A national surveillance system for respiratory epidemics could use machine learning to analyse real time data and give early warning of emerging events such as thunderstorm asthma, an expert panel has been told.
A national workshop of stakeholders convened in the wake of Victoria’s catastrophic thunderstorm event of 2016 identified three key challenges to setting up a more timely respiratory syndromic surveillance system.
While there are currently good systems in place to identify and track specific respiratory epidemics such as influenza, there are no national real time systems that can predict and co-ordinate rapid responses to respiratory syndromes such as thunderstorm asthma, according to a new report of the workshop held in Canberra.
Bringing together stakeholders including respiratory physicians, public health authorities, ambulance services and data analysts, the workshop concluded that the three key issues for a national surveillance system would be reliable and timely data sources, accurate analytic methods and communication of surveillance information to allow rapid responses by health services.
The workshop heard that current surveillance systems are based around diagnostic-based data sources such as hospital discharge data. A real-time system for rapidly emerging respiratory syndromes may need to be built around new sources such as 000 calls, ED admissions, GP sentinel data and social media reports.
For analysing large amounts of real time data, machine learning appears to be a promising approach to quickly detect anomalies and signals, the workshop heard. However machine learning results would need to be robust enough to avoid false alarms and have expert oversight to be able to interpret and establish how severe an event is.
Communication of information about emerging respiratory syndromes would require a system that is simple, effective and which delivers relevant advice to diverse groups of responders including doctors, hospitals, first responders as well as to the general public, schools and government departments.
“Syndromic surveillance … will help to drive investigation of the cause of the event and mobilisation of the appropriate response,” concluded the report section led by Professor Guy Marks of the Australian Centre for Airways disease Monitoring (ACAM).
“Specifically, it will enable the health services to achieve situational awareness and activate appropriate responses to deal effectively with the increased number of cases. It will also allow health protection epidemiologists to identify the cause and attempt to mitigate the event by controlling the exposure or limiting its impact. As well, it will facilitate effective public communication. “
However further work and consultations will be needed to determine which issues are of highest priority to overcome and where the responsibility lies for doing this, the report concluded.