A woman sneezes into a tissue.

Detecting outbreaks of flu and other illnesses

Health services need to know when a flu outbreak is normal or not so they can, if needed, quickly implement measures to contain it.

• 1 September 2010 | Updated 14 October 2011

• The winter of our discontent
• Rapid flu data analysis

Each year, the flu season arrives in our neighbourhoods, schools and workplaces. Some of us have to take time off work or school and some have more severe cases that take us to hospital.

The winter of our discontent

Each year the flu season is different. It can vary by such things as:

• when it starts and ends
• how many people require hospitalisation
• which viruses are causing illness
• which population groups are worst affected.

Influenza or the flu can spread rapidly within a family, within a city, and across the world. When a 'superflu' arrives like Pandemic (H1N1) 2009 (also called swine flu) or avian flu, health authorities need to act quickly to prevent a full-on epidemic and develop strategies to reduce its harm.

To track a flu season, hospitals routinely count the number of admissions to emergency departments with various symptoms, capturing information in their e-Health databases. But technology alone is not enough to make sense of the data and guide decision making.

CSIRO statisticians have developed a system that alerts health departments when there is an outbreak and diagnose its nature.

The statisticians have analysed data about flu cases gathered over several years to see what a normal flu season looks like. Then they developed a method to analyse data as it comes in, so that deviations from normal can be detected quickly.

CSIRO has past and current projects in outbreak detection for state health departments in:

• New South Wales 
• Queensland.

Rapid flu data analysis

CSIRO's surveillance methods can characterise a flu season as behaving normally or not, at least a half-day ahead of other methods. They can signal when something more serious is occurring, such as an outbreak of swine flu or even a disease yet unknown to medicine.

Even before a cause is known, the CSIRO surveillance system can detect if an outbreak is occurring or has started. This allows health measures to be taken earlier, giving health services a better chance of reducing its spread in the community.

This system uniquely models a range of patient features like:

• age
• gender
• ethnicity
• pregnancy status
• location of home/work/school.

Patients can be characterised this way and groups most at-risk can be provided with tailored information by health authorities.

Similar statistical surveillance methods can be used on health data to detect:

• other natural outbreaks like gastroenteritis
• intentional outbreaks like bioterrorism or food tampering.

The aim is to eventually develop reliable systems that can be built into software for hospitals or health departments. It will provide intelligence for emerging e-Health technologies and give Australians a better chance of staying healthy this winter.

Read the Quantitative biosciences research overview.