Google Predicts Spread of Flu Using Huge Search Data

Google has applied its massive data-collecting power for the first time to the prediction of the spread of disease, with the launch of a new site that claims to be able to raise the alarm over flu outbreaks up to two weeks in advance of existing public services.

Google Flu Trends takes the general search tracking technology pioneered by Google trends and applies it specifically to influenza. The firm's engineers claim to have devised a way of analyzing millions of individual searches related to the disease that in tests proved to correlate closely with the actual incidence of illness.

That gives them the potential ability to predict spikes in flu cases that in turn could be used by health professionals to warn the public or plan their responses.

Google found that if it assembled a cluster of queries people used when they were worried about flu - such as "fly symptoms", "chest congestion" or "where to buy a thermometer" - the aggregated trends were a strong indicator of flu levels across America.

"We wanted to step back and see if we couldn't model a real-world phenomenon using search query data," Jeremy Ginsberg, a Google engineer involved in developing the new system, told the Guardian's science podcast. He added that flu had been chosen as the pioneer illness to be tracked as it was a serious disease that killed hundreds of thousands of people around the world each year.

To ascertain the potential accuracy of the data, Google compared its figures against statistics filed over the past five years by the US Centers for Disease Control and Prevention that has a network of 1,500 doctors across the US who provide weekly reports on the number of patients complaining of flu-like symptoms.

"We found we could highly accurately estimate what the flu activity levels would be in subsequent years," Ginsberg said.

They also found that the Google statistics, which can be gathered on a daily basis, were up to two weeks ahead of the federal government's data due to the time lag in assembling information from so many doctors.

The ability to speed up the response of health services by up to a fortnight could prove invaluable in the event of a vicious outbreak, or the emergence of a new virulent strain.

The results of its comparisons with official health statistics will be published in the science journal Nature.

So far Google has only rolled out the service to the US, though it hopes to extend it to other countries, and may in time also widen the net to include other illnesses. At present, the flu data is given for each of the 50 states, though questions have inevitably been raised about whether in time it could be applied to individual cities or even neighborhoods and if so what possible side-effects that could have for local economies.

The company stresses that there are no privacy issues involved as the trends are gathered through combining millions of anonymous searches.

But civil liberties groups are already watching Google closely as it rapidly diversifies its data storage capacities.

The company has already spread its tentacles into the areas of individual and public health. It is exploring ways of compiling health information and fusing it with its other services, such as Google maps, and is experimenting with a tool that allows users to store their personal health information through Google sites.

The innovation forms part of a frontier technology known as digital detection that is designed to apply online information to public health mapping. Google recently announced that its philanthropic arm was investing millions of dollars in companies that are at the forefront of tracking the spread of diseases such as Health Map and ProMED that sound the alarm as new outbreaks occur.