Google Search Trends Reveal Flu Outbreaks
With flu season fast approaching, people who feel themselves getting sick are turning to the web to self-diagnose. Aggregating Google's flu-related seraches can yield interesting trends about where outbreaks are occurring — and might even provide earlier regional warnings than the CDC.
Miguel Helft, who wrote an article in The New York Times about Google's Flu Tracker, talks about the information that can — and can't — be gleaned from aggregated search data.
NEAL CONAN, host:
This is Talk of the Nation. I'm Neal Conan in Washington. Here are the headlines and some of the stories we're following today here at NPR News. Treasury Secretary Henry Paulson reversed course today saying the government will not buy up bank's troubled assets. Instead, he said the administration will use part of the bailout money to buy stock in banks in an attempt to encourage them to lend. Also today, the Dow Jones industrial average is down more than three percent, down 304 points. And the Supreme Court heard a case today on a religious group's attempt to place a monument in a public park in Utah. The Summum religious group argues that once a government accepts any donations for display in the public park, it must accept them all. Details coming up later today on those stories and of course, much more on All Things Considered.
Tomorrow on Talk of the Nation, many Americans agree this election changed race relations in the US. They also agree there's more work to be done. Your calls on the election and race relations, what's changed, plus Howard Dean, the outgoing head of the DNC, joins us tomorrow on Talk of the Nation from NPR News.
And from tracking voters on the web to tracking influenza when we feel crummy, fever, chills, muscle aches, runny nose, those ominous flu-like symptoms. Before we call the doctor, many of us self-diagnose and start off to Yahoo or Google to search under a few predictable keywords - thermometer maybe, muscle aches, congestion or just plain flu symptoms. I have a few. And when a lot of people in the same city search those terms, they may inadvertently provide important public health information. Search peaking with Google, track flu symptom searches against actual flu outbreaks, and guess what? Here to tell us more about Google Flu Tracker is Miguel Helft. He covers Google and Yahoo for the New York Times business desk and joins us today from the studios of member station KQED in San Francisco. Nice to have you on the program.
Mr. MIGUEL HELFT (Reporter, New York Times): Thanks for having me.
CONAN: And how closely did the outbreak of searches under flu symptoms track actual outbreaks of the flu?
Mr. HELFT: Very closely actually. Google - what it did is create a basket of these keywords that you just mentioned that are related to flu symptoms, the types of things we might normally type into Google when we get sick and the things that we might do even before we call our doctor. And Google, of course, has a database of all the searches that have been carried out on its search engine. And based on that idea that if a lot of people are sick, a lot of people will be searching for these keywords. It created a basket and tracked the basket over five years and then it matched it against the CDC data on actual flu incidents and it found the correlation to be very, very strong and the advantage, of course.
CONAN: Excuse me. I hate to interrupt. It's retrospective, if they have to use it prospectively to say look out, New Mexico.
Mr. HELFT: Well, yeah, that's exactly the idea. They - not only they found that it correlates with the CDC data, but because the Google data is in real time, you know, they have this data as soon as you type something in. They know that you did so. It was able to predict the spread of the flu about two weeks before the CDC reports did. And so yesterday, they unveiled this tool that's available on the web and you can go look at it and see if flu is spiking or not in your state or your region of the country.
CONAN: Right now, we just looked at it earlier today. It's low in most of the country, moderate in Maine, Arkansas, Mississippi, Kentucky and Hawaii. So what is the benefit if you know ten days or two weeks in advance that there might be a flu outbreak in your area?
Mr. HELFT: Well, the public health officials I talked to said, you know, the sooner we know about the flu spreading, the sooner they could take measures, perhaps inform the public through the media to take precautions or to make sure they get vaccinated, inform doctors and hospitals so that they know when patients come in, that you know, they might have the flu and - so, early detection is always good for prevention and measures by public health officials.
CONAN: We're going to bring a doctor into the conversation in just a moment. But we also want to hear from you. Do you think Google's tracking system would inspire you to protect yourself if you saw there was going to be an outbreak in your area? Wash your hands more, get stocks of Tema flu, whatever. Our number here in Washington is 800-989-8255. Email us, email@example.com. And you can leave a comment on our website, too. Just head over to npr.org and click on Talk of the Nation. And now, Dr. Farzad Mostashari joins us. He's the assistant commissioner of the New York City Health Department and he joins us by phone from his office in New York. Pleasure to have you on the program today.
Dr. FARZAD MOSTASHARI (Assistant Commissioner, New York City Health Department): Hi, Neal. Thank you very much for having me.
CONAN: And is this information that Google Flu develops is of interest to you?
Dr. MOSTASHARI: Well, it's very interesting in terms of kind of a new way both of monitoring information, also the visualization tool. And frankly, maybe the most interesting part of this is for the first time, it helps us get information from and hopefully give information directly to the public.
CONAN: How would you give information directly to the public as a result of this tool?
