An American flag flies at half-staff over Ohio in 2021. The Army could better tailor suicide prevention initiatives by using machine learning tools to predict which soldiers are at higher risk for harming themselves, according to a new mental health study. (Juliet Louden/U.S. Air Force)
Machine learning tools could help the Army identify which soldiers are at higher risk of self-harm and use the knowledge to improve suicide prevention efforts in the service, a new study found.
Released this month by the Nature Mental Health journal, the study looked at Army records to determine whether an artificial intelligence model could be developed to better predict who would make a suicide attempt within six months of their annual health assessment.
Researchers’ algorithm found that 25% of soldiers determined to be at the highest predicted risk accounted for 69.5% of suicide attempts in the given time frame, the study said.
The Army’s yearly Periodic Health Assessment, which screens for problems, was of little value on its own for predicting future suicide attempts based on its risk questions, according to researchers.
For example, 95% of suicide attempts occurred among soldiers who had denied being suicidal in their health assessment, the report said.
However, when the researchers plugged in thousands of other variables drawn from 54 Army and Defense Department data systems that house soldier records, machine learning proved effective at predicting suicide attempts, the study said.
The researchers drew their sample group from Army records between 2014 and 2019, looking at 452,473 active-duty soldiers and 2,096 recorded soldier suicide attempts.
“This model could be used to identify soldiers who should be referred to behavioral health treatment, as well as to suggest which soldiers already in treatment need more intensive treatment,” said the report, which was produced by more than a dozen researchers from an array of universities and institutes.
The findings come as the U.S. military as a whole grapples with suicide, a point of long-standing concern that was brought into the spotlight following the high-profile New Year’s Day suicide of an Army Special Forces member outside Trump International Hotel in Las Vegas.
Master Sgt. Matthew Livelsberger, who authorities say shot himself moments before his rented Tesla truck burst into flames, was on leave from his unit in Stuttgart, Germany.
The Army said Livelsberger, who had an extensive history of combat deployments, had used a Special Operations Command program focused on “holistic care” related to service members’ individual physical and mental needs.
However, he did not display any concerning behaviors at the time his leave was approved, the Army said.
Researchers in the Nature Mental Health study said data show that soldiers rarely speak up about having suicidal thoughts.
The findings suggest that despite Pentagon attempts to increase awareness and boost mental health services, rank-and-file service members remain concerned about being stigmatized if they speak up about their problems.
The researchers said a limitation of their study was that it did not analyze how much the more intensive interventions based on the model might cost the Army.
However, such interventions could be effective and inexpensive “relative to the value of preventing the negative outcome” of mental health struggles in the ranks, the study said.
