Mental disorders affect a huge number of young people. An estimated 9 million people between the ages of 10 and 19 live with a mental disorder. Half of them struggle with depression. According to 2021 UNICEF report, suicide is the leading cause of death among young people in Europe.

The development of artificial intelligence could improve mental health treatment, a new study conducted by researchers at UCLA Health in the US, published in the journal JMIR Mental Health, shows. It found that machine learning can detect symptoms of self-harm in children more effectively than humans.

The researchers analysed clinical notes from 600 emergency department visits made by children aged between 10 and 17. They wanted to check the extent to which patients' mental health assessment systems help identify symptoms of self-harm and assess suicide risk.

The findings are surprising. The clinical notes missed worrying signs in 29% of children who came to emergency departments with thoughts of self-harm, while statements made by health specialists flagging at risk-patients overlooked 54% of patients, Euronews reports.

And the reason professionals did not detect thoughts or behaviours associated with self-harm? Children are reluctant to report such distressing thoughts during their first visit to the emergency department. Those at risk of being missed were more likely to be boys than girls. Also black and Latino youth were more likely to be overlooked than white children and adolescents.

Researchers at UCLA Health have figured out how to help with the problem. They created machine learning models that analysed data such as previous medical care, types of medication taken, residence, and laboratory test results. In this way, the models estimated the level of suicide-related thoughts or self-harm behaviour.

The models identified patients at risk better than traditional methods of analysis. The authors of the study are now committed to improving the level of prediction of patients at risk and lower a chance of false positives – patients to be falsely flagged by AI models. The team will focus on further developing the technology.

Source: Euronews