September 21 2015 Department of Biomedical Informatics BMIF

September 21, 2015 Department of Biomedical Informatics BMIF 6300 Standards and Terminologies S. Trent Rosenbloom, MD MPH Associate Professor and Vice Chair Departments of Biomedical Informatics, Internal Medicine and Pediatrics Vanderbilt University Medical Center

Standards and Terminologies Standards are principles and rules designed to ensure that methods used and products created reliably and consistently conform to expectations Software Standards Detail: minimum set of functions provided methods used to achieve those functions formatting of the data structure

Standards and Terminologies minimum set of functions provided “Content standards” methods used to achieve those functions “Functional standards” formatting of the data structure “Syntactic standards” ← Messaging, e. g. , HL 7 XML “Semantic standards” ← Terminology, e. g. , SNOMED CT

Terminologies can provide formal and machine-computable representations of knowledge and data Such representation can facilitate interoperability, dissemination, decision support, research

Terminologies are formal representations of entities and their interrelationships. Embodied as concepts, terms, linkages ▪ Concepts are the cognitive representation of entities or meanings ▪ Terms are evocative words or phrases ▪ Linkages are explicitly defined relationships

Terminologies Concept - ischemic injury and necrosis of heart muscle cells resulting from absent or diminished blood flow in a coronary artery Terms – ▪ Myocardial Infarction ▪ Heart Attack Linkage – ▪ is_a Disease of the Heart ▪ has_severity Severities

Morning Star Evening Star The second planet from the sun, having an average radius of 6, 052 kilometers (3, 761 miles), a mass 0. 815 times that of Earth, and a sidereal period of revolution about the sun of 224. 7 days at a mean distance of approximately 108. 2 million kilometers (67. 2 million miles).

Physical Entity The second planet from the sun, having an average radius of 6, 052 kilometers (3, 761 miles), a mass 0. 815 times that of Earth, and a sidereal period of revolution about the sun of 224. 7 days at a mean distance of approximately 108. 2 million kilometers (67. 2 million miles). Representative Terms Conceptual Experience Morning Star Evening Star Venus Adapted from Campbell, ‘Representing thoughts, words, and things in the UMLS’, 1998.

Planets of the Solar System inside outside Mercury Jupiter Venus Saturn Earth Neptune

Concept: Myocardial Infarction CUI: C 0027051 Semantic Type: Disease or Syndrome Entity: Gross necrosis of the myocardium, as a result of interruption of the blood supply to the area. (Dorland, 27 th ed) Representative Terms (synonyms): Myocardial Infarction Attack coronary Cardiac infarction Heart attack Infarction of heart MI - Myocardial infarction Myocardial Infarct Myocardial infarction (disorder) Myocardial infarction syndrome myocardium; infarction Adapted from the UMLS Metathesaurus. More Specific Concepts (children): Acute myocardial infarction Old myocardial infarction Microinfarct of heart True posterior wall infarction Aborted myocardial infarction Other specified anterior myocardial infarction Silent myocardial infarction Subsequent myocardial infarction Postoperative myocardial infarction First myocardial infarction Myocardial infarction with complication Non-Q wave myocardial infarction

There a lot of terminologies In 2003, the National Committee on Vital Health and Statistics (NCVHS) recommended a subset of existing terminologies as: “uniform data standards for patient medical record information (PMRI) and the electronic exchange of such information”

Terminologies PMRI standards: SNOMED CT (as licensed by the National Library of Medicine) - for the exchange, aggregating, and analysis of patient medical information. Logical observation Identifiers Names and Codes - for the representation of individual laboratory tests Federal Drug Terminologies: ▪ Rx. Norm; ▪ The representations of the mechanism of action and physiologic effect of drugs from NDF-RT; ▪ Ingredient name, manufactured dosage form and package type form the FDA

UMLS (Unified Medical Language System) The UMLS is a terminology collection Concepts are unique No formal relationships among concepts present, per se

Using the UMLS: Semantics and relationships from source terminologies lost (or implied) May mix up different levels of detail from different terminologies Can loose link with source terminology, which can hinder maintenance

Terminology [old] History Classification scheme for the London Bills of Mortality - 16 th century John Gaunt’s refinement - middle of the 17 th century International Classification of Diseases (ICD) first adopted in Paris in 1900 Multi-axial Standardized Nomenclature of Diseases (SND) – 1928 Standardized Nomenclature of Diseases and Operations (SNDO) - 1933

Terminology History “Modern era for clinical descriptions” With SND and SNDO ▪ Multiaxial: users could model complex concepts by constructing them from more primitive building blocks ▪ Designed to classify diseases based on: Etiology Manifestations Relationships between them

Terminology Desiderata Statement of purpose, scope, and comprehensiveness Complete coverage of domain specific content Use of concepts rather than terms, phrases and words (concept orientation) Concepts do not change with time, view or use (concept consistency) Concepts must evolve with change in knowledge Concepts identified through nonsense identifiers (context-free identifier) Representation of concept context consistently from multiple hierarchies Concepts have single explicit formal definitions Support for multiple levels of concept detail Absence of or methods to identify duplication, ambiguity, and synonymy Integration with other terminologies Mapping to administrative terminologies Adapted from Cimino, ‘Desiderata for controlled medical vocabularies in the twenty-first century’, 1998.

