On Wednesday, Google is making county-level data on symptom searches in the United States available to researchers and the public. The information goes back three years and includes up to 400 symptoms and conditions, not just those known to be associated with covid-19. It will have information on searches for things such as stress and diabetes so experts can learn more about secondary health impacts.
The company has done some work to clean up the data for researchers, such as figuring out which search terms all describe a single symptom. For example, few people are going to type in “anosmia” when they lose their sense of smell or taste. They’re more likely to type something like “loss of smell covid” or “why can’t I taste this cantaloupe.”
All the data is anonymized and not connected to individual users, says the company. Instead of actual numbers of people looking up a symptom in a single area, Google will present the volume of searches for each symptom as a range.
The company tried to do something similar with symptom search data in the past. Google Flu Trends was a program that launched in 2008 to use symptom searches to predict flu outbreaks. It was originally accurate, but after a number of years, the results diverged from what was being found by public health organizations such as the Centers for Disease Control and Prevention.
Eventually the program was shut down. However, Google has continued to provide flu and other health search data to researchers who have used it to monitor things such as Lyme disease and the spread of foodborne illnesses.
With the coronavirus data set, Google says it is not trying to make any predictions itself, but allow outside researchers to dig in, combine it with other data for increased accuracy and learn more about what is still a relatively new disease.