Data61, UNSW create tool that uses AI to predict disease outbreaks from Twitter

Australian researchers have created a tool that monitors Tweets to predict outbreaks of diseases.

The tool may be used to detect outbreaks such as Influenza, Ebola and the Zika virus.

Australian researchers have created a tool that monitors Tweets to predict outbreaks of diseases. Researchers from Data61, the data research arm of Australia’s national science organisation CSIRO, say the tool uses artificial intelligence to analyse Twitter posts to establish a pattern of a disease outbreak.

The agency has been working on the project with the University of NSW Sydney’s Kirby Institute

Data61 says it tested the tool on a breakout of a thunderstorm asthma epidemic that hit Melbourne exactly three years ago.

Thunderstorm asthma is the triggering of an asthma attack by environmental conditions directly caused by a local thunderstorm.

Data61 says the sudden outbreak in Melbourne on 21 November 2016 inundated emergency services and hospitals resulting in over 8,000 hospital admissions by 6pm that day.

The agency hopes that the tool can be used to detect other outbreaks early.

However success will depend on the public taking to social media to post about their health. Researchers appear convinced there are enough posts for the tool to work.

Aditya Joshi, Postdoctoral Fellow at CSIRO’s Data61, says the popularity of social media makes it a valuable source of information for epidemic intelligence. He says a key challenge in the case of acute disease events is to detect them as soon as possible to assist health agencies to respond swiftly in emergency situations.

“We developed a technique that was able to detect the disease outbreak up to nine hours before it was officially reported and before the first news story broke,” Dr Joshu says.

Using anonymised and publicly available Twitter data, the tool analysed more than 3 million tweets containing keywords related to asthma such as “breath” and “coughing”.

“We can draw upon informal sources such as social media data to understand how acute disease events occur, and we can detect when and where an outbreak is likely to occur,” he says. “This means hospitals and public health agencies can be as prepared as possible.”

Data61 says the technique combines two fields of artificial intelligence — natural language processing and statistical time series modelling — and a four-step process to ensure the tweets containing the keywords are reports of health conditions and to remove duplicates where an individual might tweet more than once about their condition.

Natural language processing is used to process human language. The tool can distinguish between symptoms and unrelated mentions of the keywords.

Professor Raina MacIntyre, Head of Biosecurity Research Program, Kirby Institute, UNSW Sydney, says this work is a remarkable contribution to public health research.

“In future, this system can be used to provide health authorities and the community early warning of a serious and sudden health event,” Professor MacIntyre said.

“Early detection could significantly improve our capability to mitigate the impact of epidemics.”

Data61 says the tool can be used to detect other outbreaks such as Influenza, Ebola and the Zika virus. It draws on Data61’s Emergency Situation Awareness system, which analyses Twitter messages posted during disasters and crises to support disaster response efforts.