Researchers studying suicidal thoughts and behavior among adolescents don’t usually take a machine learning approach, but that’s what BYU, Harvard, and John Hopkins researchers did in a new study. They had thousands of students answer a series of questions, and BYU professor and study co-author Quinn Snell said they let the algorithm do the talking.
“There's the typical things that you think of when it comes up, right,” Snell said. “But we decided to stay away from that and just let the data tell us what happened, and let the data tell us what was the most predictive.”
What they found is that certain risk factors stood out from others, says BYU professor and study co-author Carl Hanson.
“Our findings were really interesting in that the risk factors tend to hang around two important things, and that's connection with peers and connection at home in the family,” said Hanson.
The algorithm predicted suicidal thoughts and behavior in teens with 91% accuracy. Snell says these results could help policy makers.
“I hope that can be something that we can get into the hands of these administrators and to help our students,” Snell said.
This topic can be hard to talk about, Hanson said, but taking a positive approach could make all the difference.
“If we could flip the script and look at the glass half full, and think about what we can do to strengthen connection amongst peers and help to strengthen those families,” Hanson said. “We begin to move towards prevention of adolescent suicidal thought and behavior and ultimately, suicide itself amongst adolescents.”
