Augmenting Product Defect Surveillance Through Web Crawling and Machine Learning in Singapore

Area of Science:

• Pharmaceutical quality control
• Public health surveillance
• Computational methods in pharmacovigilance

Background:

• Substandard medicines pose significant public health risks.
• Manual monitoring of pharmaceutical quality is labor-intensive and time-consuming.
• The complexity of pharmaceutical supply chains necessitates efficient detection methods.

Purpose of the Study:

• To develop and evaluate an automated system for detecting substandard medicine alerts.
• To compare the performance of machine learning algorithms against keyword-based methods.
• To improve the efficiency and scope of environmental scanning for substandard medicines.

Main Methods:

• A web crawler was designed to extract alerts from regulatory agency websites.
• Data were classified using expert-derived keywords and machine learning algorithms.
• Model performance was assessed using recall, precision, and F1 scores on validation datasets.

Main Results:

• The web crawler extracted over 12,000 unique alerts, identifying 1160 substandard medicine alerts.
• An ensemble model combining machine learning and keywords achieved high performance.
• The best models demonstrated recall of 94-97% and precision of 80-85% on temporal validation.

Conclusions:

• Automated systems combining web crawlers with machine learning and keyword filtering enhance substandard medicine detection.
• This approach significantly improves horizon scanning capabilities for public health.
• Robust automation is crucial for monitoring pharmaceutical quality in complex supply chains.