Researchers from the School of Clinical Medicine at the University of New South Wales have used machine learning (ML) to develop more accurate predictors of self-harm and suicide in adolescents.
Adolescent mental health has seen a global decline and Australian statistics are a fair reflection of those in many other societies. Suicide is the leading cause of death among Australians aged between 15 and 24.
Current predictors like past self-harm or suicide attempts have been found to be only slightly more effective than relying on chance. The new predictive model the researchers developed using ML is considerably more effective.
The dataset comprised feedback from questionnaires and interviews with 2,809 adolescent participants in the Longitudinal Study of Australian Children.
Among the participants, 5.2% reported attempting suicide at least once in the previous 12 months and 10.5% reported acts of self-harm.
By using machine learning to analyze the data the researchers found key predictors that were more accurate than those previously used by mental health practitioners.
The performance of the ML model was measured with the Area Under the Curve (AUC) metric. It’s a figure ranging from 0.5 to 1 where 0.5 is as good a guess as a coin flip, and 1 is a 100% accurate prediction.
Relying solely on previous history of self-harm and suicide attempts achieved an AUC between 0.63 and 0.647. This was only slightly better than guessing and fell below the 0.7 to 0.8 range considered acceptable for predicting risk.
The ML model’s predictors achieved an AUC between 0.722 and 0.74, which is significantly better.
The model surprised the researchers as it showed that previous self-harm or suicide attempts were not a high-risk factor and that environment and parental support played more important roles.
Dr. Lin, one of the researchers, said, “We found that the young person’s environment plays a bigger role than we thought. This is a good thing from the standpoint of prevention because we now know that there’s more we can do for these individuals.”
This application of AI in mental health care will help clinicians be more accurate in assessing at-risk adolescents and making earlier interventions.
Dr. Lin said that, “Based on patient information the ML algorithm could calculate a score for each person, and that could be integrated into the electronic medical records system. The clinician could quickly retrieve that information to confirm or tweak their assessment.”
The model isn’t ready to be rolled out in practice yet but the promising results indicate that it’s an avenue worth pursuing.
There’s a lot of data sitting in medical records and using AI to analyze it will undoubtedly deliver more surprises for health practitioners.
