Google search and social media data can predict disease outbreaks, University of Waterloo study finds

The researchers say their study could be used by public health authorities to develop a real-time surveillance system to respond quickly to emerging infectious diseases. (Sarah Blocksidge/Pexels)

Data from Google searches and social media posts can serve as an early warning about disease outbreaks, researchers at the University of Waterloo have found.

Looking at data from January to March 2020, scientists found a correlation between an increase in people Googling symptoms like cough, runny nose or loss of smell and daily COVID-19 cases.

Because people were likely to search for information on their symptoms before seeking a test, which could take days to return results, the Google search data registered the spike first.

“We were able to see that you could actually show, I would say nine to 10 days before an increase in the number of cases, that something is going to happen,” the study’s lead investigator, epidemiologist Zahid Butt, said.

“So in that sense, the signals you’re getting from Google Trends can help you as kind of an early warning system.”

The team also found a similar correlation on Twitter, although with a smaller lag time, Butt said.

“Twitter is more reactive in a sense that when something is happening then you see a lot of people posting about that particular disease. So if you look at this study, Google Trends data was better at forecasting increases in COVID-19.”

While Butt said digital surveillance like this can’t replace traditional methods of monitoring for disease outbreaks, it could help the healthcare system prepare for an outbreak.

“During the COVID-19 pandemic, what was happening was more reactive. When you had cases, then they were trying to increase the number of hospital beds or trying to add more resources to a place where they saw a lot of cases,” Butt said.

“If you’re using this system, you can look at the signals and say ‘Okay, so we’re seeing an increase in people using these sorts of symptom keywords and we can basically inform public health authorities or hospitals that there is going to be an increase in the number of cases of a particular disease… so be prepared.”

The next step for researchers will be testing if the model can be used to forecast outbreaks of other respiratory diseases with different symptoms.

Butt said they’re also investigating whether it could be used to monitor for food-borne illness or sexually transmitted infections.

Ultimately, the researchers say their findings could be used by public health authorities to develop a real-time surveillance system to flag when diseases are spiking.