Introduction to SNOMED CT
Cornerstone AI is publishing a blog post series focused on the various terminologies and dictionaries used in healthcare. This is the first of the series.
If you're a healthcare data scientist or analyst, chances are you've come across SNOMED CT. But what exactly is SNOMED CT, and why is it so important in the world of healthcare data? In this comprehensive guide, we'll delve into the intricacies of this terminology, exploring its purpose and other details.
What is SNOMED CT?
SNOMED CT stands for Systematized Nomenclature of Medicine - Clinical Terms. It is a comprehensive, multilingual clinical healthcare terminology that encompasses a vast array of medical concepts and relationships. The main objective of SNOMED CT is to provide a standardized language for recording, retrieving, and analyzing clinical data across different healthcare systems and settings. International Health Terminology Standards Development Organization (IHTSDO), a non-profit organization, manages the monthly distribution and maintenance of SNOMED CT.
SNOMED CT US Edition is a standalone release that combines the content of both the US Extension and the International releases of SNOMED CT. This edition of SNOMED CT is managed by National Library of Medicine (NLM), with additional concepts that closely relate to RxNorm and LOINC to cover medications and laboratory test needs in the US.
Why is SNOMED CT used?
The utilization of SNOMED CT brings numerous benefits to the healthcare industry. This powerful terminology provides a means of accurately representing clinical concepts and their relationships with unparalleled specificity. Unlike other coding systems that primarily focus on billing and administrative purposes, such as CPT, HCPCS and ICD-10, SNOMED CT offers more information by capturing the detailed clinical meaning behind each term. Eventual aim to improve patient care through the development of a terminology to record health care encounters more accurately in electronic health record (EHR) systems.
What do SNOMED CT Codes Look Like?
Entities within SNOMED CT are assigned unique numeric codes, known as concept identifiers or SCTIDs. For example, the SCTID for the concept representing "pneumonia (disorder)" would be: 60363000. If your team is working with SNOMED CT verbatims but notice missing or incorrect SNOMED CT codes, feel free to get in touch with Cornerstone AI to explore possible solutions. Cornerstone AI uses an NLP approach with just the raw verbatim as parameter to predict the best standard term and code to the SNOMED CT value set.
How is the SNOMED CT Hierarchy & Code Organized?
SNOMED CT follows a hierarchical structure to summarize the over 300,000 unique concepts. At the top level, it is divided into broad categories called hierarchies, which represent different areas of clinical medicine. These hierarchies include clinical findings, procedures, pharmaceutical substances, and more.
Each hierarchy is further divided into narrower clinical domains and organized using various types of relationships. With over 903,000 relationships in SNOMED CT, the connections between concepts allow for the representation of complex clinical scenarios and precise descriptions of diseases and their associated clinical manifestations. For example, a concept "cellulitis of foot" can belong to both "cellulitis" and "disorder of foot" hierarchies.
The SNOMED CT codes themselves have four components:
Concepts
Concepts are associated with unique descriptions that contain human-readable terms and respective unique code ID. Concepts are mapped to other concepts using hierarchical relationships to allow for ease of aggregation to a higher level.
Description
Descriptions provide additional human readable terms to these concepts. A concept can have multiple synonyms from different languages to these SNOMED CT concepts.
Relationships
Relationships links and describes the association of concept to another, such as using Is a or Finding site.
Reference sets
Reference sets represent a grouping of terms within SNOMED CT. These may be a collection of terms related to different medical specialties, laboratory values or mappings to other coding systems
Here’s an example of how these four components are related for a single SNOMED CT term:
Concept
Fully Specified Name: Pain of joint of knee (finding)
Snomed CT ID: 1003721002
Descriptions
US Synonym: Pain of joint of knee
UK Synonyms: Pain of joint of knee
Relationships
Is a (attribute): Pain (finding)
Finding site (attribute): Knee joint structure (body structure)
Reference sets
Simple RefSets: General Practice / Family Practice reference set, International Patient Summary, and Global Patient Set
Map RefSets: SNOMED CT to MedDRA simple map: 10033445 and SNOMED CT to ICD-10 extended map: M25.56
How Often are SNOMED CT Terms Updated?
SNOMED International releases monthly updates. These updates introduce new concepts, modify existing ones, and enhance the relationships between concepts based on community input and consensus.
SNOMED US Edition is updated twice a year to align with use in US healthcare systems.
What Mappings Exist to Other Dictionaries?
SNOMED CT is highly compatible with other widely used healthcare terminology systems, including ICD-10 (International Classification of Diseases, 10th Revision). ICD-10 offers less clinical specificity, but provides important information about reimbursements. If your team is struggling with mapping SNOMED CT to ICD-10, feel free to get in touch with Cornerstone AI to explore possible solutions. Cornerstone AI employs an algorithmic approach and leverages official mapping files to identify the most suitable ICD-10 code for over 120,000 SNOMED CT codes.
Additional references
Note: This blog post is not affiliated with or sponsored by any organization. It is intended to provide educational information for healthcare data scientists and analysts.
Cornerstone AI is an AI-assistant purpose-built to clean Real World Data (RWD) in healthcare. Our proprietary ML models automatically identify dirty data in each dataset and generate unique data cleaning rules for those data points.