An Introduction to SNOMED CT

A brief introduction to SNOMED CT - the ontology based medical terminology. This covers the basic definitions, the difference between SNOMED CT and ICD9, Post co-ordination use-cases and some general information.
This is not an extensive guide for SNOMED CT adoption in a system

1. Introduction to SNOMED Guru Kini

2. What is SNOMED Applications of SNOMED Concepts, Hierarchies & Relationships SNOMED vs. ICD9 Post Co-ordination and Pre Co-ordination Things to consider before opting for SNOMED This may be useful for requirement analysts, project managers or software developers who are faced with implementing SNOMED support in the healthcare related software. This is not a How-To guide or a detailed technical analysis of SNOMED. However, there are several useful references in the notes and the slides for these contexts. Overview Apologies for the heavily bulleted slides

3. S ystemized NO menclature of MED icine It is an organized lists of a wide variety of clinical terminology defined with unique codes Perhaps the most comprehensive clinical terminology in the world It is older than you think! Started as early as 1965 Has had several iterations since Currently used variant: SNOMED Clinical Terms (SNOMED CT) Provides cross maps to various terminologies such as ICD9, ICD10 and LOINC What is SNOMED? Continued…

4. SNOMED covers a wide no. of medical terminologies for Disorders and finding (what was observed) Procedures (what was done) Event (what happened) Substance/Medication (what was consumed/administered) Pretty much anything that may be used to capture Medical data SNOMED is designed as an Ontology Each Concept could have relationships with other Concepts What is SNOMED? (cont d .)

5. Ontology is a formalized model for particular domain Consists of 4 basic elements: Individuals/Objects – an actual, concrete instance of something that can described as a part of the ontology (viz., Angina, Class 1) Classes – abstract concepts (Procedure, Disease, etc.). Essentially a “category” for a set of Individuals Attributes – certain properties associated with a class Relations – define who certain classes (and objects) are related to each other Yes, a class diagram of a program written in an OO programming language is a visual representation of an ontology What is an ontology?

6. National Library of Medicine’s UMLS is the one stop shop for SNOMED codes SNOMED is now freely available for use for U.S. users It is now maintained by  International Health Terminology Standards Development Organization ( IHTSDO ) Where can I find SNOMED codes?

7. A “Concept” is the basic unit in SNOMED Has a numeric representation (Concept ID), which is assigned arbitrarily Can represent anything that may have a possible use in recording clinical information The same Concept could have several “Synonyms” to accommodate variations in name. E.g., “Myocardial infarction” could also be called “Infarction of heart” or just “Heart Attack” All Concepts are divided in “Hierarchies” Hierarchies do not overlap Clinical Finding/Disorder, Procedure, Substance, etc. are all examples There are some 20+ main hierarchies, more can be added over time “Relationships” between Concepts can be defined “Is a” is most common relationship Other relationships could be defined as “Attributes” of a Concept Concepts, Hierarchies & Relationships

8. Neonatal diabetes mellitus Diabetes mellitus syndrome in newborn infant Diabetes mellitus Neonatal disorder Endocrine pancreatic structure Found in Is a type of Is a type of AKA Disorder of endocrine system Is a type of Disorder of body system Is a type of Disorder of foetus or newborn Is a type of Disorder of body site Is a type of Disease Is a type of Is a type of Example: using SNOMED relationships Body Structure Disorder Synonym Relation Legend

9. Top SNOMED Hierarchies

10. Disorders and findings often used interchangeably “Finding” is a general observation or a judgment of the patient’s physical, mental or social condition (current or historical). A finding need not be an “abnormal” state and can be somewhat vague. E.g.: Patient complaints/Symptoms (e.g., cough, shivering) Lab result observations (e.g., Allergy Skin Test Positive) Social setting (e.g., Unsafe play area, Patient’s dependents) A “Disorder” or “Disease” is a sub-set of “Finding” concept that are necessarily abnormal physical or mental conditions for the patient. E.g.: Tuberculosis Angina, Class I A Finding may be the initial diagnosis of the patient’s condition which may lead to the discovery of a Disorder. E.g., A complaint of Chest pain (Finding) may lead to a final diagnosis of Angina, Class I (Disorder) Bleeding of Gums (Finding) may lead to Hematoma of gingivae (Disorder) Cough (Finding) may lead to Tuberculosis (Disorder) Disorder vs. Finding

11. ICD is a relatively ancient code family Late 19 th century roots ICD9 was developed in 1970s! Even ICD10 is ~30 years old! ICD is a classification whereas SNOMED is a terminology ICD tends to be more abstract. With SNOMED the user can get a more accurate description ICD9 (or ICD10) tend to have a “unspecified” catch-all slot for most disorders. SNOMED is far more extensive than ICD9 ICD only covers disorders SNOMED is implemented as an ontology Any number of relationships can be defined for each concept SNOMED vs. ICD Continued…

12. SNOMED CT – is better suited for capturing relevant data during an encounter Allows the user to capture the various aspects associated with a disorder (Post Coordination) This encourages the user to capture associated information like Severity, Body part affected, Cause (force or substance), laterality (viz., left or right), Morphology (form) in structured form ICD9/10 – used in cases where data need not be very granular Each code is very rigidly defined and does not support qualifiers Used in Insurance billing, Morbidity recording (death cause etc.), Epidemiological tracking (public health surveillance) These use-cases usually can work with a general disease class Usually, SNOMED CT is considered a good way to enter the medical information and ICD9/10 is considered a good way to export information SNOMED vs. ICD (cont d .)

13. Why convert ? Business requirements - e.g., your Billing Clearing house may insist on ICD9/ICD10 Interoperability – to import CCD/CCR from other sources that use ICD9/ICD10 Future-proofing – ICD10 may eventually be the de facto standard SNOMED CT provides cross-mapping with ICD9 and ICD10 terminologies Since SNOMED CT is much larger in size and because ICD9/10 primarily deal with disorders, only a portion of SNOMEC can be mapped. ICD9/ICD10 to SNOMED is mostly 1:1, however SNOMED to ICD9/ICD10 may have a number of 1:Many mapping May need user input or context based intelligence to convert SNOMED CT to ICD9/10 Conversions

14. Post Coordination Post Co-ordination: representing a medical term using two or more SNOMED concepts. E.g., “Wound in the right hand” could be coded in parts SNOMED’s ontological structure naturally lends itself to Post Co-ordination based capturing of data. However, it can cause confusion as the same concept can be represented in more than one way! E.g. Wound (Disorder) 399963005 Hand ( Body Structure) 302539009 Right ( Qualifier) 24028007 Found in Laterality Wound in hand (Disorder ) 283059006 Right ( Qualifier) 24028007 Laterality Continued…

15. ICD9/10 is design to be Pre Co-ordination oriented. That is, there is usually one unique code for a given medical term. It also implies that one cannot capture much of term’s associative data using ICD9/10 itself Some of the SNOMED “leaf” concepts are specific enough to be used as Pre Co-ordinated terms For any system, Pre- vs Post- arguments essentially deal with one question: “For capturing a medical term, what do you need more: uniqueness or granularity”? Post Coordination (cont d .)

16. Familiarize yourself Use a SNOMED CT Browser Look at some existing SNOMED subsets (CORE, VA/KP) Consider your use-cases Determine which hierarchies will the system use (Finding, Procedure, Disorder) Determine which concepts will the system use (Create a Subset) Determine when will your system use primitive (generic/vague) and non-primitive (specific) concepts Consider your conversion requirements Look at the UMLS Metathesaurus and see if you can use that instead of SNOMED Allow room for updates SNOMED CT releases updates twice a year Your system should be able to accept the updates without going out-of-action Things to consider before implementing SNOMED CT in your system

17. http:// protege.stanford.edu/publications/ontology_development/ontology101-noy-mcguinness.html - What is an Ontology http://perspectives.ahima.org/a-comparison-between-a-snomed-ct-problem-list-and-the-icd-10-cmpcs-hipaa-code-sets / - Comparing ICD10 and SNOMED CT http://www.connectingforhealth.nhs.uk/elearning/snomedct/flash / - SNOMED Intro by NHS References