Suicide is the top cause of death for Australians between the ages of 15 and 44, with nine people taking their own lives on average every day.
According to the University of New South Wales (UNSW), some estimates show suicide attempts happen up to 30 times more frequently than fatalities.
“Suicide has large effects when it happens. It impacts many people and has far-reaching consequences for family, friends, and communities,” UNSW Sydney PhD candidate in psychiatry at the Black Dog Institute Karen Kusuma said.
Kusuma investigates the issue of preventing teen suicide.
Recent research by Kusuma and a group of scientists from the Black Dog Institute and the Centre for Big Data Research in Health looked at the evidence supporting machine learning models’ capacity to forecast future suicidal behaviours and thoughts. The effectiveness of 54 machine learning algorithms designed by researchers in the past to forecast suicide-related events, such as ideation, attempt, and death, was assessed.
Machine learning algorithms beat conventional risk prediction models in forecasting suicide-related outcomes, which have historically performed badly, according to a meta-analysis published in the Journal of Psychiatric Research.
“Overall, the findings show there is a preliminary but compelling evidence base that machine learning can be used to predict future suicide-related outcomes with very good performance,” Kusuma said.
According to the UNSW, identifying people who are suicidal is critical for preventing and controlling suicidal behaviour. However, predicting danger is challenging.
Clinicians frequently employ risk assessment tools in emergency departments (EDs), such as questionnaires and rating scales, to pinpoint patients who are at a high risk of suicide. Evidence, however, indicates that they are insufficient in determining suicide risk in the real world.
“While there are some common factors shown to be associated with suicide attempts, what the risks look like for one person may look very different in another. But suicide is complex, with many dynamic factors that make it difficult to assess a risk profile using this assessment process,” Kusuma stated.
According to a post-mortem review of suicide deaths in Queensland, 75 per cent of individuals who had undergone a formal assessment for suicide risk were deemed to be at low risk, and none were deemed to be at high risk. Previous studies that looked at quantitative models for forecasting suicide risk over the previous 50 years discovered that they were only marginally more accurate than chance at doing so.
“Suicide is a leading cause of years of life lost in many parts of the world, including Australia. But the way suicide risk assessment is done hasn’t developed recently, and we haven’t seen substantial decreases in suicide deaths. In some years, we’ve seen increases,” Kusuma added.
Traditional suicide risk assessments continue to be used routinely in healthcare settings to establish a patient’s degree of care and assistance despite the paucity of supporting data. Generally, those who are determined to be at high risk get the best care, while those who are determined to be at low risk get dismissed.
According to Kusuma, unfortunately, with this strategy, those who truly require assistance aren’t receiving high-level treatments. Therefore, we need to look for ways to improve the process and prevent suicide.
More creativity in suicidology is required, according to Kusuma, and existing models for predicting suicide risk need to be re-examined. Her research using artificial intelligence (AI) to create suicide risk algorithms is the result of efforts to improve risk prediction.
“Having AI that could take in a lot more data than a clinician would be able to better recognise which patterns are associated with suicide risk,” she added.
The traditional clinical, theoretical, and statistical suicide risk prediction methods in the meta-analysis study were outperformed by machine learning models. In contrast, they accurately predicted 87 per cent of those who would not suffer a suicide event. They properly identified 66 per cent of those who would experience a suicide outcome.
“Machine learning models can predict suicide deaths well relative to traditional prediction models and could become an efficient and effective alternative to conventional risk assessments,” Kusuma said.
Machine learning models are not constrained by the rigid assumptions of conventional statistical models. Instead, they can be adaptably used to predict complex interactions between numerous risk factors and suicidal outcomes in huge datasets. They can also use social media and other responsive data sources to pinpoint peak suicidality risk and signal when interventions are most necessary.
“Over time, machine learning models could be configured to take in more complex and larger data to better identify patterns associated with suicide risk,” Kusuma stated.
With 80 per cent of the discovered studies having been published within the last five years, this study area is still in its infancy. Machine learning algorithms are used to predict outcomes connected to suicide. Future studies, according to Kusuma, will also aid in addressing the risk of aggregation bias discovered in algorithmic models thus far.
“More research is necessary to improve and validate these algorithms, which will then help progress the application of machine learning in suicidology,” Kusuma said. “While we’re still a way off implementation in a clinical setting, research suggests this is a promising avenue for improving suicide risk screening accuracy in the future.”
