Researchers saw a correlation between traffic to flu-related pages on Wikipedia and subsequent reports of illness by the Centres for Disease Control. Photo: David J. McIver, John S. BrownsteinThe number of hits recorded on Wikipedia articles could track the spread of flu and other illnesses faster than existing systems, research says.
Researchers in the United States have developed a highly accurate computer model for estimating levels of flu-like illness in the American population by analysing internet traffic on flu-related Wikipedia articles.
The research, published in PLOS Computational Biology on Friday, found the Wikipedia-based model estimated flu levels up to two weeks sooner than data from the Centres for Disease Control and Prevention became available. The Wikipedia-based model was also more accurate at estimating the timing of peak flu activity than Google Flu Trends, a service developed by the internet search giant that draws on Google search queries.
Google Flu Trends has been found to be susceptible to error at times when there is heavy media coverage of the flu, such as during the 2009 H1N1 pandemic and during the unusually severe 2012-13 northern-hemisphere flu season.
But the researchers found their Wikipedia-based model performed well even during these times. Over 294 weeks of data, the average difference between the estimate from the Wikipedia-based model and the Centres for Disease Control and Prevention data was just 0.27 per cent.
Currently the United States monitors flu cases by collecting data from almost 3000 healthcare providers, but this comes with a lag of up to two weeks, hampering efforts to distribute vaccines, staff and other healthcare resources to where they are most needed.
The researchers, David McIver and John Brownstein of Boston Children’s Hospital and Harvard Medical School, said they hoped their Wikipedia-based model could help overcome this problem.
“Each influenza season provides new challenges and uncertainties to both the public as well as the public health community,” the researchers said. “We’re hoping that with this new method of influenza monitoring, we can harness publicly available data to help people get accurate, near-real time information about the level of disease burden in the population.”
The researchers acknowledged a limitation of their model was that it was unable to identify whether article visitors were located in the United States or in other English speaking countries such as Australia, because Wikipedia did not make this information readily available.
Between 250,000 and 500,000 people die worldwide each year from the flu.
The researchers said a similar method could be used to monitor other health conditions such as heart disease, diabetes or sexually transmitted infections.