JMIR Publications

ABSTRACT

Background:

By August 2022, more than 93 million people were infected with COVID-19 in the US, while the death rate in rural areas (325/100,000) was much higher than in urban areas (248/100,000). As the pandemic spread, people have used social media platforms to express their opinions and concerns about COVID-19 related topics.

Objective:

This study aimed to 1) identify the key COVID-19 related topics in the Contiguous US communicated over Twitter and 2) compare the sentiment about these topics between urban and rural users.

Methods:

We collected tweets containing geolocation data from May 2020 to January 2022 in the Contiguous US. We relied on tweet’s geolocation to determine if its author was in an urban or rural setting. We trained multiple word2vec models with several corpora of tweets based on geospatial and timing information. Using word2vec model built on all tweets, we identified hashtags relevant to COVID-19, and performed a hashtag clustering to obtain related topics. We then run an inference analysis for urban and rural sentiments with respect to a topic based on the similarity between topic hashtags and opinion adjectives in the corresponding urban and rural word2vec models. Finally, we analyzed the temporal trend in sentiment using monthly word2vec models.

Results:

In a corpus of 407 million tweets, 350 million (86.0%) were posted by users in metropolitan areas, while 18 million (4.4%) were posted by users in rural areas. There were 2,666 hashtags related to COVID-19, which clustered into 20 topics. Rural users expressed a stronger negative sentiment than urban users about COVID-19 prevention strategies and vaccination (P < .001). There was an apparent political divide in the perception of politicians by urban and rural users, who communicated a stronger negative sentiment about Republican and Democratic politicians, respectively (P < .001). Regarding misinformation and conspiracy theory, urban users exhibited a stronger negative sentiment about covidiots and China virus topics, while rural users exhibited a stronger negative sentiment about Dr. Fauci and plandemic topics. Finally, we observed that urban users’ sentiment about economy appeared a transition from negative to positive in late 2021, which was in line with the US economic recovery.

Conclusions:

This study demonstrates that there is a statistically significant differences in the sentiment of urban and rural Twitter users regarding a wide range of COIVD-19 related topics. This suggests that social media can potentially serve as the basis for a tool to monitor public sentiment during pandemics for disparate types of regions. In turn, this may assist in the geographically-targeted deployment of epidemic prevention and management efforts. Though this research focused on COVID-19 specifically, it can be readily reused to investigate other topics without additional data collection or model training.