this is an often debated question, and after coming across (for the 100th time) just such a debate recently online, I thought it might be interesting to try to get to the bottom of the question in some way. The basic idea posted here. It is of course not scientific work, but I would be interested in the views of others on this concept.
(copy of comment posted onto Toms blog)
Thanks Tom,
We share some common thinking on this, though I should make some additional points which I hope add to this important debate.
You didn't mention the complex adaptive system nature of healthcare systems, which I know you are well aware of. The complexity of healthcare leads me to believe trying to agree and impose a single agreed healthcare information model is impossible to achieve.
I would argue that the right solution in this healthcare space will evolve over time, albeit this could be accelerated if there was better working between eHealth standards bodies and the innovators at the frontline.
You rightly mention "patterns" which should inform the best fit information model for the complexity of healthcare.
This approach to looking for patterns in healthcare led me to openEHR when I first came across it.
The patterns I was looking for.. was within the generic process of healthcare and I would suggest that any healthcare service oriented architecture and related standards should primarily be aimed at supporting the core service of healthcare, ie the healthcare process.
Thankfully here the openEHR reference model comes into its own as the primary patterns it fits with is the generic clinical process cycle, which I know informed the openEHR reference model (you may have explicitly/inadvertently omitted that in your explanation).
I make the related case for generic process oriented approach to healthcare standards here ..
<a href="http://frectal.com/book/healthcare-change-the-way-forward/healthcare-chasing-the-right-fit-between-process-and-it/" title="Healthcare: chasing the right fit between Process and IT..">Healthcare: chasing the right fit between Process and IT.</a> and here..
<a href="http://frectal.com/book/healthcare-change-the-way-forward/healthcare-openehr’s-potential-to-handle-complexity-diversity/" title="Healthcare: openEHR’s potential to handle complexity & diversity">Healthcare: openEHR’s potential to handle complexity & diversity
I won't argue that all standards bodies should come around to our way of thinking in one go, our approach should be a continued drive for Healthcare Improvement, supported by Information Technology (& related standards inc openEHR).
I hope/believe in time the power of the openEHR reference model/archetypes/templates etc show such benefit that others come around to this complex adaptive system/generic process oriented approach in due course as things evolve..
Of course where healthcare improvements evolve from other shared standards efforts, we should look to learn from them too..
Regards,
Tony
Hi Thomas,
I’ve left a comment in your blog but is not appearing, so I comment your idea here.
I don’t think today it can be possible to have one information model agreed by all the medical informatics community, but I think if we can agree in a common metamodel like an ontology that represent the more generic concepts in medicine, like people, processes, resources, records, etc, we will be one step closer to a common IM. Because if we can agree on that ontology, all the information models in healthcare MUST follow the ontology, so, different information models can live together, but they model the same concepts (semantically speaking). With different models, but semantically equivalent, the point of convergency will be closer.
Hi Pablo,
At a very high-level you may be correct but the big problem with most medical record structures is that they are currently too variably defined used and understood to be ontologised. Try thinking about the challenges of ontologising the structures of governments across perhaps Europe and you will get a flavour of the challenge. One of the benefits of the archetype approach is that it allows us to focus on small areas but with a broad audience to try and get small pockets of usable consensus. Perhaps in the future much of this will be ‘true’ enough to define ontological relationships but we are a long way from that position.
Ian
Dr Ian McNicoll
office +44 (0)1536 414 994
+44 (0)2032 392 970
fax +44 (0)1536 516317
mobile +44 (0)775 209 7859
skype ianmcnicoll
ian.mcnicoll@oceaninformatics.com
Clinical Modelling Consultant, Ocean Informatics, UK
openEHR Clinical Knowledge Editor www.openehr.org/knowledge
Honorary Senior Research Associate, CHIME, UCL
BCS Primary Health Care www.phcsg.org
Hi Thomas,
I’ve left a comment in your blog but is not appearing, so I comment your idea here.
