=Paper=
{{Paper
|id=Vol-410/paper-10
|storemode=property
|title=Essential SNOMED: Simplifying SNOMED-CT and supporting Integration with Health Information Models
|pdfUrl=https://ceur-ws.org/Vol-410/Paper10.pdf
|volume=Vol-410
|dblpUrl=https://dblp.org/rec/conf/krmed/MacIsaacWRGEP08
}}
==Essential SNOMED: Simplifying SNOMED-CT and supporting Integration with Health Information Models==
https://ceur-ws.org/Vol-410/Paper10.pdf
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
Essential SNOMED: Simplifying SNOMED CT and Supporting Integration
with Health Information Models
Peter MacIsaac, MB.BS, FRACGP, MPH1, Don Walker, MB.BS 2 , Rachel Richesson PhD,
MPH3, Heather Grain FACHI4,Peter Elkin, MD5, Jon Patrick PhD6
1
Terminology Central, Canberra, Australia, peter@macisaacinformatics.org;
2
University of Adelaide, Adelaide, South Australia, donald.walker@adelaide.edu.au;
3
University of South Florida College of Medicine, Tampa, FL, USA, richesrl@epi.usf.edu;
4
LaTrobe University, Melbourne, Victoria, Australia, h.grain@latrobe.edu.au;
5
Mayo Clinic, Rochester, MN, USA, elkin.peter@mayo.edu
6
University of Sydney, NSW, Australia, jonpat@cs.usyd.edu.au
ABSTRACT
SNOMED CT (SCT) has been designed and INTRODUCTION
implemented in an era when health computer SNOMED CT (SCT) is becoming the international
systems generally required terminology standard clinical terminology with a new
representations in the form of singular pre- international licensing and governance process
coordinated concepts. Consequently, much of SCT which makes it widely accessible. The adoption of
content represents pre-coordinated concepts and SCT by multiple countries was influenced by many
their relationships. In this conceptual paper the published studies demonstrating its comprehensive
role of pre- and post-coordinated terminology coverage [1-4] and advanced structural features.
expressions are considered in the context of the SNOMED CT has antecedents in the College of
current development direction of Electronic Health American Pathologists family of terminologies, the
Records and the use of communications UK National Health Service Read Codes. As with
and knowledge repositories. The move from any living language, it has absorbed content from a
current SCT structures to an implementation number of other terminologies and classifications.
form of SCT that focuses on “atomic concepts” SCT contains concepts and terms that describe the
will support post-coordination and terminology “language of use” as well as concepts which define
binding to information models. This core or the “language of meaning”[5-7]. Consequently, SCT
“essential” SNOMED CT - called SNOMED contains many pre-coordinated concepts that have
Essential Terminology (S-ET) - would be varying levels of semantic complexity alongside the
smaller in terms of core concept numbers, component or essential concepts which are
simpler, easier to maintain and more intuitive themselves the building blocks of these complex
for implementers. Our proposed clinical expressions. While there are sound
implementation form of SNOMED CT would historical and ongoing pragmatic reasons for this
contain only “atomic concepts” with their evolutionary development, the resulting mix of
attendant hierarchies and relationship data. These concept structures makes implementation within
would be supported by a strict model for various information models complex and prone to
representing current and future pre-coordinated variation. Currently, SCT is “cluttered” with pre-
concepts based on the use of an existing coordinated terms that are incompletely defined by the
specific post- coordination expression, internal information model that exists within SCT,
grammar, or representation. The resulting making transformations between existing pre-
concept expressions would be post- coordinated terms and post- coordinated
coordinated from a smaller core of atomic representations difficult to achieve. This result
components. Using definitional relationships, the limits opportunities for interoperability across
proposed implementation form could equate systems, [8] which is one of the key objectives of a
existing pre-coordinated terms with post- controlled terminology.
coordinated representations, allowing SCT to
maintain links with legacy data. A strategy for This conceptual paper brings to notice issues that
testing and implementing this approach is
discussed and empirical research and feasibility
testing is recommended.