Mr. HELFT: Well, yeah, one thing that might be interesting would be, we haven't done this yet but now we're all just kind of thinking about possibilities is if someone types in a search term for something that might connote an illness, we might have public health ad up on the top right corner and ad that says this is what you should do or maybe you should go to a doctor if you have these symptoms. Or we're concerned about these symptoms occurring together. If someone, if you can imagine, several years ago were to have typed in Vioxx and chest pain, maybe it would have been useful to have been able to get direct input from the patients about maybe this is something that should be reported to the FDA.
CONAN: And Miguel Helft, those pop-up ads both in Google and Yahoo! - those search engines are smart enough, I assume others are too - that if you do type those, they have those ability to specify ads that would pop-up and inform you maybe that there's a flu outbreak in your area.
Mr. HELFT: Yeah. The search engines already do that very well. I mean, they correlate the ads that they show you with the terms that you type into the search engine so that's definitely a possibility.
CONAN: And as you look at this tool, as I understand it, Dr. Mostashari, the CDC already get a lot of this information and about the same time that Google does. They just report it a little bit.
Dr. MOSTASHARI: There's traditional food surveillance which relies on manual reporting. Over the past seven or eight years, there are literally of state and locals as well a federal program to get ads electronic information for real-time to these surveillance tapping in to things like pharmaceutical sales, emergency rooms and so forth. So these are not, in terms of the specific application. It's not the first time where public health officials have had access to quicker information than traditional surveillance using existing electrons out there. But I think what is new about this particular application is that it gives a consistent national view to CDC or to the public.
CONAN: And we'd like to get the public involved in this conversation. If you have questions about this Google Flu Tracker, if you think you might use it, give us a call, 800-989-8255. Email us, firstname.lastname@example.org. And Miguel Helft, these kinds of search engine, I think it's data mining in a way. They're often used by people who are interested in marketing much more than public health.
Mr. HELFT: Absolutely. Google and Yahoo! and all these companies have this data, all sorts of data, they collect as we make our way through the web. Some of the most powerful data is the data in the search engines themselves because that is when we do a search, we're often telling the computer some of our most personal things. We search for baby names because we're about to have a baby. Or we search for a symptom because we suspect we have a disease. And the companies have been using these data internally for marketing purposes. Yahoo! is using it, for instance, to find out what people are interested in at a certain time and then program its website accordingly and the question is what will happen if more of these data gets released to the public?
CONAN: Let's get a caller on the line. This is Stephen and Stephen is with us from Goldsboro in North Carolina.
STEPHEN (Caller): Hi. Yes, I have a question for the gentleman representing Google.
CONAN: There's nobody representing Google. Miguel Helft works for the New York Times. But go ahead.
STEPHEN: OK. Anyway, regardless, my question is: this information, will it not encourage people to go online and look for terms regarding the flu stuff and how that's going to affect the statistics that Google is going to take? I mean, will it show false outbreaks because people are looking for the application rather than actual fly symptoms?
CONAN: It's an interesting question, Miguel Helft, because this is something that works sort of inadvertently. People do this on their own. It's this idea called collective intelligence.
Mr. HELFT: Yeah. Absolutely. And the Google people as well as the CDC which helped put this altogether where they cautioned us that past success is not is not a predictor of future success and they're going to be monitoring this tool over time. The example that the caller makes is that could be for instance a newspaper article saying that there's very little flu and that sends people to the web to search for flu and perhaps, it would skew the data. So I think they're going to be monitoring this and making sure that it still correlates going forward.
Dr. MOSTASHARI: This is Farzad. I do want to support the caller's question. We've seen when we monitor, for example, over the counter medication sales that there is poor absenteeism, things like the day before the big baseball game or when something goes on sale. There are many ways in which this nonspecific information can create false positives. And I think it raises the really important question of what we need to think about is how would we respond when we see an unusual trend? And I think that question, we should think about it both in terms of the public as well as for public health.
CONAN: That's interesting. You monitor for example if there's a spike in DayQuil sales, you worry about an outbreak of a lot of colds or flu?
Dr. MOSTASHARI: That's right. That's part of these nontraditional surveillance systems that I mentioned that also have provided an early warning once or two weeks before the traditional flu data.
CONAN: So this might buttress other sources of information. They're not going to stop gathering other data.
Dr. MOSTASHARI: That's right.
CONAN: It's interesting. And as you look at this Miguel Helft, it does have applications not only for marketing and other things but for other public health applications too.
Mr. HELFT: Yeah. The hope of Google and some of the other public health people I talked to is this might be useful to track other diseases. And Google, already makes - for about two years, it's had a product called Google Trends where anybody can go to their website and type in any keyword and see the relative popularity of those terms, how they ebb and flow. And a lot of people are already using that to make predictions, if you see a lot of queries going up for HDTV, you would suspect that people are shopping for those things. If you saw a lot of searching for HDTV discount, you suspect people are looking for bargains. And so I think this data is quite valuable and is being used more and more.
CONAN: Stephen, thanks very much for the phone call.
STEPHEN: Thank you.
CONAN: We're talking with Miguel Helft of the New York Times about the Google Flu Tracker. Also with us is Dr. Farzad Mostashari who's the assistant commissioner of New York City Health Department. You're listening to Talk of the Nation from NPR News.
