Lord of the Terms

FHIR Meetup March 3^rd 2022

Grahame Grieve

Lord of the Terms

• Your Challenge:
• Everyone uses terminology
• Entry level cost = 0 (enumerations)
• Complexity grows and grows (Ontologies, SNOMED CT)
• Cost is infinite?
• You have to deal with terminologies well
• Are you a Lord of the Terms?
• Or will the terminologies master you?

Forging your Terminology

• A well-made terminology lasts�through the ages
• Invest in your treasure, put your �power in there
• Cimino’s Desiderata

Cimino's Desiderata

• Domain completeness – cover your scope with depth and breadth (Use a grammar)
• Unambiguous – define your terms with a clear concept, and close the definition
• Non-redundant – don’t have more than one way to describe a concept
• Synonymy – clearly document terms that have the same meaning
• Multiple classification – allow multiple axes in your hierarchy
• (don’t reflect heirarchy in your codes)
• Consistency of views – use a common approach ruthlessly
• Explicit relationships – make sure you’re explicit about how your concepts relate to each other

​

​

http://www.cs.man.ac.uk/~jeremy/HealthInf/RCSEd/terminology-desiderata.htm

Forging your Terminology

• A well-made terminology lasts�through the ages
• Invest in your treasure, put your �power in there
• Cimino’s Desiderata
• Consider how you’ll manage and�implement/distribute it

CTS2 - Common Terminology Service

• The original terminology service
• Focus: Managing and creating terminologies
• Fully engaged with the complexity of terminologies

​

• FHIR does not cover this
• Just publishing and using terminologies
• So how are you going to do it
• how’s HL7 going to do it?

FHIR Terminology Approach

​

​

• The whole process matters
• All parts should be in harmony with each other
• Done well, terminology can save your empire

Terminology Harmony

Terminology Sub-system

• SNOMED CT / LOINC / RxNORM
• HGVS, ICPC, MIMS + 100s more
• ICD-X+
• ANZSCO, METEOR
• A drug formulary
• A config table in an application
• A list of enums in a java class
• US state codes

​

​

Code System:

Defines a set of concepts with a coherent meaning��Code�Display

Definition

CodeSystem Resource

• Define Concepts
• Codes
• Displays
• Properties
• May be small (2 codes) or large (500k+ codes)
• Some defined by implication (e.g. SNOMED CT)

​

10

Terminology Sub-system

Value Set:

A selection of a set of codes for use in a particular context

Code System:

Defines a set of concepts with a coherent meaning��Code�Display

Definition

Selects

ValueSet Resource

• A set of codes
• Include codes from other code system (e.g. LOINC)
• List the codes
• Choose a set of codes by rules
• Used to define conformance rules, UI functionality, system capabilities
• Functional service built around this

Code System vs Value Set

• Often mixed in common usage
• E.g. am application table that mixes LOINC codes and custom (self defined) codes
• The only way to know it’s a LOINC code is it has the distinctive NNN-N syntax
• HL7 says: keep your definitions clean, or you’ll get in trouble when you exchange data
• Separate the definition and use of a ‘concept’

Terminology Sub-system

Code System:

Defines a set of concepts with a coherent meaning��Code�Display

Definition

Element Definition: �Type and Value set reference

Value Set:

A selection of a set of codes for use in a particular context

Selects

Binds

Element Definition

Terminology Sub-system

Code System:

Defines a set of concepts with a coherent meaning��Code�Display

Definition

Element Definition: �Type and Value set reference

Value Set:

A selection of a set of codes for use in a particular context

Selects

Binds

Element: �code/�Coding/�CodeableConcept

Refers to

Conforms

Coding

Each “use of a code” (a reference into a code system) has 4 properties:

• system: URL of the code system
• version: stated version of the code �system (optional)
• code: the symbol defined for the �concept (code/expression)
• display: a human readable representation �of the concept (optional – debugging/display)

Examples

• Code: "status" : "confirmed"
• Coding: {� "system": "http://www.nlm.nih.gov/research/umls/rxnorm",� "code": "C3214954",� "display": "cashew nut allergenic extract Injectable"�}
• CodeableConcept: {� "coding": [{ � "system": "http://snomed.info/sct",� "code": "39579001",� "display": "Anaphylactic reaction"� }],� "text" : "Anaphylaxis"�}

Binding Strength

Expansions

• Lots of power & complexity in the rules for what’s in a value set
• Late binding of versions
• Hard to compute expressions (ECL)
• Modular code systems
• Human Language variability
• It’s intended for specialist software
• Most software ‘just wants to get the list of actual codes’
• That’s the ‘expansion’ of the value set – the outcome of applying the rules against actual code systems and local context

Concept Map

• A set of mappings from one code system to another
• Defined for use in a known context
• Sometimes take advantage of local rules
• Many to Many mapping
• Mappings may depend on properties or produce multiple properties
• Rules for unmapped codes

Terminology Subsystem

Value Set

• Enumerated code lists
• Code System filters
• Expansions

Functions

• UI Lookup
• Validation
• Subsumption tests

ConceptMap

• Mapping between terminologies
• Mapping between models

​

Functions

• Translations
• Closure Table maintenance

Terminology Server: Expand

• What codes are in this value set?
• E.g. populate a combobox
• Answer can be really complicated to determine
• Pass:
• Identity of valueset
• Details of context, text filter
• What kind of response wanted
• Get back:
• List of codes
• Possibly heirarchical

Terminology Server: Validate

• Is this code valid in this value set?
• E.g. check a code
• Decide whether to take an action
• Pass:
• Identity of valueset
• Details of context + code
• What kind of checking wanted
• Get back:
• True | false + errors description
• Display in language of choice

​

Terminology Server: Subsumes

• Is this code subsumed by that code
• E.g. decision making (simpler case than the previous)
• Answer can be really complicated to determine
• Pass:
• Identity of valueset
• Two codes
• Get back:
• True | false | unknown
• Note: codes do not need to be in the same code system

​

Terminology Server: Translate

• Can I get an equivalent for this code in some other code system
• E.g. fill out a form / message
• Prepare for a decision making service
• Pass:
• Code, Identity of target code system
• Details of context, maybe a concept map reference
• Get back:
• 0 or matching codes
• Details about the matching

​

Terminology Server: Closure

• Support SQL joins.
• Which (resources in set) have a code that (meets criteria)
• Terminology server knows ‘meets criteria’
• SQL Server has to evaluate the joins
• Build a ‘closure table’ incrementally. (why?)
• Pass:
• Identity of closure table
• Code added to the closure table
• Get back:
• List of new relationships between codes
• Add to your sql table, and join across it

​

​

Using a terminology service

• Just need to know value set URLs (and versions)
• And code system and concept map URLs
• And you need to know Codings (system + version + code + display)
• Let the terminology system deal with anything else
• You don’t have to write a sub-system for managing them
• You don’t have to ‘import’ them
• You don’t have to have a specialist team

​

​

​

Terminology Service

• Just make it simple
• Let the weirdo experts �handle terminology details
• App writers write great�apps instead
• There are good servers �available today – use one!

Lord of the Terms

• Your terminology usage tells p�eople about your values
• Invest in good terminology�management
• Use a terminology server
• You too can rule the known �universe of health IT

​

​