51
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
are well known within the health informatics coordinated concepts, and an implementation
community and proposes what may, at first glance, perspective. Simple examples have been selected,
seem to be radical surgery. This proposal is in not to prove the feasibility of this approach, but to
reality an extension and combination of existing illustrate the principles. The need for a more
features of SCT to create a more tractable solution technically challenging and quantitative approach
to support both SCT development and the art and to evaluation of this proposal is recognized and
science of terminology development. This paper is discussed.
not a report of a quantitative analysis of SCT
structures or experimental results of the types of TERMINOLOGIES AND INFORMATION
change proposed. These should come later, if the MODELS
fundamental proposition is believed to be sound It is now widely accepted that health information
and a potential contribution to terminology storage is achieved through a combination of the
development and maintenance methods. use of controlled terminologies and standardized
data models or architecture, yet the boundaries
The computational representation of data is a between the models used for terminology
combination of the use of information models construction and health record construction are
and terminology. We propose a variation, blurred. [9-12] The HL7 TermInfo project
restructure and extension o f t h e c u r r e n t attempted to resolve this by providing guidance
SNOMED-CT terminology to support on how SNOMED CT could be used in HL7
implementation in various information models. version 3 messages and data structures. [12-14]
Using a pragmatic approach the SCT would be An example of this terminology model -
altered in that existing pre-coordinated information model interface is the question of
concepts would be identified, flagged and then whether concept negation should be managed
defined through linkage to their atomic within the terminology or within the data model.
concepts and relationship types. The atomic or Should the negation be expressed as part of the
essential concepts would continue to be terminological unit: “no history of breast cancer”,
placed in logical and definitional hierarchies or as different components within an information
and relationship structures and subject to the model: “history of breast cancer” + “negative”?.
use of description logic for definition, [15] The semantics can be represented in the
inference, and classification purposes. terminology as a pre or post coordinated concept or
Existing pre-coordinated SCT concepts would in a combination of the data model and
retain their identifiers and be linked to the modified terminology. The machinery to support this latter
terminology as “pre-defined-post-coordinated approach is contained in standard information
concepts”, and would be logically equivalent to any models such as the HL7 Reference Information
post-coordination representing the same meaning. Model (RIM).[12] HL7’s TermInfo working group
The retention of pre-coordinated concepts and the has recommended that when SNOMED is being
specification of their computational definitions used in HL7 V3 models, negation be managed in
would allow pre-coordinated terms to be used in the terminology and that the model based approach
interface applications, as pre-coordinated terms to attaching a negation indicator be deprecated.
can be useful in helping data entry to be This issue points to the need for sufficient
more consistent: supporting the language of flexibility in the management of post-coordination
use. If users have a retrieval list of pre- to allow for the transformation of concept
coordinated concepts that have post-coordinated structures and modifiers between the various
equivalents, application developers can encourage options. The existence of other data and
users to use a more consistent post-coordinated information models (e.g., CDISC) – which might
form or to use entry terms that have relationships develop and endorse their own guidance for use
to post-coordinated expressions using fully- of complex terminologies such as SCT -
defined atomic concepts. suggest that standardization of SCT terminology
use in HL7 (RIM-based) applications might not
This approach to the re-organization of SCT with guarantee interoperability with applications
the formal expression of the canonical form for using other information models. [16]
pre-coordination is described as “ S N O M E D
Essential Terminology” (S-ET), or While the issue of terminology and information
simply “Essential SNOMED”; the name model interaction is somewhat independent of the
coined by Dr. Walker when first describing way that coordination of complex concepts occurs,
this approach. This paper describes the case there is a need for both pre and post-coordinated
for change in SCT representation and advantages approaches to co-exist to fully support the spectrum
of moving to this representation, the of information representation. It is also recognized
background to the development of this that equivalence between pre-coordinated and post-
approach, a representation model for pre- coordinated concepts has to be established to
52
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
maintain consistency in interpretation of types were accommodated. The Essential
terminology and between concept representation SNOMED notion, which was initially
using different combinations of terminology and canvassed informally within the health
information model binding. In the current SCT terminology community in 2001, was further
infrastructure this is achieved using computation developed following a comparative technical
and testing the equivalence of the canonical form of analysis of several terminology options that were
the two terminological variations. This requires then being investigated for use in Australian
that all of the atomic or component concepts General Practice [23] and which subsequently
and pre- coordinated concepts are fully defined recommended use of SCT leading in time to
- not the case in practice. Having a formal Australia becoming an early adopter of a national
definition explicitly developed for current and SCT license. A review of candidate
future pre- coordinated concepts within terminologies at that time for use in General
SNOMED would support the recognition of Practice examined several options. One terminology,
equivalence. [8, 17, 18] DOCLE, was constructed of atomic concepts,
joined by operators using a Bachus Naur Form
ISSUES WITH SNOMED-CT (BNF), a standard system for representation of
IMPLEMENTATION computable expressions using syntax or rules.
