Google flu tracker had trouble with latest virus

[Sally Furniss]

Google flu tracker had trouble with latest virus

By Elise Ackerman

Mercury News

Updated: 05/01/2009 06:52:40 PM PDT

In November, Google proudly announced it had discovered a new use for its famous search engine: tracking the flu.

But the sudden appearance of the swine flu has thrown the effectiveness of Google's detection system into question.

In a paper published Nov. 21 in Nature, Google researchers said an extensive analysis of hundreds of billions of Internet searches showed that flu-related queries, such as "runny nose," "sore throat," and "dizziness," closely tracked the number of people who were experiencing flulike symptoms in particular communities, according to reports collected by the Centers for Disease Control and Prevention from doctors around the country.

Google's computers not only appeared to be accurate, they were fast. While it took government epidemiologists as long as two weeks to collect and release their data, Google could show where flu outbreaks were occurring on the fly. By providing warnings in advance, Google claimed it could help save lives.

Google was one of "the most timely, broad-reaching influenza monitoring systems available today," company researchers boasted in the Nature paper.

"Our up-to-date influenza estimates may enable public health officials and health professionals to better respond to seasonal epidemics and — though we hope never to find out — pandemics," Google engineers Jeremy Ginsberg and Matt Mohebbi, who were among the authors of the Nature paper, wrote in a post on the official

Google blog on Nov. 11.On March 10, Google CEO Eric Schmidt told students at Stanford's Graduate School of Business that by alerting health care professions to an outbreak of flu in their area, Google could help save "somewhere between 10,000 and 30,000 lives a year because we use that information to get people ahead of the cycle." Schmidt said Google's data gave as much as six months' lead time.

But when the swine flu began sickening hundreds of people in the Mexican town of La Gloria in the Veracruz state in early April, Google Flu Trends missed it. There wasn't a significant increase in search activity on Google's "Experimental Flu Trends in Mexico" Web page until April 19, three days after the Mexican government had alerted the World Health Organization about a rash of flu and pneumonia.

"If you are looking for Google to give you a clue about the swine flu, I don't think you are going to get it," said Henry L. Niman, a biochemist who studies the molecular evolution of flu cases.

"The system isn't going to register such a small amount of people," said Jamie Yood, a Google spokesman who defended the service. "We think Experimental Flu Trends in Mexico shows an increase in flu activity before media reports of swine flu began."

Some other professional flu trackers agree with Niman. Sharon Sanders, who runs Flutrackers.com, said the hundreds of volunteers who contribute to her site do a better job than Google Flu Trends by relying on news reports about the flu in countries around the world. "Hopefully, humans are better than the machines," she said.


http://www.mercurynews.com/business/ci_12275174

 

[sharon sanders]

Re: Google flu tracker had trouble with latest virus

There goes any chance of a grant from google! lol

 

[sharon sanders]

Re: Google flu tracker had trouble with latest virus

We noticed some strange respiratory and hospital activity in Mexico on April 17th by searching original articles in Spanish.


http://www.flutrackers.com/forum/showthread.php?t=99635

 

[BeauD]

Re: Google flu tracker had trouble with latest virus

Check out Hawaii. Yet there are no confirmed cases of H1N1 in Hawaii. Color me skeptical.

 

[Laidback Al]

Re: Google flu tracker had trouble with latest virus

. . .

"If you are looking for Google to give you a clue about the swine flu, I don't think you are going to get it," said Henry L. Niman, a biochemist who studies the molecular evolution of flu cases.

"The system isn't going to register such a small amount of people," said Jamie Yood, a Google spokesman who defended the service. "We think Experimental Flu Trends in Mexico shows an increase in flu activity before media reports of swine flu began."

Some other professional flu trackers agree with Niman. Sharon Sanders, who runs Flutrackers.com, said the hundreds of volunteers who contribute to her site do a better job than Google Flu Trends by relying on news reports about the flu in countries around the world. "Hopefully, humans are better than the machines," she said.

Computer tabulations and frequency counts on key words such as "runny nose," "sore throat," "dizziness," or "atypical pneumonia" will never presage the start of an epidemic or pandemic. While Google touted its ability to give planners a six month lead time, in truth, the Google experiment couldn't even give a six day lead time for this outbreak of S-OI A/H1N1.

Yood, the Google representative, suggests that the system can't register small amounts of increases in the searches. However, all novel influenza outbreaks will begin with a small number of cases.

Another false assumption is that the once such an outbreak occurs sick individuals will turn to the internet to search for information about treatment. This is a very western ethnocentric point of view. Some cultures have different perspectives about what causes diseases and how they are treated. Why does Google believe that people in the small village of La Gloria in Mexico would immediately sign on and search on certain key words relating to their health conditions rather than consulting with local HCWs? In fact, one could ask how many of the people in the village had computer access to the world wide web to even search for information on their symptoms.

FluTrackers members stumbled on news reports of the first apparent, or at least the first recognized, death in Mexico associated S-OI A/H1N1. A search for "atypical pneumonia" pointed to an article on a 39 year old woman from Oaxaca who died on April 13. The news article was published on the 16th of April and was translated and posted here at FT on the 17th.

Within 2 days, FluTrackers members recognized that Mexico represented a developing situation that required further attention. About the same time, the first two cases of S-OI A/H1N1 in southern California were confirmed and reported by the CDC. It was Niman, who is quote in the article above, that was the first to hypothesize in this post here at FluTrackers that the California cases and the Mexico cases were infected by the same strain of the virus that we are now calling S-OI A/H1N1. Not only did Google fail to identify the initial outbreak in Mexico, it also failed to make the significant connection that the cases in the USA and the outbreak in Mexico resulted from the same novel strain of A/H1N1 virus.

The human element, the volunteers here at FluTrackers, are clearly better at discerning trends and making inferences than the Google translation machines. What all of the volunteers at FluTrackers bring to the process that Google can never duplicated is the ability to evaluate and discriminate important facts from filler copy in news articles from around the world. The volunteer participants at FluTrackers come to the site with a diverse set of skills, background, experience, and knowledge that can not yet be duplicate in a computer algorithm.

These members, from all around the world, deserve a round of applause and our gratitude for their tireless effort in searching for information on emerging infectious diseases in hopes of reducing world-wide morbidity and mortality.