term list(s) vs. snomed -ct ® subset. 2 nd aaha software vendors summit – april 21, 2009
TRANSCRIPT
Term list(s) vs. SNOMED -CT® subset.Term list(s) vs. SNOMED -CT® subset.2nd AAHA Software Vendors Summit – April 21, 2009
Lists of words…Lists of words…• Nomenclature
• The system or set of names for things
• Vocabulary• A collection or list of words with explanations of their meanings
• Classification• The result of classifying; a systematic distribution, allocation,
or arrangement, in a class or classes; esp. of things which form the subject-matter of a science or of a methodic inquiry. (SNOMED)
Lists of words…Lists of words…• Terminologies are about information
sharing, retrieval, aggregation and analysis.• It’s difficult if not impossible to justify the
effort required to “do terminology right” from a data entry perspective.
• A functional terminology must attend to both perspectives.
What do we need?What do we need?• Nomenclature ONLY
• Provides a simple list for data entry
• Vocabulary / Classification• We can be CERTAIN that the “term” (description in
SNOMED) means what we think it means.• We can develop rules that allow us to combine concepts
to express ideas more complicated than those contained in the nomenclature.
• We can use the knowledge base supported by the vocabulary/classification to search, retrieve and analyze our data.
Why a controlled nomenclature?Why a controlled nomenclature?
• Aggregation of text-based content from multiple sources• Multiple individuals• Multiple institutions
• Any time you rely on a computer to manipulate language and “meaning” is critical.
Why a controlled vocabulary / nomenclature?Why a controlled vocabulary / nomenclature?
• Controlled vocabularies should automate recognition of and accurate substitution of synonyms.
• Controlled vocabularies should facilitate retrieval and categorization.
Complaints about SNOMEDComplaints about SNOMED• It’s too…
• Big• Complicated• Expensive
• Yes but…• We can make it smaller (sort of), and use small pieces (for
most purposes).• Use it in simple and straightforward ways• But nothing, it’s expensive.
• Not the license fees, the cost of making it work.
Why pick something as big and complicated as SNOMED?Why pick something as big and complicated as SNOMED?• Viable nomenclatures must be maintained.• SNOMED is the ONLY actively maintained
nomenclature that has veterinary content.• Veterinary medicine CANNOT afford:
• to build it’s own competent nomenclature• to continue to live without a competent
nomenclature
SNOMED historySNOMED history
Reduce storage size Reduce Storage size Storage not an issue
Categorize information Multiple code-based mono-hierarchies
Poly-hierarchical categorization
Functional Subsets
Pathology content “All Medicine” Veterinary content separate, then integrated
SNOVET DOES NOT EXIST
Integrated content
“Computability” for retrieval.
Natural language, artificial intelligence, decision support
SNOP
SNOMED I & II
SNOVET
SNOMED III
SNOMED RT
SNOMED CT
1965 2000
IHTSDO
Development historyDevelopment history• SNOP
• Morphologies• SNOMED
• Morphologies, Etiologies, Locations• SNOVET
• Same structure as SNOMED• Mix of existing SNOMED, additional veterinary content
• SNOMED III• Disorders, Morphology, Living organisms, social context• Veterinary content re-integrated
• SNOMED RT• Logic based approach to SNOMED. Axes became hierarchies. Most
significantly, the poly-hierarchic approach to classification.• SNOMED CT
• SNOMED RT on steroids. Post merger with CTv3.
What do we get ?What do we get ?• Sound technical solution to synonyms.
• Ability to “localize” the synonyms
• Compatibility with other “lists”• Ability to merge AAHA-based records with others
(e.g., a cardiology specialty subset) • Functional Sub-setting• Enhanced queries
Solution to synonymySolution to synonymy• Obvious duplicates in AAHA draft list:
• HYPERGLYCEMIA, BLOOD GLUCOSE INCREASED
• AAHA Category = Hematology, Lymphatic, Endocrine • HYPERGLYCEMIA, BLOOD GLUCOSE INCREASED
• AAHA Category = Metabolic
• (NOT) Obvious duplicates in AAHA draft list• Thyroid gland mass
• AAHA Category = Hematology, Lymphatic, Endocrine
• Mass, thyroid• AAHA Category = Neoplasm
• In SNOMED, both = 237557003 = Mass of thyroid gland (finding)
“Local” Synonyms“Local” Synonyms• It is POSSIBLE to allow practitioners to
add their own favorite description of a concept.• Analysis / transmission by conceptID.
Compatibility with other listsCompatibility with other lists• AAHA list can be part of “mixed animal”
system
• AAHA list would integrate (could be used to query) a more granular specialty list.
Functional Sub-settingFunctional Sub-setting
• We only need PORTIONS of SNOMED• DIFFERENT portions of SNOMED needed for
different contexts in HIS.• Retain the ability to use ALL of SNOMED to
search, retrieve, analyze data produced using sub-sets.
• Be prepared to transfer (copy) from SNOMED to subset as needs change.
FunctionalSubsetsFunctionalSubsets All of SNOMED
VetSubset
“Cardiovascular disease” subset Algorithm
Cardiovascular Diseases
Intersection = Veterinary Cardiovascular Diseases
SNOMED SubsetSNOMED Subset
• “…a set of Concepts, Descriptions, or Relationships that are appropriate to a particular language, dialect, country, specialty, organization, user or context.”
• “…simplest form, the Subset Mechanism is a list of SNOMED identifiers (SCTIDs).”
• “…may be used to derive tables that contain only part of SNOMED CT.”
• Can be selected by clever query, if underlying definitions in SNOMED are sound.
Existing Subset(s)Existing Subset(s)
• Non-human subset• This subset assists applications that desire to exclude
concepts which are not human medical concepts (i.e., paw and fin).
• Note that this is NOT a veterinary subset as that subset would include terms shared with humans such as brain and eye.
• Pathology subsets (3)• CAP Cancer checklists• Allergen subsets
SubsetsSubsetsAll veterinary content Root Veterinary Subset
(large)
100 k ?
Bacteria Living organism automated subset
8500 concepts
Abnormal Morphologies Body structure automated subset
4000 concepts
Respiratory Findings Findings/disorders automated subset
850
Severities Automated from qualifiers 5
AAHA Subset(s)AAHA Subset(s)• SNOMED then remove hierarchies that are
NOT of interest.• Someone has to decide what’s “not of interest”
• Someone familiar with SNOMED• Someone with domain knowledge
• Desired functionality• We think it’s important to distribute a subset of the
hierarchy above the AAHA subset with relationships– Facilitate retrieval queries, may be possible to use the hierarchy to
control “lists in correct context” (this does not currently exist.
Subset development (Ideal)Subset development (Ideal)
• Build a competent Veterinary Subset of SNOMED• Veterinary subset a resource shared by the
profession.• Managed by central “authority”• Distributed by SNOMED?
• Use algorithm approaches to create “microsubsets”
What we’re doing instead…What we’re doing instead…• Intellectual investment (by AAHA) in a list of terms
representing desirable small animal medical content.
• “Mapping” by VTSL and EHRTF• Add “missing” content through SNOMED
Extension mechanism.• AAHA terms expressed as SNOMED descriptions.• Permanent identifiers
Mapping (why we didn’t just map AAHA’s list).Mapping (why we didn’t just map AAHA’s list).
• Mapping is directional• Largely the result of differing granularity between “target”
and “source”• 1:1 – Concept is the same
– Term may be identical or synonym – remember to distinguish on CONCEPT not on string
• Narrow to Broad – Source concept is more specific than target
• Broad to Narrow – Source concept is more general than target
• Two maps may be needed for bi-directional functionality (unless entire map is 1:1)
MappingMapping
• 1:1 maps will represent a majority• Broad (source) to narrow (SNOMED)
• Good argument that SNOMED needs more content
• Narrow (source) to broad (SNOMED)• SNOMED may need/want the content• Map to a post-coordinated concept may be
required
AAHA terminology developmentAAHA terminology development
• There is no “final version”
• Walk don’t run• No syntax (post-coordination) just yet• Breadth first, depth later
SNOMED ExtensionsSNOMED Extensions• Enable authorized organizations (VTSL maintains two
namespaces) to add Concepts, Descriptions, Relationships and Subsets to complement those that are centrally maintained as the core content of SNOMED CT.
• specialized terminology needs of an organization.• ISIS / ZIMS• USDA• FDACVM
• Extensions maintain unique identification across organizations.
SNOMED ExtensionsSNOMED Extensions• Distinguishable from the main body of SNOMED CT
• in the thesaurus• when stored in a patient record, query or decision support
protocol.
• Distinguishable from other Extensions, in the same way as they are distinguishable from the main body of SNOMED CT.
• Able to be distributed and processed in the same way as equivalent components from the main body of SNOMED CT without requiring specific adaptations of SNOMED-enabled applications.
Existing Extension(s)Existing Extension(s)• US Drug extension
• List of drugs marketed in the United States• Veterinary drugs have not been maintained
in some time.
• UK Drug extension
What are we doing to the AAHA Diagnostic Terms list?What are we doing to the AAHA Diagnostic Terms list?
• Two reviews by VTSL veterinarians, third review by AAHA team.
• Determining what each term MEANS• Mapping each term to SNOMED• Editing the terms
• Slightly more natural English• Separating list of synonyms into individual descriptions• Limiting commas to one “use” only• Converting to sentence case
• Providing SNOMED identifiers for each description
AAHA Terms (version changes) AAHA Terms (version changes)
• AAHA terms will have SNOMED-based identifiers
• AAHA terms will be mapped to SNOMED concepts
• Phrasings more like natural English• Only one use of commas• “Within-term” synonyms will be separate
descriptions.
Future project(s)Future project(s)• Plan / build user request system
• Characterize AAHA content• Patient findings• Laboratory findings• Morphologies
• Add, then “clean up” upper hierarchy• Hierarchy to display "in appropriate context"
• (Liver things show up when vet wants liver things).
• Create similar specialty-based subsets• Increased specificity/granularity
• Cardiology, Neurology, etc.