Several studies have shown that inter-rater [24] What was notable was the extensive use of
reliability of SCT coding is poor, at least in part pre-coordinated terms that were constructed from
due to the complexity of the SCT structure and the atomic elements. For example “cancer@breast”
inconsistency of existing content. [3, 19-21] This was a pre-coordinated concept for “breast cancer”,
paper proposes that a simpler, more consistent yet it is constructed using a post-coordination
representation of SCT will reduce confusion and model of atomic concepts and the location operator
improve the quality of SCT implementation. .“@”. The process of normalization of DOCLE
This would need to be tested once working for inclusion in a terminology service found that
subsets of the S-ET have been developed and so a number of atomic concepts needed to be
examine the impact on coding consistency of the created to support existing content. Considering
interaction between the information model and this approach and drawing on prior experience
the use of differently coordinated terminology. with the development of a medicines
SCT size will certainly grow as new countries terminology requiring a full set of atomic
adopt it, especially when it becomes the elements which were combined to create fully
terminology to support the many uses of coded defined pre-coordinated medicines concepts,
clinical data, such as public heath. Trying to it was postulated that the SCT terminology
keep up with the need for language of use could be significantly simplified by creating a
through definition of pre-coordinated concept separate data structure for the pre-coordinated
phrases is a recipe for “combinatorial explosion” concepts where these were parsed and then
in the size of a terminology. This is bad enough described in a post- coordination grammar. [25]
in a terminology of simple structure, yet in one
of SCTs complexity and richness of function, DESIGN OF ESSENTIAL SNOMED
the impact is especially significant. A key Essential SNOMED would contain a complete set
technical challenge involves keeping the of “atomic concepts” from which all other concepts
terminology to a manageable size and level of could be constructed by post-coordination. These
complexity so that it is both maintainable and atomic concepts would be carefully crafted into
supports end users’ applications. A second their hierarchies and defined by their relationships.
challenge for SCT maintenance is to allow SNOMED CT most likely contains many - if
compatibility with historical versions used by not most - of these atomic concepts. They
legacy applications while maintaining would consist of both primitive and fully
relevance as the core terminology resource defined concepts. The large number of pre-
for the current and future generations of health coordinated concepts that are not in the above
information systems. The model proposed in group should be “flagged” in the complete
this paper will support both of these SNOMED CT data structure as “predefined-post-
objectives. coordinated equivalent concepts”, and eventually
associated with their post-coordinated defining
BACKGROUND TO S-ET DEVELOPMENT atomic concepts using a formal post
In 1999 a combined pre and post-coordinated coordination or compositional syntax as is
model for a medicines terminology was proposed already described. This group of pre-
by two of the authors (DW and PM) for coordinated concepts would not rely on their
Australia, based on an architecture designed hierarchical position or SNOMED relationships for
earlier by DW for a proprietary drug information their definition - instead they would be defined by
service. Both pre-coordinated concepts and their the compositional expression (or formalism) used
contained atomic components and relationship to construct their atomic post-coordinated concepts
53
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
(combination of existing atomic concepts), as The advantages of this proposed structure for SCT
described above. All the existing pre- are reduced size and complexity for ease of
coordinated concepts in SCT could remain, along implementation and maintenance. An inevitable
with their identifier for use where the situation outcome would be a reduction in the
required this approach. Pre-coordinated concepts combinatorial explosion that occurs when rampant
being added to SCT would also follow this pattern. pre-coordination of concepts and phrases occurs,
yet this comes at the cost of introducing a new
For example, the pre-coordinated concept element in the post coordination expression that
“Cellulitis of the left foot with osteomyelitis of the links the pre-coordinated concepts to their atomic
third metatarsal without lymphangitis” can be elements. The core terminology concepts and
expressed using atomic concepts and relationship hierarchies should be however much simplified..
