Yes, I agree we need a global, open terminology now, but realistically
it will take a decade of work to produce such a thing. In the meantime,
we need a usable, albeit imperfect terminology, at least the national
level.
AFAIK, SNOMED is owned by ACOP (American College of Pathologists), which
is a non-profit organisation.
My understanding is that the US open-licence period for SNOMED CT begins Jan 1, 04. I have also been told that Centers for Medicare and Medicaid Services are planning a number of "demonstration projects" early in 2004, involving EHR systems and SNOMED CT terms. Therefore, I would agree with Ed and others who seem to be suggesting that we move forward with whatever steps are necessary to get SNOMED CT into production systems.
In the context of a reasonably homogenous terminology space like the U.S., what would you see as the main problems with SNOMED that would have to be "fixed"? Pre-coordination issues have been mentioned along with some even more fundamental (??) issues that Thomas Beale suggests will take several years or possibly a decade to straighten out. Of course, from my point of view, my biggest problem will likely be scarcity of special vision care concepts/terms.
So, what would we have to do exactly? And which standards organization would coordinate such a project... HL7?
-Chris
Christopher J. Feahr, O.D.
Optiserv Consulting (Vision Industry)
http://Optiserv.com
http://VisionDataStandard.org
Office (707) 579-4984
Cell (707) 529-2268
Christopher Feahr wrote:
My understanding is that the US open-licence period for SNOMED CT begins Jan 1, 04. I have also been told that Centers for Medicare and Medicaid Services are planning a number of "demonstration projects" early in 2004, involving EHR systems and SNOMED CT terms. Therefore, I would agree with Ed and others who seem to be suggesting that we move forward with whatever steps are necessary to get SNOMED CT into production systems.
In the context of a reasonably homogenous terminology space like the U.S., what would you see as the main problems with SNOMED that would have to be "fixed"? Pre-coordination issues have been mentioned along with some even more fundamental (??) issues that Thomas Beale suggests will take several years or possibly a decade to straighten out. Of course, from my point of view, my biggest problem will likely be scarcity of special vision care concepts/terms.
So, what would we have to do exactly? And which standards organization would coordinate such a project... HL7?
my feeling (with apologies to those who have been working for years in terminology and I realise know the semantic space much better than I do) is that a new process could emerge:
a) some people develop some archetypes , e.g. American College of Opthalmologists (not sure what the proper title is)
b) these archetypes will have local vocabulary which defines meanings for exactly what terms need to mean in the exact context of the archetypes.
c) The mundane task of mappings to ICD or similar classifiers needed for reimbursement and various population & efficiency studies is easy. This will take care of the practical need for these codes.
d) mappings to ontologies are more challenging, and it may well be that local archetype terms form "capsule vocabularies" that could be the basis of change requests to developers of ontologies. Snomed is probably more in this space than pure terminology, so it may be that we send change requests of some kind to them, based on archetyps.
e) due to d), ontologies may change over time in such a way that more direct mappings from archetypes become possible.
- thomas beale
Dear all,
Thomas in particular to remind me to send the message to all, instead only to him 
Some possible additions to Thomas procedure:
a) some people develop some archetypes , e.g. American College of
Opthalmologists (not sure what the proper title is)
Yes, any professional organisation, or multidisciplinary group, or even patient organisations develop such things.
b
) these archetypes will have local vocabulary which defines meanings
for exactly what terms need to mean in the exact context of the archetypes.
Yes, but, if this is going to be constructed, it is wise to already look at well defined terminology and proceed as follows:
c
) The mundane task of mappings to ICD or similar classifiers needed for
reimbursement and various population & efficiency studies is easy. This
will take care of the practical need for these codes.
Given the earlier suggestion to already use such terms, a part of the coding has been done. Otherwise, depending on purpose (clinical trial is different from international prevalence study) map from scale / local term to classification.
d
) mappings to ontologies are more challenging, and it may well be that
local archetype terms form “capsule vocabularies” that could be the
basis of change requests to developers of ontologies.
Yes, a validated scale on a particular issue around human functioning could be part of an ontology, but perhaps not always. The Barthel index or the APGAR score e.g. have distinct and different variables that probably would not stand beside each other in an ontology. Or, it would be an ontology with many to many parent - child relationships.
Snomed is probably
more in this space than pure terminology, so it may be that we send
change requests of some kind to them, based on archetyps.
Yes, the process would become interactively with knowledge determiners (ontology, scales) terminology developers (semantics) and information modellers (archetypes as constraining mechanisms for what a record system / messaging system must do with this particular grouping of patient data.
e
) due to d), ontologies may change over time in such a way that more
direct mappings from archetypes become possible.
