How Google Can Get the Flu Right

An anonymous reader writes Google Flu Trends was developed in 2009 to improve forecasts of flu levels in the U.S. by utilising Google search data. This early example showcased the potential which lies in the exploitation of human digital traces which all of us leave behind by using online services. The rise of Google Flu Trends was only stopped when the service dramatically overestimated the number of flu incidences recently. The fall raised questions about the value of online data for predictions in general. However, a study published yesterday demonstrates that it is not only about data but also about the adaptiveness of algorithms used for predictions. Scientists combined historic flu levels as reported by the CDC with Google Flu Trends data using an algorithmic framework which is able to adapt to changes in human search behaviour. Their results show that Google Flu Trends data sets significantly add information to the forecasts of current flu levels.

Comments

I'm trying really hard

[qrwe]

...to come with a good anti-virus joke.

Re:I'm trying really hard

[Anonymous Coward]

McAffee or Norton?

  Oh wait, those are bad jokes.

Re:I'm trying really hard

[Applehu Akbar]

So THIS is the Google-flu that people keep claiming they have!

overtraining?

[goombah99]

we will see if this works next fall

Indirect measurement

[Dan East]

This is a very indirect way to measure and thus predict flu outbreaks. There are many things that could prompt people to search for and otherwise discuss the flu, beyond a person or their family actually having the flu. All it would take is a few major news sources to make mention of the flu, and there's a good chance there would be a surge in people searching for terms related to the flu. Further, other illnesses could be going around that have some flu-like symptoms which would also increase search results for the flu. Can't we just get this information from healthcare providers or insurance companies directly? (although the latter is probably charging an arm and leg for that information)

Re:Indirect measurement

[Njorthbiatr]

I don't think Google was wrong in underestimating the number of cases, they just had trouble putting the human element. How many people are actually going to go to the doctor?

It assumed people would do the sane thing and consult a medical professional instead of trudging into work chugging DayQuil with a box of tissues so they're not seen like a slacker (because it's just a cold and you can't skip work for just a cold, c'mon).

Re:Indirect measurement

[Bite The Pillow]

We don't want hard data, we want predictive data that health care or insurance companies don't have yet. Or maybe charge too much for.

So scientists, linked in the article, are trying to add data points outside of searches and make better predictions.

In fact, they may have read your comment and gone back in time to address your concerns. Well done!

Re:Indirect measurement

[aztracker1]

Not to mention the fact, when people have a common cold for a couple days, they often refer to it as "having the flu" ... Having gotten the flu back in January, and the two and a half weeks it took me out for, I can say it really isn't that mild generally.

Re:Indirect measurement

[aztracker1]

I think the opposite happens far more, that people refer to a cold that has them down for a day or two as the flu so as not to seem like a slacker by the same metric. I had the flu in January, and could not really function well at all for two and a half weeks.

Google ads now serving actual viruses?

[Anonymous Coward]

RTFA is too risky given my rough calculations...

Who cares?

[Anonymous Coward]

I'm tired of people losing their shit over the flu. This didn't happen a decade ago. Unless you're elderly, an infant, or severely compromised, getting the flu is not a big fucking deal. You throw up a couple times and feel like shit for 24-48 hours *IF* you even get it. And then you move the fuck on with your life.

People act like if you don't get a flu shot, it's going to be the end of the world and you're going to die a miserable death.

Re:Who cares?

[zippthorne]

Perhaps, but if you don't get the flu shot, you're not only a higher risk of catching the flu, you're at a higher risk of transmitting it (by virtue of being a higher risk of catching it). Especially if you're a hero-worker who heroically goes in spread your mucous around.

Most of the flu hysteria I've seen has been more anti-vaccine than anti-flu.

Re:Who cares?

[Anonymous Coward]

anti-vaxxers can kiss my ass. There is a surge of whooping cough cases in 90210.

Re:Who cares?

[slimjim8094]

250,000-500,000 people a year die from the flu, more than 50,000 in the US (that's more than both traffic accidents and gun deaths combined). It's not something to fuck around with.

It's sweet that you think the people giving you a hard time about the flu shot care about you. Sorry to burst your bubble, but nobody cares about you. They care about your elderly neighbor or the kid a few seats away on the bus, or the coworker with the pregnant wife. And you can and will give it to them before you even know you have it, so it's not a question of "well *I'm* not the problem" - yes, everybody without a flu shot is potentially the problem, because they can get the flu and infect others before there's any hope of avoiding it. Before there was an effective vaccine that was cheap and easily available, the flu was just something that happened - but now that there's something to do about it, choosing not to do it is a pretty dickish move.