I don’t think today it can be possible to have one information model agreed by all the medical informatics community, but I think if we can agree in a common metamodel like an ontology that represent the more generic concepts in medicine, like people, processes, resources, records, etc, we will be one step closer to a common IM.
yes, that’s pretty much what I was suggesting.
Because if we can agree on that ontology, all the information models in healthcare MUST follow the ontology, so, different information models can live together, but they model the same concepts (semantically speaking). With different models, but semantically equivalent, the point of convergency will be closer.
information models, at least abstract ones are in effect an ontology in themselves: they are a description of information that either exists, or we want to exist. So it seems reasonable that a pragmatic UML model, with an appropriate level of abstraction can be used for just this purpose - to describe and agree on key patterns.
If this were true, it would mean that the challenges for agreement are:
An example of one of the patterns, modelled in UML is the ‘history of events’ one here. Could this or something like it be agreed across e-health for interoperably representing the common concept of a history of events?
If sufficient patterns could be agreed, then an ‘information model’ consisting of these would in effect be a ‘common information model’ for the medical informatics community - whose scope is interoperable representation of the patterns contained within.
It seems to me that this would be a great step forward.
Hi,
I have been reading architecture overview of openEHR, and I would like to make some comments, questions with respect to the ontological separation:
a) There has not been an international agreement on the Reference model, that is supposed to be stable, (openEHR RM vs HL7 RIM vs …)
I am not surprised as it is usually this level (RM) that is implemented in software according to the openehr architecture overview. But I would like to make clear for those that were reading the posts of “one model vs one framework in e-health” that I was not referring to that level of modeling.
My interpretation on this issue is that we have many health standards at that level 
b) There has not been an international agreement on the “Domain Base Concept Model”, level 2 invariant domain concepts according to openEHR ontological layering, where the archetypes are based on (clinical care entries – instructions, evaluations, observations, actions, etc), administration entries (admission, registration, accounting, etc)
My interpretation on that issue is that we have many health standards at that level too
DO NOTE also the comment on the presentation of Ocean Informatics at UCL in the year 2005:
This level must be standardised and agreed for archetypes to be sharable. So what has been the progress on that ?
c) There has not been an international agreement on the variant re-usable domain concepts, openEHR level 3, openEHR ARCHETYPES (e.g. medication, symptoms, imaging, diagnosis, procedure, etc…).
My interpretation on this issue is that we have many health standards at that level too
OK that is perhaps the reason that many presentations end up with the tower of BABEL
Therefore domain concepts are defined differently in HL7 and openEHR and the same is true for any other organization that attempted to write specifications for that level. But It is common sense I believe for any architect/developer/analyst that worked in industry, or even an IT database course student, to attempt solutions at a business domain and to define, to model entities at semantic level as a starting point. In fact if you study many of the commercial ehealth systems and open source ehealth software you will do of course find classes (objects) and tables that represent such entities.
So why is not there ONE information model at that level for many if not all to agree on ? Because I suppose many will answer there is great variability and here it comes the solution of many organizations including openEHR to specify FIRST persistence level (see also comments and discussion on one information model vs one framework in e-health), meaning to deal with data types and then start building the other layers. Invariant parts come first no matter how close they are to the health domain !
OK let us focus on the clinical content for a moment and suppose we start modeling this FIRST as it has already happened. Think also about the conversion issue of legacy, old systems.
For this argument I will choose existing clinical models, archetypes defined at the clinical knowledge manager:
Medication description (Dose Unit, Dose instruction, indications, etc…)
Symptom Features (Locations, Variation, Severity, etc…)
Imaging Data (Imaging procedure, anatomical site, location, etc…)
You noticed of course that I have taken content from different sections of a typical EHR.
Questions
Hi,
I have been reading architecture overview of openEHR, and I would like to make some comments, questions with respect to the ontological separation:
a) There has not been an international agreement on the Reference model, that is supposed to be stable, (openEHR RM vs HL7 RIM vs …)
I am not surprised as it is usually this level (RM) that is implemented in software according to the openehr architecture overview. But I would like to make clear for those that were reading the posts of “one model vs one framework in e-health” that I was not referring to that level of modeling.
