A use case study on late stent thrombosis for ontology-based temporal reasoning and analysis

Area of Science:

• Biomedical Informatics
• Medical Device Safety
• Clinical Data Analysis

Background:

• Medical device adverse event reports contain complex temporal information crucial for safety analysis.
• Extracting and analyzing temporal relationships from unstructured narratives is challenging.
• Existing methods may not adequately capture the nuances of event timing in adverse event reports.

Purpose of the Study:

• To apply the Clinical Narrative Temporal Relation Ontology (CNTRO) and its Timeline Library to extract and analyze temporal information from medical device adverse event narratives.
• To evaluate the accuracy of the CNTRO Timeline Library in identifying, ordering, and calculating durations of temporal events.
• To demonstrate the utility of temporal data extraction for verifying clinical recommendations and identifying new trends.

Main Methods:

• Annotation of 238 late stent thrombosis adverse event narratives from the FDA MAUDE database.
• Application of the CNTRO Timeline Library for temporal event identification, ordering, and duration calculation.
• Evaluation of the library's accuracy in event ordering and duration determination.

Main Results:

• The CNTRO Timeline Library achieved 95% accuracy in correctly ordering temporal events within the narratives.
• Accuracy in determining documented event durations was 80% for 41 narratives.
• Accuracy in determining durations between events was 76% for 77 narratives.

Conclusions:

• The CNTRO Timeline Library demonstrates high accuracy in extracting and analyzing temporal information from medical device adverse event narratives.
• This temporal analysis can support the verification of clinical recommendations, such as drug administration lengths.
• The proposed methodology offers potential for uncovering previously unknown temporal trends in other medical device adverse event data.