Australia: New software based on statistical probability principles is helping forecast the flu season, and could also identify the level of threat of a bioterrorist attack, write Liz Wells and Holly Bennett, University of Melbourne
A forecasting tool that uses data routinely collected by Australian state health departments could be used to forecast disease in the event of a bioterrorist attack or new influenza pandemic.
EpiFX, which was co-developed by researchers at the University of Melbourne and Australia’s Defence Science Technology (DST) Group, is already predicting the start and extent of the winter influenza season, sometimes up to weeks in advance. In the event of a pandemic, it can also be used to identify an appropriate response.
The collaboration has drawn the attention of the US Government’s Department of Defense, which is now working with researchers to develop a new tool to identify the level of threat from a bioterrorism attack and recommend appropriate response options.
James McCaw, Professor in Mathematical Biology at the University of Melbourne, has been at the forefront of infectious diseases modelling since its early days in Australia about 12 years ago, and leads the University’s EpiFX team. “We have to start with what we know about these pathogens and how they might spread,” he says.
“When an outbreak occurs, either naturally or through an act of bioterrorism, the forecasting tool provides the crucial link between scenario planning, which has been conducted in preparation for future events, and real-time data analysis.
“That kicks in when people affected by the virus start to arrive at emergency departments, or when in a particular environment, such as a barracks, soldiers start displaying symptoms.
“Analysis of that data shows us how quickly a virus is spreading, and initiates an appropriate and proportionate response.”
In extreme cases, this could mean recommending the closure of schools or shutting down public transport to minimise opportunities for a virus to spread.
Dr Tony Lau, who leads DST Group’s epidemic detection and forecasting programme, says the tool combines the concept of probability inference – which updates the probability of a hypothesis as more information becomes available – with susceptible, exposed, infected and recovered (SEIR) compartmental disease models.
An illustration of the July 2016 influenza forecast for Melbourne (image: University of Melbourne)
“This provides a mathematical framework for understanding the establishment and spread of infectious diseases,” according to Dr Lau.
This modelling can potentially predict how quickly an epidemic will spread, and identify those most at risk so preventive action can be taken through measures like targeted vaccination.
“In general, it is the young, the old and the weak who are most at risk from viruses like influenza,” Dr Lau adds. “Better forecasting means we can use state or national resources more efficiently, with those people in mind.”
Colourised structure of a prototype for a universal flu vaccine. The nanoparticle is a hybrid of a protein scaffold (blue) and eight influenza haemagglutinin proteins on the surface (yellow). The haemagglutinin was specifically engineered to display antibody binding sites common to all human influenza subtypes. The particle designed by Jeffrey Boyington of the US Vaccine Research Center (VRC) has been shown to be an effective immunogen in mice and ferrets. The three-dimensional structure of the particle was determined by cryo-electron microscopy by John Gallagher and Audray Harris (Laboratory of Infectious Diseases). More information here.
In Victoria, EpiFX has accurately predicted flu outbreaks up to five weeks in advance. Weekly forecasts of the incidence of flu are shared with the health sector to gain further insight into the influenza season and continue its refinement.
Despite early success, challenges remain. The 2017 influenza season in Australia was severe, with an extreme case count recorded in many jurisdictions. Dr Rob Moss, mathematical biologist at the University of Melbourne and EpiFX technical project lead, says the underlying causes for this increase are not yet understood but may stem in part from behavioural changes: “Our statistical tools are similar to those used to predict the weather, but we face the additional challenge of having to account for changes in human behaviours.”
“Individuals may change their health seeking behaviour based on media attention and general practitioners could have also changed their testing practices. We’ve only just started to scratch the surface in unpacking these complex interactions.”
Responding to bioterrorism
This new tool being worked on with the US Government’s Department of Defense is a partnership between the DST Group, the University of Melbourne and the University of Adelaide. Highly contagious diseases and viruses, like Ebola and influenza, as well as emerging diseases such as Zika, are being investigated as part of its development. DST Group began working with the University of Melbourne research team on disease modelling in 2014.
“We are still developing and refining our forecasting, but in the event of a health emergency, we are in a better position to respond than we have ever been, as we improve our ability to integrate forecasting with our scenario analyses,” says Professor McCaw.
The success of these forecasting tools involves a huge amount of collaboration from researchers and public servants working in fields including computer science, physics, mathematics, electronic and software engineering, epidemiology and medicine.
Other countries including the United Kingdom, South Korea and Singapore are interested in using the tool’s epidemic modelling capabilities.