by Bill Snyder

Google’s Panicky Flu Estimates Were Dead Wrong

Feb 14, 20132 mins
Consumer ElectronicsInternet

Google Flu Trends overestimated the spread of the flu epidemic by nearly 100 percent. So much for the wisdom of crowds.

Don’t you just love it when the smartest kid in the class (or office) gets one wrong? Really wrong. Of course you do. So when Nature, the respected science publication, found that Google Flu Trends had wildly overstated the prevalence of flu cases this year, I have to admit that I felt a little pang of satisfaction.

Google Flu Trends actually has had an excellent record of keeping tabs on the spread of influenza over the years, and is a model for similar systems around the world. But this year, it was off by an embarrassing 100 percent when its estimates of flu cases in the U.S. were compared to the traditional data gathering by medical professionals across the country, according to long-time Nature writer Declan Butler.


Flu Trends, as Nature explains, “relies on data mining records of flu-related search terms entered in Google’s search engine, combined with computer modeling. Traditional flu monitoring depends in part on national networks of physicians who report cases of patients with influenza-like illness  — a diffuse set of symptoms, including high fever, that is used as a proxy for flu.”

So why did Google blow it so badly this year? In a word, panic. This year’s flu season started early and was predicted to be quite severe. Although it’s not possible to know for certain, it appears that the widespread publicity around the outbreak, and actions like the declaration of a public-health emergency in New York, scared people who, like many of us, quickly went online and started searching for information about the disease.

It’s likely that many of those people weren’t ill, but by performing searches that the algorithm assumed were made by people who have flu-like symptoms, the data was skewed. Google declined to comment to Nature about the error, Butler said.

A similar error occurred in 2009, when Flu Trends badly underestimated the prevalence of that year’s outbreak, the very scary swine flu pandemic. That time, though, the publicity apparently changed the way people were searching, which also fooled the algorithm, but in the opposite direction, Nature reports.

I don’t mean to pick on Google, and Butler notes that public health experts are still quite bullish on Flu Trends. But this mishap is a good reminder that the “wisdom of crowds” isn’t always all that wise.

Photo: John Angelillo/UPI/Newscom