Every day, millions of people around the world go to the search engine Google to look for information. When they're sick, they look for information about their illnesses. Now Google says all those searches can help predict when flu epidemics are happening - maybe even sooner than public health officials can learn about them using traditional data gathering. Rose Hoban reports.

In the United States, the Centers for Disease Control and Prevention - or CDC - maintains a surveillance system which talks to doctors across the country every week. The doctors report what percentage of their patients that they have seen have symptoms of the flu. Then epidemiologists at the CDC combine all this data and are able to survey how widespread the flu is.

Meanwhile, Google software engineer Jeremy Ginsberg says Google noticed that during flu season, searches about the flu and its symptoms increased sharply. So he compared five years of Google's data with five years of data from the CDC.

"We compared this week by week, region by region, and we found that there are some themes which occur whenever flu season is popular," he says.

"Effectively, by counting and measuring how relatively popular these things are, we can make accurate estimates of how much flu there is in every state across the U.S.," Ginsberg says.

Not only did the Google data match that of the CDC's, but it tracked flu trends in a more timely manner.

"It takes about a week or two for all of the doctor reports to trickle in. And so by the time they are collected, another week or two has passed," Ginsberg says. "Because we make our data available so quickly, we were able to detect the start of flu season one to two weeks before the CDC's own data."

Using this concept, Ginsberg and his co-workers created an application called Google Flu Trends. At www.google.org/flutrends, Americans can see whether lots of people in their communities are searching for information about the flu - and get a clue as to whether influenza has come to town.

Ginsberg says Google plans to gather U.S. data for a year or two. Then he wants to expand the application to help predict flu trends in other countries, too.