Using algorithms to predict flu outbreaks

A computer algorithm originally developed to model the West African Ebola pandemic in 2014 is being used to predict flu outbreaks in Australia months in advance, and could help in the fight against bioterrorism.

Developed by Australian Defence scientists, the tool was originally used to forecast the number of people infected with Ebola up to two months in advance.

The numbers predicted by the algorithm closely matched the cases later reported by the World Health Organization.

So the scientists adapted the algorithm to model other infectious diseases, such as influenza.

“We’re taking advantage of the more sophisticated electronic health data becoming available,” says Tony Lau from the Defence Science and Technology Group.

By mining data already being collected by health organisations and public authorities the team can predict flu outbreaks up to seven weeks ahead.

This allows them to provide weekly flu forecasts to public health authorities.

Improving our capacity to predict flu peaks also aids our ability to detect bioterrorism attacks.

When people are exposed to conventional biological warfare agents they initially show flu-like symptoms, making it important to distinguish between a naturally-occurring flu epidemic and a biological attack so authorities can respond accordingly.

“We are the first ones to use this type of approach for the detection of a biological warfare agent release,” Tony says.

The tool has outperformed surveillance systems currently used by the US military, and the US Department of Defense is now funding the research.

For more information:

Defence Science & Technology Group
Darryl Johnston
+61 2 6128 6385

Tony and his team are developing a bio-surveillance portal. Credit: Defence Science and Technology Group