As we first saw in March, aggregate data from Kinsa’s network of smart thermometers can serve as an early warning system for COVID-19 spread. Since that time, even as social behaviors and policy measures have evolved, Kinsa’s illness data has continued to be an early indicator of community spread. Kinsa has published this aggregate illness data on healthweather.us since March as a public service.
Today, in order to more clearly communicate the insights from our data and help communities respond earlier to concerning trends, we are re-launching healthweather.us to visualize the results of our latest work: a prediction model that tells you, very simply, where we expect new case growth to accelerate in the coming weeks.
The model uses Kinsa’s well-known illness measures – influenza-like illness and Rt and their derivatives – and combines them with measures derived from COVID-19 case data. Working forward from the beginning of the epidemic in March, the model accurately predicts case growth 79% of the time, an average of 10 days in advance. Methods and results are detailed in our technical write-up.
We believe this new model makes our data more accessible and understandable. We expect to continue evolving the model, and the visualizations on the site, as we move into flu season.