Use Cases for Structured Terminology

Overview

Use cases for terminologies themselves as well as used in conjunction with information models and instance data.  

For example, in the healthcare space a terminology defines the universe of conceptual meanings that apply in, say, a clinical context.  Information models define the nature and structure of data to be gathered (e.g. a blood pressure measurement).  Instance data exists in formats described by information models, is linked to a medical record (which itself is an information model), and represent individual clinical interactions (e.g. an actual blood pressure measurement of patient X).

Information models are "bound" to the terminology in a variety of ways.  First, the name of the model and its fields should exist in the terminology, and the types of values used for qualitative data gathering should come from the terminology.  For example, the means by which a blood pressure measurement was taken (e.g. an arm cuff) should, in the information model, be represented by a concept in the terminology itself.

There is a complicated longitudinal maintenance process required to facilitate reporting across time as new concepts are added to the terminology, others become obsolete, and some are refined to become more specific.  A well maintained terminology will indicate to end users (within the terminology itself) how obsolete codes are now supposed to be used.   Sometimes manual intervention and review is required.

Use cases

Following are use cases for use of structured terminology.

• Ability to share a conceptual space across an organization (or organizations).
• Ability to allow individual stakeholders to "use their own language" when interacting with data while retaining the ability to have a single conceptual model behind those interactions.
  
  • A special case of this is a multi-lingual thesaurus that supports terms in multiple languages for the same "concept".
• Harmonizing lexicons across organizations
  • Only when organizations have formally defined lexicons
• Concept-based indexing and retrieval
  • Can be used to support "lay person" searching within an otherwise highly technical content domain (e.g. medicine).
• High level semantic categorization
  • e.g. UMLS semantic type, or "top ontology" like BFO
• Semantic-search (requires an ontology, not just a simple taxonomy or other "lighter" structured terminology)
  
  • Ability to ask questions like: "How many patients suffered a disease/disorder whose finding site is in the kidney?"
  • Ability to ask questions like: "What are all the prescribable beta blockers (e.g. in the formulary)?"
• Information model binding
  • Support expression of data gathered in a way that is bound to a terminology (for later semantic analysis)
• Linked data
  • Expressing data gathered in terms of information models bound to a terminology (this is really a precursor to "semantic search"). 
• Free-text analysis (and concept extraction), Mining of unstructured text
• "Levels of aggregation" analysis
  • e.g. "UKTC" categories tool that produces a sub-set of SNOMED where each "leaf" node has about the same number of children.  These are analytically derived "categorizations" within the terminology.
• Research analytics
  • E.g. “Given two medications for treating the same condition, which one is more effective?”
• Predictive/Trend analytics
  • E.g. “how likely is patient X to become obese given family history and outcomes of similar patients?”
• Cross-terminology mapping
• Reporting
  • Quality reporting
  • Mortality and epidiomology reporting
  • Public health surveillance

Metathesaurus use cases

• Interoperability across different systems, different organizations, or data captured at different points in time.
  • e.g. LOINC/CPT/ICD/SNOMED
  • e.g. "How many MI were treated last year"
• Coding for billing/reimbursement purposes
  • e.g. ICD9/10CM - very important in medicine.
• Structured quality reporting
  • e.g. "Meaningful Use"