Uniformed Services University-led studies leverage data predicting Soldiers at highest risk for suicide

The Uniformed Services University in Bethesda, Maryland, published research Jan. 6, 2025, that showed machine learning models, a type of artificial intelligence, could predict Soldiers at highest risk for suicide attempts following a periodic health assessment. In January 2026, this research evolved into controlled trials with Soldiers called the Suicide Avoidance Focused Enhanced Group Using Algorithm Risk Detection, or SAFEGUARD, initiative.

James Naifeh, a Ph.D. clinical psychologist, assistant scientific director at the USU Center for the Study of Traumatic Stress, and professor of psychiatry at USU’s F. Edward Hebert School of Medicine, is part of the team that will leverage data from SAFEGUARD. They’re looking to “target suicide prevention programs to Soldiers at highest risk, and to develop decision support tools to help providers identify which intervention is most likely to help a particular Soldier, also known as precision or personalized medicine,” he said.

SAFEGUARD tests mental health interventions in Soldiers at different touchpoints, such as first duty station, annual PHA, or discharge from psychiatric hospitalization. It intends to gauge those at risk for suicide, Naifeh said.

“Suicide is a devastating event, but it is also a rare and very difficult-to-predict event,” he said.

With the changing digital landscape of military medicine, “machine learning offers a way of targeting more intensive, evidence-based interventions to Soldiers at high risk of suicidal behavior,” said Naifeh.

SAFEGUARD designed to protect those at risk

SAFEGUARD uses machine-learning models originally developed by the U.S. Army Study to Assess Risk and Resilience in Servicemembers, known as STARRS.

The STARRS team integrated more than 50 U.S. Army and Department of War administrative data systems with the records of all Soldiers on active duty from 2004 to 2019. Using those data, the machine learning models assign each individual a risk score, representing their likelihood of future suicidal behavior.

The SAFEGUARD initiative will take the STARRS models that were developed using historical records and update them using real-time medical and personnel data to deploy suicide prevention interventions to Soldiers with high predicted suicide risk.

SAFEGUARD currently has three components focused on different military career touchpoints:

• Level Up
• Operation Life Force
• Pathfinding

Level Up is designed for Soldiers in their first assignments delivered in a group orientation session on arrival. “It delivers practical, cognitive-behavioral (talk therapy) skills-based tools to manage stress, regulate emotions, think tactically, and build healthy relationships,” according to Naifeh. In other words, Level Up teaches life skills to young service members.

Operation Life Force is a virtual program that delivers five group interventions for Soldiers with high predicted risk of suicide-related behaviors. Its skills training and suicide safety planning are framed in “mental toughness to minimize stigma while teaching soldiers skills to manage harmful behaviors,” Naifeh said.

Fort Hood, Texas, is enrolling Level Up and Operation Life Force participants for the controlled trial.

Pathfinding, which Naifeh leads, is recruiting Soldiers nationwide who had inpatient psychiatric care at military hospitals. Participants receive either treatment as usual or treatment as usual plus an intensive case-management intervention, said Naifeh. Centrally trained and supervised mental health professionals run the six-month telehealth program.

It’s based on two existing evidence-based case-management programs and adapted for active duty Soldiers.

“By providing that additional support and helping Soldiers identify resources that can help them, we hope to reduce the rate of suicidal behavior in that group,” Naifeh said. “This approach aims to improve Soldier well-being and contribute to a healthier force.”

Enhancing suicide risk predictors in screening

One reason SAFEGUARD uses machine-learning algorithms based on administrative data is that previous research indicated Soldiers “are less likely to report mental health symptoms and suicidal thoughts when the Army may see their responses than when their responses are anonymous or confidential,” according to Naifeh.

"We found very few soldiers reported suicidal thoughts on their PHA," Naifeh said in a May 28 USU article on his research, highlighting a serious gap in the current screening process. "In fact, 99.8% of soldiers denied having suicidal thoughts when they completed the PHA,” he said. As a result, 95% of suicide attempts that occurred in the 6 months after a PHA were among Soldiers who denied having suicidal thoughts during their PHA.

With the implementation of machine-learning models, the ability to identify those at risk for suicide in the study increased substantially.

"Based on our preferred model, the 25% of Soldiers with the highest predicted risk account for about 70% of the suicide attempts that occurred in the six months following the PHA. This is known as concentration of risk,” he said.

SAFEGUARD and military readiness

SAFEGUARD contributes to warfighter readiness by “reaching Soldiers with the skills and support they need at key points in military service,” Naifeh said. “If it’s successful in preventing suicidal behavior, SAFEGUARD will contribute to warfighter readiness by reducing the devastating family, unit, and organizational disruption associated with suicide.”

Following the SAFEGUARD trials, the initiative could roll out to the other services, Naifeh suggested. However, this effort would require “some initial work to develop and test machine learning models using service-specific data.”