My interpretation on this issue is that we have many health standards at that level
Hi Ian,
As I see the issue, the medical record internal structure (as a model) could be an instance of “the healthcare ontology”, and if some of these models do not follow the general semantic rules of the ontology, then we can say that they are bad defined. I mean that the ontology shouldn’t model the internal structure of the medical records, I think it have to model the big record entities (maybe compositions, sections and entries), and other related entities like persons, organizations, roles, resources, processes, etc. There’s one (bad) thing we do over and over again while modeling clinical records: separating the clinical process from the record of the process. If this “almighty healthcare ontology” can be created some day, I think we must model the clinical process first, and then the clinical record that tells the story of one excecution of the process (because the record is just that, information about an instance of the clinical each process).
Just my grain of sand 
Is there a ‘healthcare workflow’ ontology?
E.g an ‘encounter’ archetype showing the possible interactions with the EHR, specialised into ‘Visit a GP’, ‘Nursing housecall’, ‘Hospital Grand Rounds’, ‘Specialist’ templates etc.
If these use cases could be agreed they could be a basis for separate compatible implementations.
Regards,
Colin Sutton
P.S. my interest is in including clinical trial registration and feedback of clinical trial conclusions and systematic reviews into the processes.
According to HL7 RIM there are four core classes
Entity, Role, Participation Act
Then one can relate healthcare domain concepts such as visit, admission, patient, health care professional, health care organization, observation, procedure based on these core classes.
If there are standard, common, general, archetypes then obviously one can use them in such a way to model clinical process !
I am not sure what is the equivalent of HL7 RIM core classes in openEHR to model clinical process and how one can use archetypes in that respect.
Athanassios
PS:
Athanassios: Why not implementing these standard archetypes with classes-objects at programming level, if we agree the names of the attributes (features) ?
Thomas: Because this is exactly what we want to get away from; that is the road to disaster - it is the 1980s approach where everything in the domain becomes a class and/or a relational table.
Fortunately there is a better approach that is emerging, whereby these classes/XSDs/etc can be generated from templates, in such a way that the data created can always be converted back to canonical form.
Athanassios: I simply cannot imagine how one can escape completely from defining the overall picture at conceptual (programming level), i.e. defining standard, common, general “archetypes” (core classes depending on the problem to use). I suppose in many cases one would like to build a completely new clinical information systems based on these core classes and I expect highly complex business logic behind that to capture both online and offline, dynamic and static, clinical, administrative processes. So If there is indeed such a different way to view software engineering as you mentioned, I think it will have to become more understandable and common across developer communities including RIMBAA and others.
Hi Colin,
The COMPOSITION class from which all Composition archetypes are derived, has a context attribute which carries a great deal of the who, why, when of an Encounter. You could create a separate archetype for each of the encounter types you mentioned but in practical applications I have found that a generic Encounter archetype is sufficient, perhaps just renamed in the template to the specific type of encounter, or by adding a ‘document type’ code in the archetype.
Ian
Dr Ian McNicoll
office +44 (0)1536 414 994
+44 (0)2032 392 970
fax +44 (0)1536 516317
mobile +44 (0)775 209 7859
skype ianmcnicoll
ian.mcnicoll@oceaninformatics.com
Clinical Modelling Consultant, Ocean Informatics, UK
openEHR Clinical Knowledge Editor www.openehr.org/knowledge
Honorary Senior Research Associate, CHIME, UCL
BCS Primary Health Care www.phcsg.org
If I might add to the debate about supporting clinical processes in e
Health standards..
I've written about the complexity of healthcare, the need to support
healthcare people (ie frontline clinical staff) and their processes with
information technology, which requires standards and better value for money.
