Adding Google searches for symptoms and other information to COVID forecast models increased their accuracy, recent UBC research has found.
What’s Google got to do with it?
Researchers led by UBC statistics professor Daniel McDonald (he/him) fed extra COVID-related information into a basic model, on top of standard data such as case numbers and deaths. This information included outpatient visits about COVID-related symptoms, visits by patients with confirmed COVID-19, the number of people who knew someone who was sick, doctor visits with COVID-like illness, and Google search trends for loss of smell or taste.
How accurate is it?
They found each of these five ‘indicators’ improved the model’s accuracy in forecasting COVID-19 cases and predicting hotspots, buying about two to four extra days of lead-time. The team also found that the indicator based on Google searches offers a notable improvement when cases are trending “up.”
Why is this important?
This could serve as a case study for adding these indicators to models currently used by the CDC and BCCDC to improve accuracy.
- Study link
Interview language(s): English