Study: Algorithm can predict when adolescents will commit suicide with 91% accuracy

DOI：10.1371 / journal.pone.0258535

Explanation

Researchers from Brigham Young University, Johns Hopkins, and Harvard have created an algorithm that can predict suicidal ideation and behavior in adolescents with 91% accuracy, as outlined in an article published in PLOS ONE.

It also details online harassment and bullying, which are major predictors of suicidal ideation and behavior in adolescents, and predicts with high accuracy which adolescents will become suicidal and exhibit suicidal behavior (attempts) based on the experiences they face. We analyzed data from 179,384 middle school and high school students who participated in the 2011-2017 Student Health and Risk Prevention Study.

The dataset included responses to more than 300 survey questions and more than 8,000 demographic information, for a total of 1.2 billion data points processed, and a variety of algorithms were applied to the data to find a machine learning model that accurately predicted which adolescents continued to have suicidal thoughts and behaviors (STB) based on the data provided.

It showed that females were more likely to experience suicidal ideation and behavior (17.7%) than males (10.8%), and adolescents without a father at home were 72.6% more likely to have suicidal ideation.

The algorithm also found which risk factors were the major predictors of suicidal ideation and behavior, and these included

・Being threatened or harassed through digital media

・Being singled out or bullied by students at school

・Exposure to/involvement in serious discussions and shouting at home.

The researchers were not surprised to see some of these risk factors, but they were a little surprised to see the large impact from family factors.

Three of the top ten predictors of STB were directly related to family situations.

These include being in a family where there are serious arguments, being in a family where people argue about the same things over and over, and being in a family where people yell at and insult each other.

It is unclear whether knowing these predictors can prevent suicidal behavior in the future, but it may be helpful to know your own situation.