Utilizing social media data for pharmacovigilance: A review

Area of Science:

• Medical Informatics
• Pharmacovigilance
• Computational Linguistics

Background:

• Automatic monitoring of Adverse Drug Reactions (ADRs) is crucial for patient safety.
• Social media data offers a vast resource for ADR detection, but system comparison is challenging.
• Existing research uses diverse data sources and techniques, hindering performance evaluation.

Purpose of the Study:

• To systematically review and characterize approaches for ADR detection and extraction from social media.
• To assess the applicability of these methods to pharmacovigilance.
• To propose a systematic pathway for ADR monitoring using social media.

Main Methods:

• Conducted a systematic literature search across major databases (Medline, Embase, Scopus, Web of Science) and Google Scholar.
• Included studies focused on extracting ADR information from user-posted social media data.
• Categorized studies by detection approach, corpus size, data source, availability, and evaluation criteria.

Main Results:

• Twenty-two studies met inclusion criteria, with a surge in publications recently.
• Supervised classification and lexicon-based approaches are popular for ADR detection and extraction.
• Publicly available annotated data remains scarce, impeding direct system performance comparisons.

Conclusions:

• Interest in using social media for ADR monitoring is growing.
• Both health-related and general social media yield valuable data, with general platforms offering higher volume.
• Increased availability of annotated data is essential to advance supervised learning approaches for ADR detection.