When Google Flu Trends launched in 2008, it was seen as a groundbreaking, and potentially more accurate way to track the spread of influenza through the prevalence of flu-related web searches. For the most part, the algorithm has paralleled the CDC's official estimates, but this year, things went haywire. As Nature reports, this year's influenza season has been unusually harsh, and Google Flu Trends has vastly overestimated its spread, with figures that are nearly twice as high as the CDC's. The company hasn't commented on the glitch, but experts suggest widespread media coverage may have skewed its estimates.
Researchers still believe the web could transform the way we track the spread of disease, though Google's snafu has underscored an important point about algorithm-based monitoring. "You need to be constantly adapting these models, they don’t work in a vacuum," John Brownstein, an epidemiologist at Harvard Medical School told Nature. "You need to recalibrate them every year."