Coverage achieved by one of two ways ▪ Post-coordination - complex concepts from different levels of detail are composed as needed from fundamental concepts (e. g. , ‘chest pain’ composed from the concepts ‘chest’ and ‘pain’ when needed) ▪ Pre-coordination - all levels of detail are modeled with distinct concepts (e. g. , ‘chest pain’, ‘substernal chest pain’, and ‘crushing substernal chest pain’ are all in the terminology)

Completeness measured by Coverage: ▪ coverage calculated as the proportion of concepts covered by a terminology ▪ multiple studies: post-coordinated terminologies generally have better coverage than pre-coordinated terminologies

Post-coordination versus Pre-coordination

Select One Flavor

Select One Topping

Select One Cone

…or…Select One Favorite

Post-Coordination ▪ ▪ ▪ Flexible Wide choice Rules implied Explicit relationships Inefficient Permits Inappropriate combinations Pre-Coordination ▪ ▪ ▪ No flexibility Limited choice Asserted knowledge Implied relationships Efficient Only appropriate combinations

Consequences of post-coordination: D 5 -46210 01 Acute appendicitis, NOS G-A 231 01 Acute D 5 -46100 01 Appendicitis, NOS T-59200 01 Appendix, NOS G-CO 06 01 In T-59200 01 Appendix, NOS ▪ Inefficient post-coordination: “too cumbersome for M-41000 01 Acute inflammation, NOS 01 Acute complex problem entry” G-A 231 G-CO 06 01 In M-40000 01 inflammation, NOS Table. Duplication due to compositionality: four ways to compose ‘Appendicitis’ in SNOMED, from the CANON Group ▪ Nonsensical Concepts ▪ Concept duplication

Rigorous development may produce terminologies unusable by healthcare providers for routine clinical tasks. Rector: tension between clinical usability and meticulous knowledge representation mirrors the conflict ▪ human users require flexible, expressive terminologies that model common colloquial phrases ▪ computer programs are generally designed to process formally defined concepts having rigidly defined interrelationships.

Rector’s six tasks for terminologies: 1) support efficient data entry and query formulation 2) record and archive clinical information 3) support sharing and reuse of clinical information 4) infer and suggest knowledge according to decision support algorithms 5) support terminology maintenance 6) create a natural language output from manual structured input

Interface Terminologies Generally a set of flexible, user friendly, colloquial terms displayed in via computer programs. Use assertional medical knowledge to support efficiency, size and focus Have been used for problem list entry, clinical documentation, provider order entry Rosenbloom ST, et al. Interface terminologies: facilitating direct entry of clinical data into electronic health record systems. J Am Med Inform Assoc. 2006 May-Jun; 13(3): 277 -88. .

Terminology Subsets SNOMED CT 2012 AA in the UMLS contains: ▪ over 311, 000 unique concepts, 10% are diagnoses ▪ almost 800, 000 descriptions ▪ approximately 1, 360, 000 links S-CT

Terminology Subsets SNOMED CT CORE subset in the UMLS contains: ▪ about 5, 000 unique concepts, primarily diagnoses ▪ covers 95% of diagnoses recorded from 7 healthcare sites (Beth Israel Deaconess Medical Center, Intermountain Healthcare, Kaiser Permanente, Mayo Clinic, Nebraska University Medical Center, Regenstrief, Hong Kong Hospital Authority) CORE S-CT Fung KW, Rosenbloom ST, et al. Testing Three Problem List Terminologies in a simulated data entry environment. AMIA Annu Symp Proc. 2011: 445 -54.

Terminology Subsets SNOMED CT VA-KP subset in the UMLS contains: ▪ about 17, 000 unique concepts ▪ primarily contains precoordinated concepts from the “Clinical Finding” hierarchy CORE S-CT VA-KP

Terminology Subsets Institutional subsets / supersets ▪ can be created locally as SNOMED CT extensions ▪ can be created locally without regard to SNOMED CT ▪ may or may not follow standard formalisms CORE S-CT local VA-KP

Terminology Subsets CORE 1, 449 2, 437 1, 756 S-CT 302, 537 527 2475 CCPSS 986 12, 675 9510 VA-KP * Statistics thanks to Lina Sulieman