Tweets more viral than COVID-19

Information shared on social media—whether official or not, scientific or not—shape beliefs regarding the novel coronavirus disease (COVID-19), according to a new study.

“The dynamics of the pandemic, news, scientific and nonscientific events, and even the related tweets already published on social media platforms may influence the health beliefs of the general public on social media to some extent,” the researchers said.

Using natural language processing (NLP) and machine learning techniques, a total of 92,687,660 tweets from 8,967,986 unique Twitter users were analysed; 5,000 such tweets were manually annotated to train the models. Each tweet was evaluated to gauge the user’s health beliefs, employing the health belief model (HBM) with four core constructs: perceived susceptibility, severity, benefits, and barriers.

The researchers also took an epidemiology approach to quantify what they deemed to be an “infodemic.” They applied the susceptible-infectious-recovered model, where those who tweeted about COVID-19 were deemed susceptible, while tweeting about health beliefs, as classified by the HBM, were defined as infected. Recovering meant that the user stopped tweeting about health beliefs. The basic reproduction number (R₀) was also calculated.

After removing those that were not in English, 51,792,817 tweets remained available for analysis, of which the HBM identified 5,585,780 that were related to health beliefs. [J Med Internet Res 2021;23:e26302]

Over half (n=3,058,121; 54.75 percent) of the HBM tweets were about perceived susceptibility to COVID-19, while 40.08 percent (n=2,239,038) were about perceived severity of the disease. Treatment-related beliefs—the perceived benefits of (0.04 percent) or perceived barriers to (0.03 percent) hydroxychloroquine (HCQ) or chloroquine (CQ)—only constituted a small minority of HBM tweets.

From January to June 2020, there was an explosion of HBM-related tweets, yielding an R₀ estimate of 7.62 for users who tweeted about their health beliefs.

In particular, tweets about perceived susceptibility to the disease would spike in accordance to global COVID-19 case increments. Tweets about perceived severity likewise shared a correlation with death trends. Public health interventions also seemed to alter disease-related health beliefs.

“Through further evaluation of HBM-related tweets, our findings demonstrated that trends in health beliefs were correlated with dynamics in positive case and mortality rates,” the researchers said. “Additionally, we observed a decline in perceived disease susceptibility during government-issued lockdowns, while perceived severity appeared unaltered.”

They then took a selection of events, categorized as scientific (such as FDA announcements, statements by medical organizations, or study publications) or nonscientific (including pronouncements of politicians, news reports, and personal tweets of personalities), and looked at their respective impacts on health beliefs and tweets.

Both types of events triggered fluctuations in treatment-related health beliefs. Where nonscientific events tended to cluster around big spikes in the number of HBM tweets, scientific events were more distributed throughout the timeline, coinciding with many gentler changes in beliefs.

The Kruskal-Wallis test showed that scientific and nonscientific events had comparable influence over the perceived benefits (p=0.78) and barriers (p=0.92) regarding HCQ and CQ.

“It is ‘unhealthy’ that both scientific and nonscientific events constitute no disparity in impacting the health belief trends on Twitter, since nonscientific events, such as politicians’ speeches, might not be endorsed by substantial evidence and could sometimes be misleading,” the researchers said.

“These findings lay the groundwork to better understand how the general public’s COVID-19-related health beliefs are influenced by case and mortality rates, government policies, current news, and significant events,” they added. “Through careful study of these observations, we may better implement management strategies to combat the pandemic and the ‘infodemic’.”