Yes, that might work two ways from archetype to ontology, but reverse to via including from ontology into archetypes.
Hope this helps,
Sincerely yours,
Dr. William T.F. Goossen
Senior Researcher and Consultant Health and Nursing Informatics
Acquest Research and Development, Koudekerk aan den Rijn, the Netherlands
http://www.acquest.nl/
&
Adjunct Associate Professor in the College of Nursing, faculty in the Organizations, Systems and Community Health Area of Study, the University of IOWA, Iowa City, Iowa, USA. www.nursing.uiowa.edu/NI
&
Country Representative for the Netherlands in the Special Interest Group Nursing Informatics, IMIA. http://www.infocom.cqu.edu.au/imia-ni/
&
Member Evaluation Committee International Classification for Nursing Practice, Geneva, ICN. International Council of Nurses http://www.icn.ch/ and http://www.icn.ch/icnp.htm
&
Associate Professor, Adjunct on the faculty of the School of Nursing,
University of Colorado Health Sciences Center, Denver, USA.
&
Bestuurslid Vereniging voor Medische en Biologische Informatieverwerking
http://www.vmbi.nl/
&
Fellow of the Centre for Health Informatics Research and Development (CHIRAD), School of Social Sciences, Kings Alfred’s, Winchester www.chirad.org.uk
Hi to all,
b
) these archetypes will have local vocabulary which defines meanings
for exactly what terms need to mean in the exact context of the archetypes.Yes, but, if this is going to be constructed, it is wise to already look at well defined terminology and proceed as follows:
- if a scientifically validated and reliable scale (mini ontology?) is available and meets the clinical need then choose that.
if this scales are not available, then use some kind of standardized terminology like LOINC, SNOMED, ICF, NANDA or whatever.
if that is not available to fullfil the needs for the subject, then choose your own wording that defines best the meanings for the clinical area.
It certainly makes sense.
From what I personnaly experienced, if you don't make the "semantisation" work at the very first stages, you will have a hard time doing it afterward because an ontology is an accurate but restricted langage ; thus you end up translating from natural langage to a more restricted langage - or worse from classification terms, triyng to give a meaning to terms that where only created for "patient grouping".
Making "semantic Archetypes" is a job that involves a doctor and a knowledge manager - this one translates between "doctor langage" and "ontological langage", but also asks for the proper level of accuracy in the concepts representation.
Very hard to do lately.
Given the earlier suggestion to already use such terms, a part of the coding has been done. Otherwise, depending on purpose (clinical trial is different from international prevalence study) map from scale / local term to classification.
We are currently experiencing such things ; it is not easy to have people understand the difference between description (As accurate as possible), local study (question 5 can be answered 5.1, 5.2...) and studies using classifications such as ICD or ICPC where you just can use concepts inside the classification (and it is sometimes complicated since, for example, "send to the hospital" as no entry inside ICPC).
I don't think you can expect adressing all these issues through Archetypes
Yes, a validated scale on a particular issue around human functioning could be part of an ontology, but perhaps not always. The Barthel index or the APGAR score e.g. have distinct and different variables that probably would not stand beside each other in an ontology. Or, it would be an ontology with many to many parent - child relationships.
The way we solve this kind of problem is that we incorporated inside the ontology concepts as "ICD10 code", "ICPC code" and so on. These ontology concepts are given the code as a "value" in the same way "patient size (cm)" would be given 180 as a value.
Philippe AMELINE wrote:
Hi to all,
We are currently experiencing such things ; it is not easy to have people understand the difference between description (As accurate as possible), local study (question 5 can be answered 5.1, 5.2...) and studies using classifications such as ICD or ICPC where you just can use concepts inside the classification (and it is sometimes complicated since, for example, "send to the hospital" as no entry inside ICPC).
I don't think you can expect adressing all these issues through Archetypes
I would not either...we just need some good oontologies...
Yes, a validated scale on a particular issue around human functioning could be part of an ontology, but perhaps not always. The Barthel index or the APGAR score e.g. have distinct and different variables that probably would not stand beside each other in an ontology. Or, it would be an ontology with many to many parent - child relationships.
The way we solve this kind of problem is that we incorporated inside the ontology concepts as "ICD10 code", "ICPC code" and so on. These ontology concepts are given the code as a "value" in the same way "patient size (cm)" would be given 180 as a value.
the ADL supports this more or less as well...
