Identifying teens experiencing suicidal thoughts and behaviors could now be easier than ever thanks to the development of a new algorithm based on machine learning .
The precision of the new algorithm is greater than that of previously developed predictive approaches. Orion Weller of Johns Hopkins University in Baltimore, Maryland, and his colleagues present these findings in the journal PLOS ONE .
A serious public health problem
The data to train the algorithm included responses to more than 300 questions each for more than 179,000 high school students who responded to the survey between 2011 and 2017, as well as demographic data from the United States Census.
The researchers found they could use the survey data to predict with 91% accuracy which individual responses from adolescents indicated suicidal thoughts or behaviors. In doing so, they were able to identify which survey questions had the greatest predictive power; These included questions about bullying or threats from digital media, bullying, serious discussions at home, gender, alcohol use, feelings of safety at school, age, and attitudes about marijuana.
Future research could expand on the new findings using data from other states, as well as data on actual suicide rates :
Our study examines machine learning approaches applied to a large data set of adolescent questionnaires, in order to predict suicidal thoughts and behaviors from their responses. We found great predictive precision in identifying people at risk and we analyzed our model with recent advances in ML Interpretability. We found that factors that strongly influence the model include bullying and harassment, as expected, but also aspects of your family life, such as being in a family with yelling and / or serious arguments. We hope this study can provide information to inform early prevention efforts.
Suicide is "multifactorial" and there are no general rules to understand all cases, but they do all have in common a "vital anguish" related to an education that hides suffering, failure or ruin. In the following video, at least, we focus on two factors that seem quite decisive: