Country / Region
APAC
Tags
Data quality, Implementation, Mapping, Research
This presentation outlines our implementation of a syntactic and systematic validation process designed to significantly enhance mapping accuracy and consistency across multiple healthcare institutions.
During the mapping process, several factors contributed to inconsistencies, necessitating the need for our validation procedures. For instance, human errors during the manual mapping process led to inaccuracies. Additionally, when multiple mappers work concurrently, the same local terms can inadvertently be mapped to different standard codes. Consequently, we identified conflicts and nonuniformities resulting from the mapping of identical local terms to different SNOMED CT concepts within and across institutions. By systematically addressing challenges such as duplicate local terms and mapping conflicts, our approach leverages the comprehensive framework of SNOMED CT to ensure the precise and consistent use of standard terminology in multi-institutional data standardization projects.
Description
Our project targets the challenges encountered in multi-institution clinical data standardization efforts, where discrepancies in mapping local terms to SNOMED CT concepts frequently occur. These discrepancies result from the involvement of multiple mappers who apply diverse mapping criteria, eventually leading to various conflicts such as the same local terms being mapped to different concept IDs. For example, the local term 24HRS AMBULATORY BP was mapped to two different concepts, 164783007 |Ambulatory blood pressure recording (procedure)| and 170599006 |24 hr blood pressure monitoring (regime/therapy)|. To tackle these issues, we developed an automated systematic validation process comprising both syntactic and semantic validation that performs two levels of verification: within institutions and across institutions
1. The principle of within/across institutional systematic validation
: Local terms that are literally identical or equivalent in meaning are consistently mapped to the same SNOMED CT concept.
2. Syntactic Validation
(1) Verify that duplicate local terms are mapped to the same SNOMED CT concept.
(2)Check if local terms, which exactly match to SNOMED CT descriptions, are assigned to concepts other than the corresponding SNOMED CT concepts.
3. Semantic Validation
(1) Semantic tag: In addition to syntactic validation, semantic validation was performed to verify whether the meanings are identical. Any mappings that deviated from our predefined semantic tags were subsequently revised to align with the appropriate SNOMED CT concepts.
(2) The results of the syntactic validation were manually reviewed one by one to ensure that they were consistently mapped based on their semantic meaning.
Scope
1. Extensive Coverage
We are mapping concepts across multiple fields, and SNOMED's extensive coverage, diagnoses, procedures, examinations, and medications demonstrates its versatility as a unified terminology system. This broad coverage is a significant advantage in our standardization efforts because it enables us to integrate diverse data from different institutions into a single unified system.
2. Detailed Modeling
One of the primary strengths of SNOMED CT is its detailed, well-organized modeling of each concept, which incorporates both core attributes and subtle semantic distinctions. This enables users to quickly and accurately understand the exact clinical meaning of each concept‚ a critical factor when mapping diverse local terms from multiple institutions to a unified SNOMED CT concept. By leveraging these comprehensive models, our mapping process achieves enhanced semantic precision and accuracy, resulting in fewer conflicts and more consistent data integration across institutions.
How SNOMED CT will be used
To cover a diverse range of medical terms, we selected SNOMED CT because it supports the largest scope of local terms available.
Utilization of semantic tag
In order to maintain mapping consistency, we restricted semantic tags for each value set. Our automated mapping tool applied these semantic tag restrictions, and any mappings that fall outside the defined range were flagged during the manual review process.
For instance, in our implementation, we constrained the semantic tags for DiagnosisCodes to include only "disorder," "finding," "situation," "event," and "person." Similarly, for ProcedureCodes as well as ExamCodes, we limited the semantic tags to "procedure" and "regime/therapy."
This structured approach to semantic tags ensures that our mapping processes not only rely on literal matches but also incorporate semantic equivalence, ultimately guaranteeing semantic interoperability while upholding a high level of consistency and accuracy across various institutions and value sets.
Utilization of synonyms for each description
SNOMED leverages the fact that a single concept can have multiple synonyms to enhance the accuracy of mapping within institutions. Even if different institutions use different local terms to represent the same concept, as long as they share the same synonym in SNOMED, they will be mapped to the same SNOMED concept, thereby maintaining consistency.
Why SNOMED CT will be used
Contact