types, is shown in Table 1. The concepts and The core of S-ET would grow some as new
relationship concepts that comprise the definition atomic concepts were added. The S-ET structure
would all be considered core Essential SNOMED would be expressively intuitive as its approach to
content. concept r e p r e s e n t a t i o n w o u l d s u p p o r t
Oper- Disorder Has- Has- concept constructions. Hierarchical
ator FindingSite Laterality simplification would result as the definition of
Cellulitis Foot Left the many pre-coordinated-concepts would be
independent of immediate hierarchies or
and Osteomyelitis third
metatarsal
relationships – S-ET would use the compositional
expression to link with hierarchies and defining
Without Lymphangitis
relationships of the atomic concepts. Existing
approaches to canonical forms would continue and
Table 1 – Definitional relationships of an existing pre- allow equivalence testing between different pre-
coordinated SCT concept. coordinated concepts and post-coordinated
The current SNOMED CT terminology model expressions. Pre-coordinated concepts would still
specifies relationships between concepts and terms, be able to be represented in a hierarchical
but does not make a distinction between post- arrangement to support inference and subsumption,
coordinated concepts expressions and pre- however these could be calculated rather than
coordinated concepts. We propose that this explicit expressed as happens currently in SCT. In
distinction be made explicitly, as a tool to assist in this model the hierarchical relationships would be
SNOMED-CT terminology maintenance and inferred rather than the canonical form.
implementation. Figure 1 describes the way
that the new architectural elements could be linked Equating pre- and post-coordination may be easier,
with existing S- CT structures which are as the computational form is actually specified for
represented by the three elements placed at the concepts within SCT. It is acknowledged that the
right hand side of the figure. current approach in SNOMED is not
comprehensive due to incomplete set of canonical
representations and possible lack of semantics to
fully describe the meaning of existing
semantically complex pre-coordinated concepts.
Both the pre-coordinated form and the various
representations of the post-coordinated concept
are valid ways of describing the same concept.
The first is more aligned with human
interpretation and the second supports computer
processing of the data. It is clear that both forms of
concept representation are needed and both have to
be supported by clinical terminologies such as
SNOMED CT. The approach recommended in
SNOMED Essential Terminology is
believed to be consistent with the current SCT
approach to canonical form definition.
Figure 1 – Conceptual terminology model for Essential
SNOMED. This model is highly dependent upon an
expressive and computable syntax for post-
DISCUSSION coordination. The process of moving to an S-ET
distribution format will highlight any
At the outset it is acknowledged that this deficiencies in the current post-coordination
proposal is grounded in the excellent overall methods and constraints as these will become
design and management features of SCT. explicit and subject to development. SNOMED –
54
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
CT authors can continue to develop pre- will meet the widely accepted need to continue to
coordinated expressions if required. End users, manage post-coordination in a modern
particularly those who rely on the use of pre- terminology to support the computer-human
coordinated concepts, will have the capacity to add interface.
locally relevant pre-coordination through a minor
modification of the SCT way of managing local Making the transformation to S-ET
extensions, and in doing so would not require The transformation to a S N O M E D E s s e n t i a l
frequent change submissions to the core essential Terminology would require a set of suitable
SNOMED terminology. As discussed earlier an “relationship-types” and an appropriate post-
S-ET model, coupled with an improved model coordination representation form or “syntax” that
for managing pre-coordination will support the catered for the “pre-defined-postcoordinated
terminology user interface. equivalent concepts”. SNOMED has published a
BNF for this syntax. This syntax describes
One of the difficulties faced by SNOMED is the the core SCT concepts, and their
need to harmonize with widely used terminologies relationships. An XML equivalent (in addition
that are heavily structured on pre-coordination. or as an alternative) may be helpful for the current
LOINC and MEDCIN and most health computer engineering environment. This paper
classifications would be examples {28]. A is not exploring the relative merits of these
S N O M E D E s s e n t i a l Terminology would approaches; however the process of defining
not need to include the pre- coordinated the post-coordination equivalents of existing
concepts imported from such concepts will also provide a validity check on the
terminologies, and could instead relate completeness of the syntax or post coordination
their concepts to SNOMED-CT by model, and as such is complementary to activities
mapping which used the post-coordination of the International Health Terminology Standards
syntax. Alternatively such pre-coordinated Development Organisation’s (IHTSDO) Concept
concepts could be placed along with Model Special Interest Group.
