Deep learning for topical trend discovery in online discourse about Pre-Exposure Prophylaxis (PrEP)

Area of Science:

• Public Health Informatics
• Computational Social Science
• Digital Health Communication

Background:

• Pre-Exposure Prophylaxis (PrEP) is a critical HIV prevention strategy.
• Social media platforms offer rich, real-time data on public discourse surrounding health interventions like PrEP.
• Traditional survey methods may not capture the nuances of online conversations regarding PrEP uptake barriers and facilitators.

Purpose of the Study:

• To explore online discourse surrounding Pre-Exposure Prophylaxis (PrEP) using computational informatics and public health expertise.
• To identify distinct clusters of conversation related to PrEP on social media.
• To understand the diverse ways PrEP is contextualized online, including misinformation and disinformation.

Main Methods:

• Collected 4,020 tweets related to PrEP using Twitter's Application Programming Interface (API).
• Employed a three-step neural network/deep learning process to analyze tweet clusters and their relationships.
• Identified and characterized 25 distinct clusters representing various facets of online PrEP discussions.

Main Results:

• Identified 25 distinct clusters of online discourse concerning PrEP.
• Key themes emerged, including medication side effects, social perceptions of PrEP use, and concerns about cost and access barriers.
• Revealed diverse online contextualizations of PrEP, alongside potential pockets of misinformation and disinformation.

Conclusions:

• Social media mining provides unique insights into PrEP perceptions and challenges beyond traditional research methods.
• The identified clusters offer valuable information for tailoring public health interventions and communication strategies.
• Leveraging this data can help pinpoint critical points for addressing and mitigating misinformation and disinformation about PrEP.