Flu prediction tool shunned

PUBLISHED : Sunday, 24 June, 2012, 12:00am
UPDATED : Sunday, 24 June, 2012, 12:00am


Hong Kong has no plans to link up with a real-time initiative called Google Flu Trends, which claims to predict an epidemic within days

A spokeswoman for the Centre for Health Protection said the city's influenza surveillance systems were extensive and it had no plans to partner with Google's flu prediction tools.

Hong Kong has one of the most efficient flu surveillance systems in the world and works with the World Health Organisation and overseas health authorities.

Last week, the centre confirmed that the city had its longest flu season in a decade with one death per day on average.

Google's model for predicting epidemics within days compares with the several weeks needed by many national health authorities to confirm an outbreak through laboratory tests. However, infectious disease and data experts are divided on how this wealth of information could be Hong Kong's next weapon against a deadly virus.

The model tracks online searches for key words related to flu such as symptoms, medication and complications of a virus. Key words include fever, bronchitis, pneumonia, robitussin (cough medicine) and amoxicillin (antibiotic). It works with data from 28 countries including the United States, Canada, Australia and many in Europe and South America.

Japan is the only country in this region using the model. A Google spokeswoman said there were no plans for Hong Kong to join.

Dr Liu Wenyin, an assistant professor at City University's computer science department, said that health authorities should consider using the Google data to supplement their own systems.

'At least the government could try it out,' he said. 'As a scientist, I think everything should be subject to experimentation so I suggest that the two parties work together. It's definitely another way to predict a flu epidemic, but only an experiment can confirm this.'

Google Australia's head of retail Ross McDonald said in Hong Kong that Google could identify a flu epidemic in about three days 'because the very first time that people start to get a sore throat, a backache, or a headache, they turn to their computer and type in those symptoms'.

But Dr Benjamin Cowling, an assistant professor at the University of Hong Kong's School of Public Health, said Google data could lead to misinformation as viruses and their symptoms were difficult to distinguish.

'People may have flu, but the figures may also be going up and down because other viruses are circulating throughout the year, so it could be an epidemic of the common cold,' he said.

The flu season is generally February to March and June to July, but other viruses are present all year. 'There's no time of the year when there aren't viruses circulating in Hong Kong, so we wouldn't see such a nice clear pattern as that in the US and Europe,' Cowling said.

'The other problem is because the model is looking at behaviour, it could be affected by other factors like if a celebrity gets the flu, or a child dies from the flu, and it [the traffic] is just people searching for information.'