Mining clinical big data for drug safety: Detecting inadequate treatment with a DNA sequence alignment algorithm

Area of Science:

• Health Informatics
• Pharmacovigilance
• Computational Biology

Background:

• Health data mining offers significant potential for enhancing drug safety surveillance.
• Identifying specific clinical event sequences is crucial for understanding treatment efficacy and adverse events.

Purpose of the Study:

• To develop and evaluate a visual analytics tool for detecting clinical event sequences within a clinical data warehouse.
• To adapt the Smith-Waterman algorithm for retrieving temporal patterns in electronic health records for pharmacovigilance.

Main Methods:

• Adaptation of the Smith-Waterman algorithm for clinical event sequence retrieval with temporal patterns.
• Development of a web-based user interface for interactive query specification and result visualization.
• Evaluation using pharmacovigilance use cases, including detection of inadequate treatment decisions.

Main Results:

• The adapted Smith-Waterman algorithm demonstrated suitability for pharmacovigilance tasks.
• Achieved a high F-measure of 0.87, indicating strong precision and recall in identifying clinical event sequences.
• The user interface facilitated rapid identification of inadequate treatment cases.

Conclusions:

• The developed visual analytics tool, leveraging an adapted Smith-Waterman algorithm, is effective for drug safety and pharmacovigilance.
• The approach enhances the ability to analyze electronic health records for critical clinical event patterns.
• This method supports improved decision-making in patient treatment and drug safety monitoring.