of a flood of government money to fund pseudo-academic rainbow chasers,) do actually work and I think its time to actually focus on whats out there working and tell the rainbow chasers to come back when they had the pot of gold firmly in their grip.

We need some pragmatism. The fundamental issue with the current V2 standards is a lack of compliance rather than the actual standards themselves. Government bankrolling of organisations such as NEHTA has us off the track, lost in the weeds, while working eHealth exists on starvation rations. New "solutions" should be explored, but only using 5% of the money, not 99.5%. The lack of delivery in the last 10 years is surely proof that rainbow chasing is not a good investment.

The level of hubris makes resolution unlikely, particularly while politicians continue to provide billions of tax payers dollars to spike the punch. The older standards, produced by sober people with real IT ability do work, if you comply with them. We need to go back to the future.

Anonymous said...

The problem with eHealth standards is that very few people or organisations really understand what is available and how and when they should be used. (with the possible exception of Mr Grieve and Mr Browne, national treasure nerds).
For those lesser beings working at the healthcare coal face on acquiring, developing, integrating and implementing systems, it is still very very hard to know what should be required in a tender specification, how systems should be configured and interfaced and exactly how to incorporate things like SNOMED, AMT etc. We all want to make sure that we can share healthcare documents with other systems and national EHRs, but where is the list of standards? Searching the Standards Australia site doesn't help unless you know exactly what you want. Checking out the NEHTA site won't tell you which set of allergies codes should be used in which context.
Like Eric, I always thought that NEHTA would help us with all of this. The National E-Health Record (PCEHR) concept of operations at one point described a Terminology Service, and this sounded promising, although there seems to have been no progression with that.
Instead, it all gets more complex and mysterious…
Don't worry Enrico, standards are not stifling innovation and expertise, because in the end we simply have to get on with it, standards or not.

Enrico Coiera said...

"Don't worry Enrico, standards are not stifling innovation and expertise, because in the end we simply have to get on with it, standards or not."

I guess for that to be true then the standards enterprise would have to be zero cost. But its not. Think what NEHTA has spent. Think what the brilliant minds of the Grieve's and Brown's might be doing if no longer trying to untangle Gordian knots. As we operate in what seems to be a zero-sum game, a dollar spent in one place is not spent in another.

Don't get me wrong - if you read my blog posting, you see I'm not anti-standard, I'm anti standards without rationale or cost-benefit analyses. Standards are sometimes essential, sometimes a fool's errand. Asking if we are in the first or second circumstance seems to be not a question often considered, and with little science to assist in answering it, if asked.

krowwolf said...

>> Drugs - there are new ones all the time, and withdrawn ones.

> So hardly news - its the framework not the items that are still not sorted


Indeed, and that's the Substances re-design (note that this is entirely separate from AMT, but the long timeframes involved are why, AFAIK, AMT was created as a separate thing).

>> Versions - this one is a mis-understanding. The technical issues for version control were fixed years ago with the RF2 file format (essentially, allowing for national extensions). The actual issue being reported here is about clarifying a set of canonical machine-readable names for versions.

> Sorry the issue is CDA documents managing different versions. I believe it is still an issue


Ah, that's what you're referring to. The issue is actually one that sits between HL7 and IHTSDO; HL7 needs to decide whether to use one or multiple OIDs to deal with the different editions of SNOMED CT and IHTSDO needs to standardise on the version strings in order for a single OID to work (although the HL7 community does a poor/non-existent job of tracking terminology versions anyway). Good news is that the IHTSDO work is essentially done and just working through the process. In addition, the approach from HL7 with FHIR builds on this and the single OID approach.

> As a note there are some very smart people commenting above and they don't buy your apologia.

I know these people and I'll let them tell me their opinions.

> It needs more work at a fundamental level or a rethink!

Which takes time, yet impatience seems to be the crux of much of the criticism here.

Dr David More MB PhD FACHI said...

"I know these people and I'll let them tell me their opinions."

I know them all too and I think they have just told you what they think.

Maybe just ease up a little on the hubris.

David.

Anonymous said...

Andrew,

I think you'll find that if you follow the money spent by Government on eHealth of Taxpayers' monies the "rainbow chasers" have already had, enjoyed and consumed the said "pot of gold"!

One organisation in particular you know and love all too well has enjoyed alone well over 50% of the gold deposited in the pot, robbed from the poor downtrodden NET TAX PAYER!

While they're profiting so handsomely from their I'll gotten gains, why on earth would they "willingly" change anything??

The "rainbow chasers" are well and truly fat, dumb and happy at all our sorry expense...

Anonymous said...

"One organisation in particular you know and love all too well ...."

Is this one of their radio ads?
http://www.waltzingrhino.com/Humor/business/accidenture.mp3

Grahame Grieve said...

One of the national treasure nerds here...

"The issue is actually one that sits between HL7 and IHTSDO; HL7 needs to decide whether to use one or multiple OIDs to deal with the different editions of SNOMED CT and IHTSDO needs to standardise on the version strings in order for a single OID to work (although the HL7 community does a poor/non-existent job of tracking terminology versions anyway). Good news is that the IHTSDO work is essentially done and just working through the process. In addition, the approach from HL7 with FHIR builds on this and the single OID approach"

It's really good when you can call your work done because you don't have to engage with the real thorny issues :-(. I personally do not think that we have begun to plumb the depths of maintenance, distribution, and extension issues in SnomedCT space yet. That IHTSDO work can be called essentially done when no one in the user community understands what it means demonstrates part of the problem.

It's easy to say "the HL7 community does a poor/non-existent job of tracking terminology versions anyway", but perhaps that speaks more to the fact that terminologies aren't easily used. Name one practical outcome you'd get for actually tracking versions... they only exist in a theoretical world where the users are perfectly educated and don't do stupid things. See, for instance, this: http://www.healthintersections.com.au/?p=567

As for FHIR - I've done what I can, but it's lipstick on a pig.




Dr David More MB PhD FACHI said...

Thanks Grahame for some clarity for the masses.

David.