existing SCT pre-coordinated content. The
end result would provide the flexibility of As the “pre-defined-post-coordinated concepts”
incorporating or mapping to external could be related back to their atomic components
terminologies, even though they may not (which are themselves part of the SNOMED
share the same data models as SCT. hierarchy and relationship structure) it would no
longer be necessary to separately define the
It is recognized that there are situations where pre- hierarchies or associations for the pre-coordinated
coordination is more efficient from a computational concepts within the terminology. This does not
perspective, as in the recognition of commonly preclude such constructs being employed, much
used text strings in natural language processing like current indexing activity at run-time. These
(NLP) applications. S N O M E D E s s e n t i a l hierarchies could be machine classified. For
Terminology will allow the further development example, if the phrase “fractured ankle” was
of such concepts without undue concern about the compiled from two concepts as follows:
combinatorial explosion that might otherwise
exist. NLP requires the consistent application of [problem, action or issue] = “fracture”
terminology and parsing of text. If a S N O M E D [which has FindingSite] = “ankle”
E s s e n t i a l Terminology model is not adopted
then it is likely that some equivalent derivative Consequently , if it was necessary to locate
product will be created by necessity by these key “injuries of the lower limb”, then the hierarchical
application areas. Having a standard form will ancestors of “fracture” would include “injuries”
support consistency of output of different NLP and those of “ankle” would include “lower limb”.
applications.
The issue of what is and what isn’t an atomic or
One of the current strategies to simplify pre-coordinated concept is subject to debate and the
SNOMED is to restructure the relationship between boundaries can be fuzzy. Is headache a single
terminologies and classifications. Removing or concept or a post-coordinated ‘pain’ with
retiring classification concepts from SCT will allow ‘location’ of ‘head’? Technically, it should
them to reside in their respective not be part of an S-ET based on atomic
classifications and have linkage to the clinical concepts but is it sufficiently common and
terminology by mapping or other formal semantically ‘simple’ enough to warrant
constructs. SNOMED Essential inclusion? Under our proposal, “atoms refer
Terminology proposes making a similar change to semantic units, not term labels or
to manage both the historical terminological compound term labels. While it is clear that
clutter resulting from SNOMED’s antecedents and even single concepts can have compound names, it
use in legacy information systems. In addition it would be a conceptual error to consider a concept
55
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
such as Hodgkin’s lymphoma to be a pre- As a result many of the pre-coordinated concepts
coordinated concept, whereas a “fractured right found in SCT diagnoses, findings and procedures
femur” is patently so. would be excluded from the atomic- concept list
and be placed in the pre-coordinated group.
Building essential SNOMED will be
necessarily a pragmatic exercise which can cope It is clear that any change to the structure or
with one or other of these forms or both. The representation forms of SCT may have an impact
consequences of the fuzziness in determining on reference set (subset) development, use within
whether an existing concept should be managed as value-sets, and mapping to classifications and use in
a pre or post coordinated concepts are not local extensions. These areas need to be further
expected to be significant. All are still included in examined, however S-ET would not have a
SCT. significant impact, as the current SCT and S-ET
would contain the same concepts and
New pre-coordinated concepts could be created relationships. There is a significant advantage for
(if r e q u i r e d b y i n f o r m a t i o n s y s t e m s o r local extensions as local terminology experts could
us er preference), although this temptation may map new local concepts to atomic elements within
best be resisted as it is expected that the SCT, hence gaining the benefits of
requirements for pre-coordination would become classification and relationship modelling,
less pressing with the introduction of standard without having to wait for formal inclusion in
information models (e.g., HL7 V3, archetypes or later releases. The development of reference sets
OpenEHR) and the advancement of Natural based on concept and hierarchy selection would
Language Processing (NLP) to support data also include related pre-coordinated concepts.
entry.[26] Complex pre-coordinated concepts can
The feasibility of remodelling large sections of
sometimes be useful in encouraging consistency in
SNOMED CT, particularly when there are
representation where small nuances may be
competing priorities for terminology development,
unintentionally instantiated where no difference in
must be assessed. While a conversion strategy has
clinical meaning exists. For example: Colon
not been covered in detail, the re-organization could
cancer can be represented either as: Malignant
consider using current SCT relationships - but with
Neoplasm - Has finding site – Colon; or as:
some care because of their known limitations. The
Colon – HasSpecimen - Malignant Neoplasm.
size of the term string, the number of individual
This ambiguity is undesirable, and the
words, the presence of relationships, and a
availability of pre-coordinated concept
comparison with lists of terms extracted from
expressions at the interface level can prevent
medical dictionaries might help identify potential
this type of variation, and set patterns for
pre-coordinated concepts. It is possible
good practice in post- coordination within
that a functional result to create S-ET
terminology services.
