Since the 2013-2014 season, scientists of Columbia University's Mailman School of Public Health have published weekly regional flu forecasts for over 100 cities in the US.
Their system employs a computer model to generate multiple simulations that mimic the behaviour of an outbreak and are then knit together to generate an overall prediction.
In the new study, the researchers used data from a network of 50 outpatient clinics and laboratory reports in Hong Kong from 1998 to 2013 as a test case to retrospectively generate weekly flu forecasts.
The technique predicted the peak timing of the outbreak three weeks in advance of the actual peak with accuracy as high as 93 per cent.
More From This Section
Prediction accuracy varied depending on the strength of the outbreak and how far in advance the prediction was made.
In general, forecasts for specific strains were more accurate than those for aggregate epidemics, and the peak and magnitude of outbreaks were more accurate than the timing of their onset or their duration.
Seasonal influenza outbreaks in temperate climates like the US are restricted to the winter months. By contrast, outbreaks in the subtropics such as Hong Kong happen year-round.
In addition, outbreak intensity, duration, and timing are more variable in the subtropics than in temperate regions.
"The irregularity of flu outbreaks in subtropical climates makes forecasting more challenging," said first author Wan Yang, an associate research scientist at the Mailman School.
The study was published in the journal PLOS Computational Biology.