Professor Svetha Venkatesh is helping to identify people in the Geelong area most in danger of making a suicide attempt. Yet she’s not a doctor. She’s not a psychiatrist. Professor Venkatesh is a computer scientist.
Under Professor Venkatesh’s direction, the Centre of Pattern Recognition and Data Analytics (PRaDA) at Deakin University, in conjunction with Barwon Health, is developing a computer program that works on the myriad data collected by hospitals in the Geelong area, to help predict the likelihood of suicide.
‘There are many problems that “Big Data” can solve in the medical field,’ says Professor Venkatesh. ‘But we chose suicide prediction because, although risk factors are known, combining them to predict suicide attempts is difficult. It’s not a disease where you can do a test and see results. Computer assisted tools providing clinical support in such problems are useful.’
‘Every time you go to hospital they code information relating to diagnosis, discharge, medications, procedures,’ she explains.
‘They have to collect all this information for administrative and funding reasons.’
The PRaDA team used this data with machine learning techniques to extract a set of features and their combinations to predict suicide attempts.
That’s where the Deakin/Barwon Health collaboration comes in.
'There are many problems that “Big Data” can solve in the medical field'
‘Barwon Health has a completely electronic medical record system,’ she explains. ‘But it has never been used in this way before. You might have gone to Emergency 20 times, and it will all be recorded, but when you come to Emergency, your doctor doesn’t have time to go through years of records to assess your risk of suicide. They will make the assessment largely on what they see before them.’
‘Electronic medical records used to be about reducing paper,’ says Associate Professor Richard Harvey, Clinical Director of Mental Health, Drugs and Alcohol Services at Barwon Health, who is working with Professor Venkatesh on the project. ‘The next step, which is where we are now, is working out how to use the data we are collecting in more useful ways.’
‘Low risk might include minor scratches, whereas high risk might be driving a car into a wall or taking dangerous drugs.’
‘It’s not about predicting suicide,’ she stresses. ‘It’s about assisting those at immediate risk. It means that limited resources can be allocated more effectively and in a more targeted way.
‘Currently the system is being trialled. Deakin is sent the data – all names and identifying information is removed and replaced by numbers – and it is returned to Barwon Health with the analysis. The next stage is to have the system deployed to a clinical setting.’
Professor Venkatesh isn’t stopping at suicide. She explains ‘we are also working on cancer mortality prediction, and we are just starting to work on predicting pre-term birth risk, so that mothers can be allocated the safest risk model.’
'Low risk might include minor scratches, whereas high risk might be driving a car into a wall or taking dangerous drugs.'