While I think its impossible to get complete agreement on the right
standard/information model required, right now I believe openEHR is a
very good fit with whats needed.
http://frectal.com/book/healthcare-change-the-way-forward/healthcare-openehr’s-potential-to-handle-complexity-diversity/
I don't suggest its the only way of doing it, but rather than debate
this ad infinitum believe that tackling healthcare process improvements
with information technology trialing these standards is the way to
progress the science. I've suggesed greater open source collaboration
between standards groups and those at frontline should help. Hence my
push for Opereffa and recently locally in my hospital an open source
portal project..
http://frectal.com/book/healthcare-change-the-way-forward/healthcare-change-why-“open-source”-is-part-of-the-recipe/
Thanks to others in this field who are also sharing their work publicly.
As to the way from where we are to where we need to be may I offer a
slide/picture that I find can help explain one way..
http://frectal.files.wordpress.com/2011/05/architecturemovetoopenplatformv2.png
FYI, It is deliberately imperfect.. so no points for spotting problems
with it.
It explores 3 options A B C which I beleive are universally applicable.
Option A is the current mess of >200 legacy systems I have in my
hospital(s) that dont connect well and pose a large cognitive load on
clinicians, .. for some this is the status quo option.
Option B is move towards "best of breed" ie integration via whatever
means are available, eg integration engines.
Option C offer
Ci) Rip and Replace with vendor X- (anyone keen on that dependency?)
Cii) Open Platform - with key open source components (inc archetypes
etc), to my mind the way forward
The jump from A to C is huge for many people, too big for one leap.
Locally here we are purusing Option B for now, ie integration with a web
UI/portal layer facing the clinicians.
I hope to be able to introduce a single archetype into that mix (eg
Allergy) as the organisation realises we need agreement around core
concepts.. Others from the "top 10", ie Problem/Diagnosis, Medication
etc would be likely to be next likely contendors.
Over time I see that an evolving standards based (eg archetype) could
replace the current architecture in a move towards an open platform..
this would/will take years to achieve, but is the best current fit the
with organic movement of the ecosystem I work within here.
I'm guessing that most folk on the list are struggling with how to move
from A to B towards Cii.. so I see that rather than waiting for
national/international agreement on the ideal reference model, that
most/all of those of us are tackling similar challenges at the frontline
will leverage standards where they help (not hinder). Anything that
offers help in sustainability, maintainability and ease of collaboration
appeals here+...
openEHR and related tooling appear to offer that potential...however I
still dont have the openEHR tooling environment I need, as Toms says
tooling costs money and as Seref notes the range of requirements are the
size of an elephant..
So for a starter.. from a common clinical process perspective, who can
offer a web-based (openEHR or other standard) widget/service to add
clinical value in supporting documentation of adverse
reactions/allergies at the frontline?
If other approaches (i.e. via related tooling, not just debate/academic
discussion/paper based standards) add value to support such common
clinical processes then the complex ecosystem of healthcare will start
to leverage..
Regards
Tony
Dr. Tony Shannon
Consultant in Emergency Medicine, Leeds Teaching Hospitals
Clinical Lead for Informatics, Leeds Teaching Hospitals
Chair, Clinical Review Board, openEHR Foundation
+44.789.988 5068 tony.shannon@nhs.net
According to HL7 RIM there are four core classes
Entity, Role, Participation Act
Then one can relate healthcare domain concepts such as visit, admission, patient, health care professional, health care organization, observation, procedure based on these core classes.
If there are standard, common, general, archetypes then obviously one can use them in such a way to model clinical process !
I am not sure what is the equivalent of HL7 RIM core classes in openEHR to model clinical process and how one can use archetypes in that respect.
Hi Thomas,
I agree that the essence of this issue is to detect “generic/reusable patters” or “ontological components”, and then derive our “information models” from these components.
Just two thoughts:
A marketing issue: If these patterns are directly derived from some existent IM, then we will have the same trouble of defining one common IM: my model is better than yours, so we’ll never agree. I think we must represent and present these patterns as ontological components, trying to avoid the copy&paste of the pattern from one o the other IM. I know that de openEHR IM is derived from an ontologial analisys of thereality,so we can see it as a concrete ontology for healthcare information, but it is not presented as a concrete ontology, is presented as an IM to be implemented on software. I don’t know if I mess up this explanation, just want to tell that we must be careful in the way we present, represent and name things if we want a global agreement.
