Characterizing Public Sentiments and Drug Interactions in the COVID-19 Pandemic Using Social Media: Natural Language...

Area of Science:

• Computational linguistics
• Public health informatics
• Social media analytics

Background:

• Traditional studies on COVID-19 medications faced limitations like reporting biases and long data collection times.
• Social media offers real-time drug-related data, crucial for understanding pandemic impacts on drug use and monitoring misinformation.

Purpose of the Study:

• To develop a natural language processing (NLP) pipeline for analyzing social media discourse on COVID-19-related drugs.
• To harness real-time social media data for timely public health insights and misinformation monitoring.

Main Methods:

• A pipeline utilizing pretrained language model-based NLP techniques was constructed.
• Key modules included named entity recognition, sentiment analysis, topic modeling, and drug network analysis for adverse drug reactions (ADR) and drug-drug interactions (DDI).
• The pipeline analyzed COVID-19 drug-related tweets from February 2020 to April 2022.

Main Results:

• Over 2.1 million relevant tweets were analyzed, with ivermectin, hydroxychloroquine, remdesivir, zinc, and vitamin D being the most discussed drugs.
• Public perception was influenced by endorsements and directives, not empirical evidence; repurposed drugs garnered more attention than approved ones.
• Analysis identified key discussion topics and potential ADR/DDI patterns, vital for public health surveillance.

Conclusions:

• An NLP pipeline provides a robust method for large-scale public health monitoring.
• This framework offers valuable supplementary data for epidemiological studies on drug safety and interactions.
• The study establishes a foundation for future social media-based public health analytics.