RSSFlu-gle? Google project maps the misery
By Angela Gunn | Published November 12, 2008, 5:51 PM
Stop calling your mom to ask her if those sniffles and aches mean you've got the flu. You're going to mess up Google's stats.
After a year of testing that involved correlating data and tweaking algorithms with the Epidemiology and Prevention Branch of the Influenza Division at the US Centers for Disease Control and Prevention (CDC), Google's got a new map that shows where people are searching for info on flu-related topics. According to its tests, the map can accurately estimate current outbreak levels two weeks faster than traditional flu-monitoring systems such as the CDC's own reports.
Google Flu Trends (it's a shame the search giant didn't snatch "flugle.com" away from the domain squatters) tracks certain high-relevance search terms, maps them by IP address, and puts forth the data in map form (and, for the incurably wonky, in downloadable files). The data is aggregated from searches done by the estimated 90 million Americans who search on medical issues every year.
It may sound like shaky science -- after all, anyone in any condition can Google "flu" -- but comparisons begun in the vicious 2003-04 flu season bear out a close correlation between what Google sees and what the CDC sees. Even better, since Google can process their data very quickly, daily reports are possible. That kind of early detection could keep more people from getting sick -- or at least provide an excellent reason for you not to make that business trip to Maine this week. (Feel better, Mainers!)
The implications for scientists and those working to contain epidemics (or the long-dreaded next pandemic) are sufficiently impressive to have gained five Google employees, along with a CDC scientist, a paper likely to be published in the prestigious journal Nature. A draft of "Detecting influenza epidemics using search engine query data" (PDF available here) goes into great detail about the project's findings, privacy-protection measures, the scope of the inquiry -- 50 million potential search queries were narrowed down to 53 -- and various factors that might taint the stats, including the dreaded "media coverage of the project."
Google Flu Trends is a production of Google.org, the company's public-good arm. The company has pledged not to sell advertising on the Flu Trends site, and there's no linkage between Flu Trends and the Google Health records-keeping project.
That kooky google just knows everything, dont they....
Score: 0
|see here is the problem with this.... when showing location, it only shows the location of the isp branch office, as that is where location tests show. so that means this map is only showing the location of isp's with infected users. big deal, when i try to find the location of my ip address, it shows me 50~ miles away. that could be a bit of a problem for pinning down flu outbreak areas...
Score: 0
|The problem is that the sampling size is far too large that it does not accurately reflect the extremely variable nature of the distribution within a given overly generalized region.
And like Skimore accurately mentions, it simply serves as yet one more way to report another crisis, even though the reputed crisis may not even affect the area within the reported region at all!
In this case its simply a case of over generalization over substance!
Like the news, in order to gain attention, everything is turned into an iminent crisis - even if the probability of occurance is infinitesimally small!
Now everyone run and hide under your bed - that is if you have been able to crawl out from under it since the last pseudo-crisis.
Score: 0
|I'd blame terrorists, Muslims or Europeans for any of these issues if I was you. You should listen more to aunty Sarah.
Score: 0
|There you have it folks, spoken like a complete idiot.
Don't you have your own small problem with cartoons, clown?
Boom!
Score: 0
|Howdy foxfyre -- gotta disagree with you re the stats. It was too much to delve into in this rather short article, but the way they're handling regional distribution is pretty impressive. Highly recommend the draft of the Nature paper, linked in the penultimate paragraph; give it a look and see what you think of their methodology. Some nice statistics work happening on this project.
As for myself, I do wish -- since it's theoretically possible, considering how they're figuring location -- that the map were more granular. I don't suppose zip-code-level mapping would be something they'd want to undertake (especially since part of the appeal of this project is the 24-hour turnaround, which as you know is practically instantaneous compared to the CDC's two-week delay), but it would be fascinating.
(And why does it have to be a crisis to want to know where the flu is kicking? Maybe a person just wants to take some extra precautions, or at least stuff some extra tissues and hand sanitizer in one's backpack, the way us hay-fever types might when the pollen report is nasty. Or maybe people who work with fragile populations -- I'm thinking of my grandmother's nursing home -- would like a little warning so they can plan for possible staff shortages. Not every piece of bad news has to cause pandemonium! That's like hating the Weather Channel 'cause it says you might get some rain. Don't blame the data for how folks react to it...)
Score: 0
|Great another way to freak everyone out about something that might or might not be happening..
Score: 0
|