The current openEHR IM is great for dealing with clinical record information and micro clinical processes (Instructions, Activities, Actions and the associated state machine), but not for the macro processes that embrace the micro clinical processes, and for building computerized information systems we need those processes modeled also. For example, if a traumatized patient comes to the ER in an ambulance, and then is derived to an ICU, we have a global process of “trauma care”, then we have macro processes like “prehospitalary care”, “emergency care”, and “ICU care”. In each of these macro processes we have multiple workflows excecuted in paralel, and different types processes but interdependent like administrative (patient identification, human resource assignation, etc), clinical (observations, actions, evaluation, etc), accounting (resource ussage), and financial (healthcare costs). so, if we model patters or ontological components, I think these must represent (in a generic way) the macro processes, not only the micro-clinical processes.
Hi Pablo,
Hi Thomas,
I agree that the essence of this issue is to detect “generic/reusable patters” or “ontological components”, and then derive our “information models” from these components.
Just two thoughts:
- A marketing issue: If these patterns are directly derived from some existent IM, then we will have the same trouble of defining one common IM: my model is better than yours, so we’ll never agree.
that’s probably true. I think the right approach would be to establish an ‘empty whiteboard’ place to create such a model from pieces of other models, however modified, so as to be consistent.
I think we must represent and present these patterns as ontological components, trying to avoid the copy&paste of the pattern from one o the other IM.
the problem with this approach is that it is not likely to be sufficiently expressive, even though I think you are theoretically correct. But software people like pragmatic models and formalisms, and presenting e.g. an OWL ontology is just making it harder to understand the semantics not easier.
I know that de openEHR IM is derived from an ontologial analisys of thereality,so we can see it as a concrete ontology for healthcare information, but it is not presented as a concrete ontology, is presented as an IM to be implemented on software. I don’t know if I mess up this explanation, just want to tell that we must be careful in the way we present, represent and name things if we want a global agreement.
I think there is no tool or formalism today that does exactly what we want. UML 2.x is weak on its definition of inheritance (it doesn’t properly distinguish is-a inheritance from other kinds of non-substituting inheritance); OWL2 is weak on the semantics of associations and encapsulation. We await a true ontological formalism Until then I think we have to make do with diagrams, and bits of OWL and UML as they suit us.
- The current openEHR IM is great for dealing with clinical record information and micro clinical processes (Instructions, Activities, Actions and the associated state machine), but not for the macro processes that embrace the micro clinical processes, and for building computerized information systems we need those processes modeled also. For example, if a traumatized patient comes to the ER in an ambulance, and then is derived to an ICU, we have a global process of “trauma care”, then we have macro processes like “prehospitalary care”, “emergency care”, and “ICU care”. In each of these macro processes we have multiple workflows excecuted in paralel, and different types processes but interdependent like administrative (patient identification, human resource assignation, etc), clinical (observations, actions, evaluation, etc), accounting (resource ussage), and financial (healthcare costs). so, if we model patters or ontological components, I think these must represent (in a generic way) the macro processes, not only the micro-clinical processes.
yes, these are good points, and still more work is needed to formalise these processes. I would however call this a different problem than that of the basic patterns. We at least need the patterns to simply build basic pieces of information.
Hello everyone
It has been very interesting to read the post and follow the subsequent
discussion about [the quest for] "one information model" in e-health and
especially the "patterns" part.
I am not sure if there can be "one information model" in e-health but
what i think that can be done is a ranking of available models according
to how expressive they are and patterns (patterns of data structures and
types) would be key to this.
