U of G computer scientist harnesses Twitter to monitor COVID-19 and other disease outbreaks.
As we have navigated a global crisis with the COVID-19 pandemic, our government officials have been challenged to quickly identify outbreaks and provide communities with warnings that could prevent further spread. Traditionally, governments rely on records provided by health departments to issue warnings, but the system is imperfect. There can be major lag between the time infected patients are logged to when an outbreak is reported to officials to when a warning is issued for the public.
Social media sites can be valuable tools for monitoring and detecting potential outbreaks to bridge the gap between logging and public notification. These sites can also be used as a surveillance tool to warn officials about outbreaks and in subsequent public health planning. University of Guelph computer scientist Rozita Dara [1] and her team have introduced an Internet-based surveillance system that monitors activities related to diseases on Twitter, sorts through that information, and provides a supplementary source of data for our officials. Previously, the team’s work focused on avian influenza, but in response to an immediate need, the researchers have pivoted to extend their framework to COVID-19.
The team’s Twitter-based data analysis framework monitors disease outbreaks in real time. Data are collected by a web crawler, which systematically visits Twitter every minute, searching for keywords. In the case of avian flu, these included words like “h5n1,” “poultry,” and “outbreak.” The framework then filtered out irrelevant posts and analyzed the information to detect the onset of outbreaks. The researchers worked to evaluate Twitter’s reliability by assessing the overlap between information gleaned from Twitter’s daily posts and official reports. With their avian influenza research, they found that 75 percent of real-world outbreaks were identifiable from Twitter, and that the notifications reported on Twitter came earlier than official reports. Now, Dara and her team are shifting focus to COVID-19, working to help officials keep communities safe.
“We have collected more than five million COVID-related tweets so far,” says Dara. “In addition to predicting the risk and magnitude of outbreaks, we aim to predict the number of COVID-19 cases on a daily or weekly basis in selected regions.”
This work was supported by Egg Farmers of Canada, Chicken Farmers of Saskatchewan, the Canadian Poultry Research Council, the University of Guelph’s Food from Thought initiative, and the Canada First Research Excellence Fund.
Yousefinaghani S, Dara R, Poljak Z, Bernardo TM, Sharif S. The Assessment of twitter’s potential for outbreak Detection: Avian Influenza Case Study. Sci. Rep. 2019 Dec 3. doi: 10.1038/s41598-019-54388-4