- thomas beale
Thomas,
Thank you... this is getting clearer. I know that there are several hundred coded terms and answer lists relating to eyeglass and contact lens products already in LOINC, and there may well be some eyeglass terms in SNOMED CT. Many of these LOINC terms, however, have incomplete/ambiguous definitions or do not fully support business messages like eyeglass purchase orders and insurance claims. So it seems that a vision industry committee could start fresh by creating ontology models (UML) for "eyeglass" and "contact lens" products, from which archetypes can be defined with ADL.
It would only seem necessary to map these archetypes to LOINC and/or SNOMED CT if there were business applications already using those terminology systems... and we wanted the old LOINC-based systems to be interoperable with the new archetype-based systems... right?
Finally, if a vision industry committee did successfully create 100 new archetypes to support doctors ordering eyeglasses from labs, would our committee be able to simply distribute the archetype-definition-list to the industry as one, big XML document? I guess, the other implied question is whether it would be necessary to set up one or more terminology servers... or could each lab management system and practice management system simply be shipped with a local copy of the current archetype list? ... with the ability to import newer versions (simple XML documents) later on?
Thanks,
-Chris
Dear All
This area is difficult and we must learn as we go. There are a few
conclusions I have come to from an EHR system point of view..
1. The data structures and term sets that are required for clinical care and
communication must be able to be instituted both prior to and after the
standardisation processes have been published.
2. Special requirements that are not contrary to agreed standards should be
able to be implemented without difficulty - this is the norm rather than the
exception.
3. Where terminologies required in archetypes are small and generally
agreed, these should be primarily expressed in the archetype itself - not to
do so is to add to the unrealistic demands on external terminologies.
4. Translations will be safest inside archetypes where the meaning is
clear - the context is highly specified. This is a reason to extend the role
of internal terminologies of archetypes.
So, the new statements I would make are:
1. Archetypes should have no language or terminology primacy - and these
should be able to be added post-hoc.
2. Terminologies internal to the archetype will always be safer to translate
and provide synonyms and specialisations.
Despite the feeling of some in the business, this does not really diminish
the need for external terminologies. I am also aware that the comprehensive
approach of Philippe and the Odyssey Project and the text processing of
Peter Elkin. I believe these efforts will remain as relevant, but more
focussed within an archetype driven information model.
Cheers, Sam Heard
Thomas had indicated that you were going to respond to the issues and
clarify the stand being evolved in this respect by EHR. Shall welcome your
comments.
Dr Grewal
Dr Grewal
Can you please outline the issues as the included email is from me...Sam
Thomas and Ed,
I just noticed this unsent message to you from 9-26 (included below) and I’d like to add an update. The vision care industry has launched a project in Open Applications Group called Vision Markup Language or VisionML. http://www.openapplications.org/projects/VisionML.htm describes our core mission and there is a short .ppt presentation on that same page under the Project Status from last update link. Anand Kumar and others will be assisting us with the ontology/terminology issues that I outline in the .ppt.
While VisionML is focused on the needs of the vision care industry, the public comment, vetting, and formal voting model that we propose, as well as OAGi’s basic development model will also be applicable to other specialty care domains. I believe that the logistics involved with bringing doctor-minds to bear on this problem are more challenging in some ways than the technical ontology modeling work. For this reason, we are also positioning VisionML as a demonstration project for providers in the ~30 other healthcare specialty domains.
OAGi and Oracle Cporporation are co-sponsoring our webcast on Dec. 9th to inform key vision industry stakeholders of the project and to make a pitch for sponsorship of VisionML through OAGi membership. Ed, I heard and agree with your comments at the IOM briefing last Thurs. regarding patient safety. I also agree with IOM that data standards and a common vocabulary are of utmost importance and that the work must be at least partially funded by the federal government. We hope to obtain CMS/HHS funding for the VisionML project early in 2004.
I would invite everyone on this email distribution list to register for our webcast at http://www.openapplications.org/webcast/VisionML/VisionML.htm
Best regards,
-C
Christopher J. Feahr, O.D.
Optiserv Consulting
VisionML Project Lead
Open Applications Group, Inc.
http://www.openapplications.org
Office (707) 579-4984
Cell (707) 529-2268
(9-26-03 unsent message)
Thomas,
I am inclined to agree with Ed (and others) who are suggesting that we at least attempt to develop a robust and nearly universal medical terminology for healthcare. Many believe that SNOMED and possibly UMLS are good places to start. CPT and ICD-type codes are barely adequate even for insurance transactions and completely inadequate for most health care processes. As long as perfection is not our immediate goal, I think we should move forward immediately into addressing the terminology component.
The most expensive part of this undertaking will be the vetting with a credible representation of the provider community. An efficient mechanism for managing the vetting will have to be designed and implemented first. I would be glad to discuss some infrastructure ideas if there is interest in this.
Regards,
-Chris