Using big data to understand the online ecology of COVID-19 vaccination hesitancy

The rapid development and deployment of COVID-19 vaccines presented a significant opportunity to control the pandemic. However, low vaccine uptake rates in many countries highlighted the challenge of achieving herd immunity. While numerous surveys explored vaccine hesitancy, research examining this phenomenon within the social media ecosystem remained limited. This study aimed to fill this gap by analyzing a large dataset of YouTube comments to understand the factors driving vaccine hesitancy and inform future public health messaging. The researchers hypothesized that analyzing real-time social media data would provide valuable insights into public perceptions and behaviors regarding COVID-19 vaccination, offering actionable intelligence for public health interventions. The study's significance lies in its potential to leverage big data analytics for a deeper understanding of vaccine hesitancy dynamics and ultimately contribute to more effective public health strategies.

The study begins by defining vaccine hesitancy according to the World Health Organization (WHO), categorizing it as influenced by confidence, complacency, and convenience. It then reviews existing literature on COVID-19 vaccine hesitancy, highlighting findings from various survey studies that identified concerns about vaccine safety, side effects, and lack of trust as major drivers. The literature review also examines previous research on vaccine hesitancy within social media contexts, noting a preponderance of sentiment analysis and topic categorization studies but a scarcity of research investigating the impact of specific themes on vaccination intentions. The authors identify a gap in the literature regarding the use of individual-level social media data to explore the causal relationships between vaccine hesitancy factors and vaccination intentions. This gap motivated the current study's methodology, which employs big data analytics to overcome the limitations of traditional survey methods, such as small sample sizes and low response rates.

The study employed a mixed-methods approach, combining thematic analysis of qualitative data with causal modeling using quantitative data derived from text mining. Data were collected from YouTube comments on videos uploaded by mainstream news outlets related to COVID-19 vaccine efficacy announcements. A total of 43,203 English-language comments focusing on vaccine-related discussions were included in the analysis. SAS Text Miner (9.4) was used for text analytics, involving text parsing, filtering, and clustering to identify recurring themes. The Term Frequency-Inverse Document Frequency (TF-IDF) algorithm was employed to assess the relevance of words within the dataset. A hierarchical clustering algorithm with Singular Value Decomposition (SVD) was used to group related terms into clusters. Eleven distinct clusters were identified, and these were then interpreted and labeled using an inductive approach, involving independent review and discussion among researchers. The labeled clusters were then used as constructs for a multiple regression analysis to examine the relationships between various factors and vaccination intention. The Health Belief Model (HBM) provided a framework for interpreting the relationships between the constructs (perceived susceptibility, perceived severity, perceived benefits, perceived barriers, trust in pharma, trust in government, trust in media, political ideologies, and vaccine misinformation). The researchers addressed concerns about reliability and validity by detailing data preparation, cluster interpretation techniques, and the validation process against the study objectives. Multiple regression analysis in SPSS was utilized to determine the predictive power of the identified constructs on vaccination intention, after addressing issues such as collinearity, nonlinearity, and the normality of residuals.

The text analytics yielded eleven distinct clusters, which were interpreted and labeled as follows: Vaccination Intention, Political Ideologies, Perceived Trust in Pharma, Perceived Trust in Government, Perceived Trust in Media, Perceived Barriers, Vaccine Misinformation (two clusters), Perceived Severity, Perceived Benefit, and Perceived Susceptibility. Multiple regression analysis revealed that perceived susceptibility and political ideologies were the most significant predictors of vaccination intention. Specifically, individuals who perceived a higher likelihood of contracting COVID-19 were more inclined to get vaccinated. Furthermore, political ideologies played a crucial role, with individuals leaning toward conservative viewpoints exhibiting lower vaccination intentions. Other factors, such as perceived trust in pharmaceutical companies, media, or government, did not show a statistically significant association with vaccination intention in this analysis. The model explained 74.7% of the variance in vaccination intention (R² = 0.747).

The study's findings highlight the complex interplay of factors influencing COVID-19 vaccine hesitancy within the online environment. The significant influence of perceived susceptibility aligns with the Health Belief Model, emphasizing the importance of risk perception in shaping health behaviors. The strong effect of political ideologies underscores the politicization of the pandemic and its impact on vaccine uptake. The lack of significant associations with other constructs, such as perceived trust in institutions, may suggest that other factors, not captured in this analysis, are at play or that these trust factors operate differently within the specific context of social media discourse. The study's methodology, using big data analytics to analyze real-time social media data, offers a powerful approach to understanding dynamic shifts in public opinion and informing targeted interventions.

This study provides valuable insights into the drivers of COVID-19 vaccine hesitancy on YouTube, demonstrating the utility of big data analytics in understanding complex public health challenges. The findings emphasize the significant influence of perceived susceptibility and political ideologies on vaccination intentions. Future research should explore longitudinal trends, incorporate data from multiple social media platforms, and investigate the role of demographic factors. Developing validated survey scales based on the identified clusters would allow for more robust quantitative studies. The results suggest that public health messaging needs to be tailored to address risk perceptions and account for the politicization of health information.

The study is limited by its focus on YouTube comments during a specific time period (November 2020, coinciding with the US election), potentially influencing the results. Sampling bias may also exist due to the reliance on a single social media platform and the exclusion of non-English comments. The lack of demographic data prevents exploring the association between individual characteristics and vaccine hesitancy behaviors. Finally, the relatively novel approach of using text analytics in predictive modeling of vaccination intentions requires further validation through comparative studies using different methodologies and datasets.