And Dr. Mostashari, let me ask you also, don't you have to be wary as a lot of us will use the web to search for flu symptoms but a whole lot of communities have much less access to the web than others do?
Dr. MOSTASHARI: I think that's certainly true although I think the algorithms that Google is using, adjusts for the overall volume of searches. We actually haven't been able to get kind of under the hood as it were, exactly how the algorithms and statistical methodology works. But at least in terms of what we've been able to see, it looks like it's done a nice job of adjusting for those kinds of things.
CONAN: And adjusting for places where there is not an awful lot of people who live there.
Dr. MOSTASHARI: Yeah. And if it's anything like Google Trends, there's going to be a limit to how small an area you can look at in terms of the minimum number of volume of searches.
CONAN: So it could be by state or by county or by large metropolis?
Dr. MOSTASHARI: That's right. Currently, it's at the state level, I believe, but we would - at least in New York City - there's probably going to be enough queries to theoretically be able to give us.
CONAN: Probably get down to zip codes.
Dr. MOSTASHARI: That's right.
CONAN: Let's get Nancy in the line. Nancy is with us from Minneapolis.
NANCY (Caller): Hello. Hi. Thanks for taking my call. I listen to the Talk of the Nation everyday.
CONAN: Thank you very much for that. But it sounds like you've got a problem.
NANCY: Well, I've got what I believe the influenza and my husband came down with it a week ago prior to myself. We had missed our shot clinic on October 25th and by next the week, we were both ill and I'm telling people get the shot because this is the worst I have felt since I was a kid.
CONAN: That's an interesting. We've had an email from Andrea in Salt Lake City, Utah. How funny it is to hear to your topic today this morning. I thought I might have the flu. I disinfected my whole office workspace and my neighbors' before they arrived to work. I just looked at Google's flu tracker and will continue to look at it through the season. I hate being sick. Nancy, I think you might want to say amen to that.
NANCY: I would say amen and I wish I had known about it because we would have rescheduled our flu shots immediately if I had known this was going to be what it was like.
CONAN: Well Nancy, we hope you feel better.
NANCY: Yeah, thank you for taking the call. And everyone, get a shot.
CONAN: OK. Thank you Nancy.
Dr. MOSTASHARI: Thank you, Nancy. Yes!
CONAN: Let's see if we can get John on the line. John is calling from Cleveland in Ohio.
JOHN (Caller): Hello. How are you?
CONAN: Very well. Thank you.
JOHN: My question is that putting me on the emergency issue. We're often on the front lines of people seeking refuge from the flu which we have really a limited arsenal. If they're coming in within a day or two, you can prescribe them Tamiflu or one of the other antivirals which really is more akin to a marketing success in what I've seen that translates in a clinical success. But this is all day to day to day and I think most epidemiologists, people with a CDC, anyone, that data is something but information is really what data needs to be and if it's not really translatable information, what's the point or the purpose?
CONAN: Dr. Mostashari, what do you think?
Dr. MOSTASHARI: I completely agree. We do have to when we say we have to think about how the public can respond to these increases. We should all recognize despite some of the news releases, about the press coverage of this, people should get their flu shot well before flu is epidemic in their communities.
CONAN: Nevertheless, some people will always be lagers in this regard. And maybe one of these announcements might get them to take their flu shot. But also I'm told that one of the best defenses is simply to wash your hands frequently during the day so you don't necessarily pass it to yourself.
JOHN: Good prevention in most diseases is worth the pound of cure. But again, when it translates to the patient frequently is they feel so bad, they come in for an average, for an instance an average of emergency department bill for just a nursing care I'd say you'd get two medicines say you get a medicine for nausea and maybe four dollars or something for the aches and pains from which you're suffering and a bag of IV fluid. The average charge for that is $954 not withstanding if you have respiratory symptoms or older chest x-ray is which is about 195 as well as then the physician charge, if it's a level two or three, you're looking at least $180. So it really translates into, again, the information. How does that translate into saving health care? The patient, those with or without insurance offers $1,500. I need to look more into the site but it seems like a great idea. Just maybe it's a tool that we need to find the right bolt to help it screw down some things.
CONAN: Well interesting, thank you very much for the input, John. Appreciate it.
JOHN: Thank you.
CONAN: There's another example of, I guess, collective intelligence. Miguel Helft, thanks very much for your time today.
Mr. HELFT: Thanks for having me.
CONAN: Miguel Helft covers internet companies including Google and Yahoo! for the business desk of the New York Times. His story on Google Flu is on the front page of today's editions of the newspaper. And our thanks, too, to Dr. Farzad Mostashari, the assistant commissioner of the New York City Health Department with us from his office in New York City. Appreciate your time today.
Dr. MOSTASHARI: Thank you and thanks to your callers.
CONAN: And let's see, it's Talk of the Nation. Get your flu shot. That is good advice. Get your flu shot. This is Talk of the Nation from NPR News. I'm Neal Conan in Washington. Transcript provided by NPR, Copyright NPR.