MedDRA® automated term groupings using OntoADR: evaluation with upper gastrointestinal bleedings

Area of Science:

• Medical Informatics
• Knowledge Engineering
• Pharmacovigilance

Background:

• Standardized medical terminology is crucial for consistent data analysis and reporting.
• Existing methods for creating customized MedDRA term sets can be labor-intensive.
• Upper gastrointestinal bleeding (UGIB) requires precise terminology for accurate case description.

Purpose of the Study:

• To propose and validate a knowledge engineering method for building customized MedDRA term sets.
• To apply the method to the specific medical condition of upper gastrointestinal bleeding (UGIB).
• To leverage semantic resources like SNOMED CT for automated term selection.

Main Methods:

• A broad list of MedDRA terms related to UGIB was compiled and a gold standard was established by experts.
• MedDRA terms were formally described in the OntoADR semantic resource.
• Two semantic queries utilizing SNOMED CT concepts (morphology, finding site, clinical manifestations) were developed to automatically select candidate MedDRA terms for UGIB.

Main Results:

• Query 1 (morphology and site) achieved 71.0% recall and 81.4% precision (F1 score 0.76).
• Query 2 (Query 1 plus clinical manifestations) significantly improved recall to 96.7% while maintaining 77.0% precision (F1 score 0.86).
• The results demonstrate the feasibility of automated term selection for customized MedDRA sets.

Conclusions:

• Knowledge engineering techniques are feasible for constructing customized MedDRA term sets.
• The proposed semantic query approach shows potential for efficient and accurate term selection.
• Further research is needed to enhance precision and recall and confirm the strategy's utility.