Case study: EpiFx and EpiDefend

Gaining control over pandemics with predictive modelling

The advantage of mathematical modelling is that the same core algorithms amount to a universal platform for exploring disease control options across disease and pandemic scenarios.

Context

The greatest public health outcome during a pandemic is achieved by rapidly controlling the spread of disease at the start of an outbreak. Achieving that overriding objective requires understanding, ahead of time, how a disease is likely to spread.

Forecasting disease spread is a formidable task, but one that caught the attention of James McCaw, Professor in Mathematical Biology at the University of Melbourne over 10 years ago.

Since then, Professor McCaw has partnered with the Defence Science and Technology (DST) Group to develop EpiFX, a modelling tool that can translate disease surveillance data into predictions of likely future spread.

Participating in that work at the University of Melbourne are Dr Rob Moss and Dr Freya Shearer, who collaborate with DST’s Dr Peter Dawson and Dr Tony Lau, alongside the US Government’s Department of Defense.

Professor McCaw was motivated by a series of epidemic outbreaks and one pandemic – avian bird flu, SARS, swine flu – that challenged nations around the world starting in the 1990s. He found that global responses lacked a system to interpret disease incidence early in an outbreak and that limited the ability to select optimal control measures.

EpiFX was developed to solve this problem. It converts surveillance data – such as hospitalisation rates or social media chatter – to produce disease spread forecasts that are finding both public health and military applications.

Project

Spearheading efforts to refine the technology and validate EpiFX forecasts is Dr Moss.

His most constant focus is seasonal influenza. Using various sources of data, he produces weekly forecasts that are helping to validate the forecast system. This is essential to build confidence in the models among public health agencies, which are the ultimate stakeholders of this development work.

The prime disease monitoring data tapped by Dr Moss is provided by state and commonwealth health departments.

However, he is finding important nuances in other sources, especially data crowdsourced from an online system called FluTracking. This additional data is helping to identify unusual epidemics earlier than otherwise possible, adding to early-detection capability. An example of an unusual influenza epidemic was the 2017 season, which was unusually long and severe.

Outcomes

The advantage of mathematical modelling is that the same core algorithms amount to a universal platform for exploring disease control options across disease and pandemic scenarios.

The EpiFX algorithms are applicable to many of Australia’s 70-plus notifiable infectious diseases, for instance. They are also useful to enhance preparedness for pandemics. And, for the military, the stakes involve the ability to protect troops from epidemic diseases that can decimate operational capacity.

The same tool also has applications defending against bioweapons, including acts of bioterrorism, especially since many bioterrorism agents result in flu-like symptoms. As such, it helps to have a forecast about the expected level of flu activity for a certain population.

Our weekly forecasts provide a measure of background flu activity, allowing for the detection of unusual spikes that may indicate a bioterrorism attack, Dr Moss says.

In these scenarios, EpiFX remains effective even if detailed information about a pathogen’s genetics, virulence or geographic spread is lacking.

Instead, the predictions are based on rates of change among those infected, those who survived infection (and are immune), and the susceptible. That requires understanding the relationship between a patient’s infectiousness and the expression of symptoms, given different modes of transmitting a disease.

In translating EpiFX for use in bioterrorism or pandemic scenarios, Dr Shearer says that extra tools are required to fully capitalise on the earliest available data of a pathogen’s transmissibility and disease burden. These tools were developed at the University of Adelaide by Professor Joshua Ross and are called First Few Hundred (FF100) algorithms.

Dr Shearer is now developing a system that combines information from EpiFX and FF100 models, making it possible to run thousands of intervention simulations to predict the effectiveness of different control options, given the spread and severity of a pandemic.

The aim is to rapidly identify the most effective and proportionate public health measures, which can include the distribution of anti-viral drugs, school closures and the cancellation of large public events.

Back to top