Training Augmented Intelligent Capabilities for Pharmacovigilance: Applying Deep-learning Approaches to Individual...

Area of Science:

• Pharmacovigilance and Drug Safety
• Artificial Intelligence in Healthcare
• Regulatory Compliance

Background:

• Increasing regulatory scope necessitates advanced drug safety processes.
• Manual collection and collation of individual case safety reports (ICSRs) are time-consuming.
• Augmented intelligence offers a scalable solution for managing growing ICSR volumes.

Purpose of the Study:

• To train and evaluate cognitive services for accurate ICSR characteristic identification.
• To assess the feasibility of implementing augmented intelligence in case processing.
• To improve operational efficiency and data quality consistency in pharmacovigilance.

Main Methods:

• A consortium of ten deep-learning trained cognitive services was developed.
• 20,000 ICSRs were transcribed and annotated for model training and testing.
• Cognitive service outputs were validated by pharmacovigilance experts against ground-truth data.

Main Results:

• All ten cognitive services achieved an F1 or accuracy score of ≥75%.
• The developed cognitive services demonstrated adequate training and functionality.
• The study confirmed the potential for augmented intelligence in ICSR processing.

Conclusions:

• Machine learning and NLP techniques provide accurate outputs for ICSR processing.
• Augmented intelligence supports pharmacovigilance professionals, enhancing decision-making and efficiency.
• Cognitive services can transform drug surveillance towards proactive decision-making and improved patient safety.