As health officials around the world track COVID-19 infection rates, two universities in the U.S. have partnered with Facebook to try to predict infections with real-time survey data.
Dr. Alex Reinhart, an assistant teaching professor of statistics and data science at Carnegie Mellon University and a member of the Delphi Group, says that as health officials were struggling with testing capacity in the spring, they realized they might be able to predict infections by analyzing social media.
"They realized that if we could know when people are experiencing symptoms, they probably experience symptoms a few days before seeing a doctor. That's probably a few days before they get test results back and so that could potentially be an early indicator," Reinhart said.
The Delphi Group reached out to Facebook, which agreed to help them survey its users.
"Every day Facebook takes a random sample of their active users that day in the United States and internationally and invites them through a little blurb at the top of their newsfeed that says, 'you can help coronavirus research' if you take this survey, which is voluntary," Reinhart said.
Once Facebook users click on the survey button, it takes them to Carnegie Mellon's page for the survey. The University of Maryland also jumped on board with the project and conducted the survey for all international Facebook users.
Facebook does not receive any survey data and only refers to the interested participants to the survey links. So far, more than 30 million people have taken the survey.
Dr. Frauke Kreuter, who is working with the University of Maryland in Germany on the international side, says she's not aware of another global survey on COVID-19.
"There are two factors globally, I would say. One, is that many countries do not have good reporting systems and so they rely even more on alternative data sources. And the other one is, you want to compare yourself to other countries, but for that you need to have kind of the same measure in each country," Kreuter said. "And that's what we're lacking with a lot of measures right now because each country does there reporting slightly different."
So far, Carnegie Mellon and the University of Maryland have been able to develop heat maps showing coronavirus symptoms across the country and world. Reinhart says they've been on par with COVID-19 infection rates being reported from health officials and says the survey has helped them identify patterns when it comes to mask-wearing and infection rates.
"In early September, we started asking questions about mask usage and we soon found that there is a striking difference in mask usage across the country," Reinhart said. "At the time, places that had lower mask usage seemed to be having a worse time in the pandemic."
Reinhart says the survey results are helping them learn more about the effects of mask mandates. Researchers are hoping to continue the survey as the pandemic evolves — for example, they hope to begin gauging users on vaccine usage and skepticism in the coming months.
The data is available for anyone to view and use.
"It's different from what you can get from cell phone mobility data that we see people use because we get to directly ask people what they're thinking, how they're feeling, what they're experiencing," Rinehart said. "We'd like as many people as possible to discover this data and use it for their own important research questions."