- Analyzed more than 12 million COVID-19 related posts on China’s Twitter-like platform Weibo
- Found that people’s posts about symptoms and diagnosis of the disease can predict case counts up to seven days ahead of official reports.
- Users turned to social media during testing shortage
Tracking social media “sick posts” could give public health officials a head start on identifying and responding to emerging disease outbreaks, researchers at the University of California, Davis, suggest in a new working paper.
Analyzing more than 12 million COVID-19 related posts on China’s popular microblog Weibo from November 2019 to March 2020, the researchers found that people’s posts about symptoms and diagnosis of the disease can predict daily case counts up to seven days ahead of official statistics.
Broad public health implications
The findings hold broad public health implications, with social media surveillance offering a fast, low-cost way to inform disease containment and mitigation efforts, the researchers said. Tracking social media can also be done from afar.
Cuihua “Cindy” Shen, an associate professor of communication, and colleagues at UC Davis and two universities in China, used machine learning to sift sick posts from other COVID-19 related posts. Because Weibo does not provide public API access to its database, the researchers built their own pool of 250 million users — close to half of active monthly users of the Twitter-like platform — and searched their posts for 167 keywords related to the virus.
The predictive pattern of the sick posts held true for both Hubei province, where COVID-19 originated, and the rest of mainland China, though to varying degrees.
Epicenter of outbreak saw testing shortages
“Being the epicenter of the outbreak, Hubei province experienced extreme testing shortages during the early stage of the study period,” the researchers wrote in their paper. “As a result, many Hubei residents turned to social media sites such as Weibo to seek help for testing and medical care.”
The researchers found similar patterns even in regions with more health care resources and fewer COVID-19 cases. “Our finding is not merely a reflection of the help-seeking behaviors at the epicenter,” Shen said.
“As COVID-19 continues to spread across the globe, countries lacking testing and screening infrastructures will become ‘dark spots,’ endangering their own people as well as the entire world," researchers said in the paper.
“It is imperative that international organizations such as the World Health Organization integrate such data into their outbreak forecasting management practices, in order to mobilize and coordinate relief efforts to help combat COVID-19.” — UC Davis researchers
Co-authors of the paper are Wang Liao, assistant professor, Jingwen Zhang, assistant professor, and Bo Feng, professor — all in the UC Davis Department of Communication; Anfan Chen of the Department of Science Communication and Science Policy at the University of Science and Technology of China in Hefei, China; and Chen Luo of the School of Journalism and Communication at Tsinghua University in Beijing.