Twitter discourse reveals geographical and temporal variation in concerns about COVID-19 vaccines in the United States

Abstract

The speed at which social media is propagating COVID-19 related misinformation and its potential reach and impact is growing, yet little work has focused on the potential applications of these data for informing public health communication about COVID-19 vaccines. We used Twitter to access a random sample of over 78 million vaccine related tweets posted between December 1, 2020 and February 28, 2021 to describe the geographical and temporal variation in COVID-19 vaccine discourse. Urban suburbs posted about equitable distribution in communities, college towns talked about in-clinic vaccinations near universities, evangelical hubs posted about operation warp speed, thanking God, exurbs posted about the 2020 election, Hispanic centers posted about concerns around food and water, and counties in the ACP African American South posted about issues of trust, hesitancy, and history. The graying America ACP community posted about the federal government’s failures; rural middle American counties posted about news press conferences. Topics related to allergic and adverse reactions, misinformation around Bill Gates and China, and issues of trust among Black Americans in the healthcare system were more prevalent in December, topics related to questions about mask wearing, reaching herd immunity and natural infection, and concerns about nursing home residents and workers increased in January, and themes around access to black communities, waiting for appointments, keeping family safe by vaccinating and fighting online misinformation campaigns were more prevalent in February. Twitter discourse around COVID-19 vaccines in the United States varied significantly across different communities and changed over time; these insights could inform targeted messaging and mitigation strategies.

Delayed medical care and underlying health in the United States during the COVID-19 pandemic: A cross-sectional study

“…Additionally, among those that delayed care, COVID-19 rates did not impact their decision to delay. This may be explained in part by geographic variations in beliefs around the severity of the pandemic ( The New York Times, 2022 , Guntuku et al, 2021 ). However, an alternative reason could be that individuals consistently delay care for reasons that existed before COVID-19, such as a lack of access to health care facilities, health insurance, or paid sick leave.…”

Spatial Analysis of COVID-19 Vaccination: A Scoping Review

“…Several studies used Twitter as the source of data for their analysis [24,29,30,35]. Social media data, particularly geotagged tweets, are valuable and cost-effective resources for near real-time spatial and spatiotemporal analyses.…”

“…The contribution of 11 articles to the spatial analysis of COVID-19 vaccination was only limited to mapping [24][25][26][27][28][29][30][31][32][33][34]. Although most mapping studies disregarded time components, and used static presentation of data, temporal variations or transmission dynamics were mapped in [24,29,30,[35][36][37].…”

“…Using heat maps, they concluded that public sentiment on COVID-19 vaccines varied significantly over time and space. Similarly, Guntuku et al [35] utilized choropleth maps to describe geographical and temporal variations of COVID-19 vaccine concerns in the US based on vaccine-related tweets. Their findings suggested that Twitter discourse regarding COVID-19 vaccines in the US varies significantly across different communities, and changes over time.…”

Tracking Public Attitudes Toward COVID-19 Vaccination on Tweets in Canada: Using Aspect-Based Sentiment Analysis

“…Hussain et al [ 11 ] performed sentiment analysis on Facebook and Twitter data to understand public sentiments toward COVID-19 vaccines in the United Kingdom and the United States. Guntuku et al [ 12 ] reported geographical and temporal variation in concerns about COVID-19 vaccines in the United States by using topic modeling. Kwok et al [ 13 ] analyzed tweets from Australian users by using topic modeling and sentiment analysis.…”