could result from flagging pre-
coordinated concepts and terms, without
The atomic-concepts included in S-ET would
substantially altering the publication
be those that are necessary and appropriate to build
structure. As with most terminology
pre-coordinated concepts that currently exist, or
development, specific tools to manage the
may be added subsequently, as well as the atomic
transition to S-ET would need to be developed,
concepts currently in use. The boundaries around
refined and the end result would need
atomic concept definition are often fuzzy as
appropriate checking and quality control
discussed above. Editorial rules would be
processes and upfront attention to ongoing
required to consider inclusion of concepts that are
maintenance.
not “semantically atomic” but are very common.
A pragmatic approach would need to be While the first efforts at instantiation of the S-ET
developed, and the following may suggest one model may involve the restructure of arbitrary
strategy: twigs and branches of the SNOMED hierarchical
1. The entries expected to be found as tree, an approach proposed would be to operate on
defined concepts in a large medical concepts identified in large sub-setting exercises
dictionary [27]; this would likely include items where terminology of use has been identified from
that have a distinct clinical meaning and are analysis of actual clinical use in a specific domain
used frequently – e.g. Lung cancer; breast such as intensive care [29] or general practice .[23]
cancer; direct inguinal hernia; chest pain.
2. Those concepts that cannot be adequately CONCLUSION
defined by the composition of their post- This paper has proposed a modest alteration to the
coordinated concepts due perhaps to use of structure of SNOMED CT so that it supports the
an uncommon or unsupported semantic type co-existence of pre and post coordination in a form
for the relationship between elements. that advances the basic structure of what might be
regarded as good terminology practice [30].
56
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
The changes do not require any fundamental in Ontologies (AOW 2006), Tasmania,
changes in SCT methods, but rather a structural Conferences in Research & Practice in
extension and the incorporation of existing post- Information Technology, 72, pp61-68, 2006
coordination methods of expression into the core 6. Rector, AL, Qamar R, Marley T. Binding
terminology. Ontologies and Coding Systems to
Electronic Health Records and Messages.
KR-MED 2006 – Biomedical Ontology in
This paper outlines a number of issues with the Action. November 8, 2006, Baltimore,
current SCT architecture and proposes a solution Maryland, USA
which is consistent with its current design and 7. Elkin PL, Brown SH, Lincoln MJ, Hogarth M,
which may have a number of advantages. If the Rector A. A formal representation for
proposed model creates resonance with the end messages containing compositional
users of SNOMED CT, it should be exposed to expressions. Int J Med Inform 2003 Sep;71(2-
empirical testing and considered by the IHTSDO 3):89-102.
and their related organizations. The authors hope 8. Andrews,J.E.,et al.,Comparing
this paper stimulates discussion and feedback. We Heterogeneous SNOMED CT Coding of
look forward to formal testing of these ideas for Clinical Research Concepts by Examining
Normalized Expressions. Journal of
feasibility and acceptance.
Biomedical Informatics, 2008. in press.
9. Rector, A.L. The Interf ac e Betwee n
ACKNOWLEDGMENTS Information, Terminology, and Inference
This paper was inspired by an original paper authored by Models. in Tenth World Conference on
Don Walker with contributions from Peter MacIsaac Medical and Health Informatics: MedInfo-
arising from the “GP Vocabulary Project” supported by 2001. 2001. London.
the Australian Department of Health and Ageing. 10. Dampney, C.N.G., G. Pegler, and M. Johnson.
http://www.adelaide.edu.au/health/gp/research/current/in Harmonising Health Information Models - A
formatics/#vocab). Critical Analysis of Current Practice. in Ninth
The authors would like to acknowledge the National Health Informatics Conference. 2001.
contributions of Tad McKeon, anonymous KRMED Canberra ACT, Australia.
reviewers, and numerous colleagues who have 11. Huff, S. and J. Carter. A Characterization of
contributed in some way to this paper and our Terminology Models, Clinical Templates,
education and research on terminology and informatics. Message Models, and Other Kinds of Clinical
Information Models. in AMIA Symposium.
This work has been supported in part by grants from 2000.
the Centers for Disease Control and Prevention 12. Chute, C.G., Medical Concept Representation,
(PH00022 and HK00014; PLE). i n M e di c al I nf or m at i c s . K no w l e d ge
Management and Data Mining in
Biomedicine., Chen H., et al., Editors. 2005,
Springer U.S. p. 163-182.
REFERENCES 13. Markwell, D., Meaning Well & Well meaning -
HL7 TermInfo. 2005, The Clinical Information
1. Wasserman, H. and J. Wang, An applied
Consultancy Ltd. www.clininfo.co.uk
evaluation of SNOMED CT as a clinical
14. HL7, Health Level Seven. 2005, Health Level
vocabulary for the computerized diagnosis
Seven, Inc.
and problem list. AMIA Ann Symp Proc,
15. Elkin, P.L., et al., A controlled trial of
2003: p. 699-703.
automated classification of negation from
2. Elkin, P.L., et al., Evaluation of the clinical notes. BMC Med Inform Decis Mak,
Content Coverage of SNOMED CT: 2005. 5(1): p. 13.
Ability of SNOMED Clinical Terms to 16. 16. Richesson, R.L. and J.P. Krischer,
Represent Clinical Problem Lists. Mayo Clinic Data Standards in Clinical Research:
Proceedings, 2006. 81(6): p. 741-748. Gaps, Overlaps, Challenges and Future
3. Warren, J.J. and R.P. Wilson. Representing Directions. Journal of the American Medical
Cardiovascular Concepts in an Electronic Informatics Association, 2007. 14(6): p. 687-
Health Record Using SNOMED CT®. 696.
in American Medical Informatics 17. Spackman, K.A. and K.E. Campbell,
Association Annual Symposium. 2006. Compositional concept representation using
Washington, D.C. SNOMED: towards further convergence of
4. Richesson, R.L., J.E. Andrews, and J.P. clinical terminologies. Proc AMIA Symp,
Krischer, Use of SNOMED CT to Represent 1998: p. 740-4.
Clinical Research Data: A Semantic 18. Rector, A.L., et al., The GRAIL Concept
Characterization of Data Items on Case Report Modelling Language for Medical Terminology.
Forms in Vasculitis Research Journal of the Artificial Intelligence in Medicine 1997. 9: p.
American Medical Informatics 139-171
Association, 2006. 13: p. 536-546. 19. Rothschild, A.S., H.P. Lehmann, and G.
5. Patrick, J. Metonymic and Holonymic Roles Hripcsak, Inter-rater Agreement in Physician-
and Emergent Properties in the SNOMED coded Problem List, in American Medical
CT Ontology, Advances in Ontologies, Informatics Association. 2005: Washington,
M.H. Orgun & Tf, Meyer (Eds). Advances
57
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
D.C.
20. Burkhart, L., et al., Mapping parish nurse
documentation into the nursing interventions
classification: a research method. Comput
Inform Nurs, 2005. 23(4): p. 220-9.
21. Chiang, M.F., et al. Reliability of SNOMED-
CT Coding by Three Physicians using Two
Terminology Browsers. in American
Medical Informatics Association Annual
Symposium. 2006. Washington, D.C.
22. Andrews, J.E., R.L. Richesson, and J.P.
Krischer, Variation of SNOMED CT Coding
of Clinical Research Concepts among
Coding Experts. JAMIA, 2007. 14: p. 497-
506.
23. Walker, D. and e. al, General Practice
Vocabulary Project. 2001, Commonwealth
Department of Health and Ageing.
24. Welcome to DOCLE Systems. 2008, .
http://www.docle.com.au/
25. Walker, D. and e. al., Essential SNOMED,
unpublished project report, University of
Adelaide.
26. Elkin, P.E., et al., An Evaluation of the Content
Coverage of SNOMED CT for Clinical
Problem Lists. Mayo Clin Proc, 2006. 81(6):
p. 741-8.
27. Elkin, P.L., et al., Guideline and Quality
Indicators for Development, Purchase and
Use of Controlled Health Vocabularies.
International Journal of Medical
Informatics, 2002. 68(1-3): p. 175-186.
28. Rosenbloom ST et al. Using SNOMED CT
to Represent Two Interface
Terminologies In review
29. Patrick J, Herkes R, Ryan A. Enhancement
technologies for clinical information
systems. HISA NSW conference
proceedings 2008.
30. Cimino JJ. Desiderata for controlled medical
vocabularies in the twenty-first century.
Methods Inf Med 1998;37(4–5):394–403
58