The flu is scary. Lots of people think Ebola is scary - sure, the flu has a lower case-mortality rate, but it's compensated for by how easy it is to spread, so overall even if we pretend this Ebola outbreak is typical, the flu is still like 100x more deadly worldwide, and something like 50,000x more deadly if we just look at the US. The 1918 flu killed 50-100 million people - that's 3-5% of humans who were alive at the time, disproportionately affecting healthy young and middle-aged adults. It killed 10x as many Americans as did World War 1. Sure, sanitation and treatment weren't what they are now, but if you get the flu, you can't do much more than supportive therapy. If that flu happened today, the death toll would undoubtedly be lower than it was, but most people would still know someone who'd been killed that year by the flu. That is, without a vaccine.

Re:Who cares?

[Anonymous Coward]

Except thimerosal a.k.a. mercury ethyl, and mild flu symptoms lasting 1 to 2 days, and cost, for a vaccine that may not even work since each season's vaccine is an estimated guess. Sorry for not feeling dickish.

Re:Who cares?

[slimjim8094]

3 estimated guesses, actually, based on the flus most likely to be seen in North America. The flu is a rapidly-changing and complicated virus - all the more reason to reduce its chances of mutating back into something really dangerous, eh?

... blah blah blah stupid thiomersal conspiracy ...

Oh. You're not careless, you're just an idiot. Carry on - just stay away from, you know, people.

Re:Who cares?

[aztracker1]

You do realize that the flu killed more people in the 1900's than all the explosives, bullets and other tools of war combined, right?

Re:Who cares?

[bigpat]

250,000-500,000 people a year die from the flu, more than 50,000 in the US (that's more than both traffic accidents and gun deaths combined). It's not something to fuck around with.

There are many different viruses and bacteria (including Ebola) that have flu-like symptoms and based on the summary of methodology for these CDC numbers I don't believe the CDC is doing enough regular randomized testing with controls to determine how many of those flu deaths are actually "the flu" or something else with "flu-like symptoms".

There should be two types of randomized testing. First general monitoring, where the CDC pays doctors to perform blind tests on patients with certain symptoms. The tests need to be blind because testing for a specific disease will bias the results... meaning doctors will see it as a test of their diagnostic skills to get it right if you hand them a test for the flu or something else specific. Rather what we should be after is a sense of what percentage of people showing up with a cough of any kind or other symptoms have certain viruses or bacteria. Just a blind test with a direction, give to someone with a cough of any severity. Or give to someone with a fever, headache etc.

The other thing would be for people who are in hospital or who die and had any flu-like symptoms to receive such randomized testing in order to gather enough data to get a specific breakdown on the cause of death beyond just flu-like symptoms. Otherwise if you just did the randomized testing of people showing up to the doctor's office and not those who are seriously ill or die, then you would mistakenly project the percentage of deaths as the percentages of illnesses circulating at the time, when we really want to know which of the viruses and bacteria are contributing to more deaths.

Re:Who cares?

[FirephoxRising]

That's probably not the influenza virus you're describing, Most people with the real flu are sick for a week plus.

Secondary problem

[Firethorn]

The other problem I have with the AC's post is that he describes a bad cold, not a case of the flue, which lasts at least 72 hours...

Per the CDC you're infectious a day before the symptoms hit, and remain infectious for 5-7 days after that.

So if you really have the flue, you should stay home a couple days longer even if you're feeling better after 3...

Re:Secondary problem

[aztracker1]

For me it was about two and a half weeks of feeling ill, back in January (apparently a more harsh variety)

Re:Secondary problem

[Firethorn]

Yeah, generally speaking if it doesn't knock you down for around a week it was at most a bad cold, not the flu. But over two weeks is a really bad one.

Google can get the pip

[Mister Liberty]

As far as I'm concerned.

But the flu IS a big deal

[PeterM from Berkeley]

You're often sick for a week, pretty commonly you get a secondary bacterial infection (like pneumonia or a sinus infection), and then you're sick for more than a week, and you might well be left with a lingering cough.

And BTW, "throw up a couple times" is NOT that common a flu symptom, though it CAN happen with flu. Typical symptoms are:
body aches
fever/chills
coughing
runny nose, sore throat, headach, pain around eyes

Vomiting/diarrhea is more common amongst children.

--PM

Every year we get this rubbish

[Anonymous Coward]

It's a puff piece for Google. I know most people can't remember anything beyond what Fox told this this morning, but come on, every bloody flu season we get this crap about how Google can save us. Surely it's time to go back to twitter as the source for all scientific data?

So far..

[Anonymous Coward]

It looks like they're predicting a fairly small flu season this year, which seems on track with reality considering Ebola has the same symptoms as the flu early on and everyone is paranoid as shit about Ebola and taking precautions for both themselves and others. This is a bit anecdotal but usually when people get the flu they just keep going to work but this year two cooworkers actually called out sick until they were completely better, quipping about Ebola via email.

The Emperor Has no Cold

[tinkerton]

Or what was it again. Is google going to guess the length of the nose of the emperor of China next, based on search items? And has it got the flu? Look , that there's a good correlation is clear, but the reliability of the results is very low. Next the NSA will be using those tools in order to find out how much they can get away with. Or wait, they already made their minds up about that.