For example, at the low end of the spectrum, we could have some really
simple models which are able to describe hierarchical information up to
a certain depth. For example, you can describe the data about a document
as a "sequence of element" where "element" can be "simple" (only one
data entry of some supported type) XOR "complex" (an entry composed of
many simple supported types). This would be less expressive than "a list
of element" where element can be "simple" OR "complex". In the latter
case complex elements can also be nested. Something more expressive than
this would be a "[structure] of element" where [structure] could be a
list, a tree, a graph or other of "complex" OR "simple" elements.
The table PATTERN / EXPLANATION in Thomas' blog post is a good start but
i think that it is mixing types with some familiar class names. These
can be broken down even further: The "data/state/protocol/reasoning" is
a Dictionary. The "History of events" is essentially the same but the
type of the "index" is now different. An "order state machine" is a
directed graph (with loops). A composition document is a Set and
participations could also be sets or dictionaries.
Could we have defined the COMPOSITION as a special case of a "set
archetype" with specific constraints? Could we have done the same for a
state machine?
Simply counting the supported patterns would not provide a comprehensive
picture though. I trust that terms are used consistently. So, a
mathematical model provides constraints over a domain. But there is
nothing stopping someone to create an overly complex model to fit
complex data with minimal error (reminds me of feature creep). You could
have two models, one with 5 parameters and one with 50 parameters that
seem to be fitting experimental data (observations from the domain)
perfectly, which one do you choose? Obviously, the one with the 5
parameters because it is _simpler_. Mathematics does have tools to
estimate model complexity versus how well the model describes the
observations and i am sure that some parallels can be drawn here. In
computer science, there are probably additional markers that need to be
taken into account...It's not only a matter of how expressive a model is
but also how "easy" it is to be used or how "cheap" it is to
implement...These are probably more difficult to quantify but key
factors for adopting a particular model.
In this way, the model(s) that rank first would be the easiest and
cheapest model(s) that can describe a domain most effectively.
We don't have a golden standard for a domain of course, but something
like this would be irrelevant. Even if we could pull all the workflows
and documents from all the healthcare systems in the world, they would
be a snapshot of what is required today** and with the exponential pace
of progress would quickly become surpassed. This means that the ranking
(without taking into account the "shape" of the domain) would only be
useful amongst two or more models.
All the best
Athanasios Anastasiou
**: But a very good indication of the patterns that are actually used
out there...maybe Google is already doing it.
Hi Athanasios,
I doubt if mathematically measuring model complexity would be a good way to determine utility of information models for use in archetype modelling, and therefore in e-health more generally. The only way to do that is to see which ones support more / better archetypes. In this sense, the information model is like a ‘theory’ - the idea is to get one that is as simple as possible for the most comprehensive explanatory power.
Hello Thomas
Thank you very much for your response.
One of the motives for what i am outlining in my last message has been
the recurring discussions in the list about the suitability of this or
that model (or approach) in e-health. So, an objective metric (a
relative or ideally, an absolute one) of the suitability of a model to
describe a domain could drive improvement and consensus.
It would be the quantification of the suitability of "...[a thoery] that
is as simple as possible for the most comprehensive explanatory power."
Anyway, i might put together something more specific about this and
share it with the list sometime later, although i admit that there are
still some gray areas especially when trying to provide some practical
examples on established models.
All the best
Athanasios Anastasiou
Hello Thomas Thank you very much for your response. One of the motives for what i am outlining in my last message has been the recurring discussions in the list about the suitability of this or that model (or approach) in e-health. So, an objective metric (a relative or ideally, an absolute one) of the suitability of a model to describe a domain could drive improvement and consensus.
certainly I agree with that - it is just that using a ‘white-box’ method is unlikely to work in my view; a ‘black-box’ method is more useful because it tests the information model(s) against their ability to support real world content models like archetypes. I am not saying that the openEHR model is perfect by any means, only that much of it was elaborated by direct archetype modelling; this provided far more feedback for changes than any standard IT modelling methods or metrics could ever ave done. As I think I mentioned in the blog post, one of the best tests for an information model in e-health is to see how you would model Glucose Tolerance Test based on it. With some it is exceedingly difficult, with some it is easy.