From Engagement to Misinformation: Analyzing the Power of Medical Influencers on Social Media

The study by the University of Bristol researchers investigates how top medical influencers on social media, identified through a comprehensive dataset, shape public health discourse during COVID-19 by using socio-semantic network analysis. This research highlights the dual role of influencers in promoting medical interventions and spreading misinformation, offering insights for designing effective health communication strategies.


CoE-EDP, VisionRICoE-EDP, VisionRI | Updated: 10-07-2024 17:07 IST | Created: 10-07-2024 17:07 IST
From Engagement to Misinformation: Analyzing the Power of Medical Influencers on Social Media
Representative Image

A study by Zhijin Guo, Edwin Simpson, and Roberta Bernardi from the University of Bristol, explores the influential role of medical influencers on social media during the COVID-19 pandemic. Researchers collected tweets from the top 100 medical influencers, identified using Onalytica’s Influencer Score, to understand their impact on public health discourse. This comprehensive dataset formed the basis for constructing a socio-semantic network, linking influencers' identities and the key topics they discuss.

Unraveling the Influence of Medical Influencers

To achieve their objectives, the researchers developed a few-shot multi-label classifier to categorize influencers and other network actors based on their identities. They also employed BERTopic, a model for extracting thematic content from tweets, which helped in identifying the discursive frames used by medical influencers. Integrating these elements into a network model allowed for a nuanced analysis of how medical influencers shape public health discourse. The study highlights the dual role of medical influencers: they can drive positive engagement and attitudes toward medical interventions but also pose risks by spreading misinformation. Examples include misinformation about measles and COVID-19 vaccines, which has fueled vaccine hesitancy. Understanding these influencers' impact is crucial for addressing misinformation and restoring trust in health authorities.

Connecting Social Relations and Semantics

The socio-semantic network analysis used in the study connects social relations and the semantics of influencers' words. Previous methods focused on the centrality of words and actors' positions but overlooked the importance of actors' identities. Identities, defined by roles, social groups, or personal characteristics, influence how individuals communicate and engage with others. By incorporating identities into the analysis, the study provides a deeper understanding of how medical influencers shape health discourse on social media. The researchers presented several key results from their project on the top 100 medical influencers during the COVID-19 pandemic. They developed a few-shot multi-label classifier to identify the identities of medical influencers and their network actors, using prompt-based methods and fine-tuning to improve accuracy. BERTopic was used to identify discursive frames from the tweets, involving generating sentence embeddings for each tweet, reducing dimensionality for clustering, and identifying key topics using c-TF-IDF. Mapping the relationships between medical influencers' identities, social network ties, and discursive frames revealed how influencers adapt their discourse to engage different audiences, providing valuable insights for public health researchers and practitioners.

From Data Collection to Network Construction

Data collection involved identifying influential healthcare professionals using Onalytica’s methodology and collecting tweets from their official accounts. The researchers manually labeled a sample of users' bios to classify their identities, resulting in 54 labels representing job occupations, organization categories, and personal attributes. Social network construction involved analyzing various types of relations, such as replies, mentions, retweets, and quotes. The study's methodology included using pretrained models for sentence embedding and a few-shot classifier for identity classification. BERTopic was employed for topic modeling, leveraging pretrained language models and clustering techniques to identify meaningful topics in the tweets. Experimental results showed that the prompt "My occupation is {MASK}" was particularly effective in identity classification, achieving high precision and recall. Topic modeling with BERTopic provided more specific and semantically coherent topics compared to traditional methods like Latent Dirichlet Allocation (LDA). The study identified significant topics such as vaccine hesitancy, equitable vaccine distribution, and the importance of N95 masks.

Shaping Health Communication Strategies

The research underscores the importance of understanding medical influencers' impact on social media, particularly in the context of public health crises. The findings can inform the design of social media campaigns to promote accurate health information and combat misinformation. Future work will focus on refining identity classification methods, enhancing topic models, and exploring the dynamics of message transmission within social networks. The study emphasizes the need for detailed analysis of medical influencers' identities and their discourse to better understand their influence on public health conversations. By mapping identities and discursive frames, researchers can gain insights into how these influencers tailor their messages to resonate with diverse audiences. This understanding is crucial for developing strategies to counteract misinformation and enhance public health communication. The integration of socio-semantic network analysis with advanced machine learning techniques provides a powerful framework for analyzing the complex interactions between medical influencers and their audiences. This approach not only sheds light on the role of influencers in shaping health discourse but also offers practical tools for health communication professionals to design more effective interventions. As the study progresses, further refinement of identity classification and topic modeling techniques will enhance the accuracy and depth of insights gained from social media data. This ongoing research will contribute to a more comprehensive understanding of the digital landscape of health communication and the pivotal role of medical influencers in shaping public perceptions and behaviors during health crises like the COVID-19 pandemic.

  • FIRST PUBLISHED IN:
  • Devdiscourse
Give Feedback