Auditing the semantic completeness of SNOMED CT using formal concept analysis

Area of Science:

• Ontology engineering
• Medical informatics
• Formal concept analysis

Background:

• SNOMED CT is a comprehensive clinical terminology crucial for health data interoperability.
• Auditing the semantic completeness of large ontologies like SNOMED CT is essential for data quality.
• Existing methods for evaluating ontological completeness can be resource-intensive.

Purpose of the Study:

• To develop and evaluate a novel approach for auditing SNOMED CT semantic completeness.
• To utilize formal concept analysis (FCA) for modeling SNOMED CT expressions.
• To identify and evaluate 'anonymous nodes' as a proxy for semantic completeness.

Main Methods:

• Developed an FCA-based model to formalize SNOMED CT expressions.
• Identified anonymous nodes within the FCA model for evaluation.
• Employed quasi-Poisson regression to assess the relationship between anonymous nodes and semantic completeness.
• Sampled data from the Procedure and Clinical Finding domains of SNOMED CT.
• Conducted case studies for validation of the anonymous node metric.

Main Results:

• A significant negative correlation was found between anonymous nodes and fully defined concepts (p < 0.001).
• The Clinical Finding domain exhibited fewer anonymous nodes compared to the Procedure domain (p < 0.001).
• Case studies confirmed anonymous nodes as an effective index for SNOMED CT auditing.

Conclusions:

• Anonymous nodes derived from FCA serve as a viable proxy for SNOMED CT semantic completeness.
• The proposed FCA-based method offers a novel approach for auditing large ontologies.
• This methodology can be applied to assess semantic completeness within and across different domains of SNOMED CT and other ontologies.