As the rate of Army suicides continues to climb, hope has arrived in an unexpected form: a mathematical algorithm. (Scott Mansfield/UpperCut Images/Getty Images)
At one time, soldiers in the U.S. Army were less likely to commit suicide than civilians, but since 2004, an alarming trend has emerged: The rate of Army suicides has increased so dramatically that it now exceeds the rate among average Americans. And the numbers are only growing — a problem that has left experts dumbfounded.
“We don’t know why the rate is continuing to grow as the war wraps up. That’s worrisome,” said Ronald Kessler, a professor of health care policy at Harvard Medical School. “When you have a volunteer army, it’s not just a random cross-section of America. So are the kinds of people who join the Army different from other people? Has that changed over time as the economy has changed?”
These questions have spawned a larger one that military officials hope can help them identify who needs mental-health treatment the most: Which soldiers are at the highest risk of suicide? As part of the Army STARRS program — a multibillion-dollar research initiative to tackle this growing issue — Kessler and a group of 25 other researchers analyzed administrative data for thousands of active-duty soldiers who’d been hospitalized for psychiatric reasons over a five-year period.
Among this group, 68 committed suicide within a year of being discharged, which translates to about 264 suicides per 100,000 hospitalized soldiers. That’s more than 14 times higher than the rate of suicide for the U.S. Army as a whole.
“We found that the strongest predictor of suicide was having been in a mental hospital recently,” Kessler told Yahoo Health. “One would say, ‘When you leave the hospital, that’s when you’re cured, right?’ But in the case of mental disorders, there’s a very high suicide risk in that transition period back to normal life.”
Even though these soldiers represent a relatively small percentage of the Army as a whole, it would still be extremely costly to implement a prevention program targeting all soldiers who’d received inpatient psychiatric care, said Kessler. To further narrow the scope of at-risk soldiers, his team set out to develop an algorithm that could predict who was most likely to commit suicide, since past research has shown that risk-assessment tools tend to be more accurate than clinicians’ judgments. “There have been several efforts to do this in the past, and they all failed,” he said. “We were successful.”
Unlike past projects, which have focused exclusively on health-care claims data — things like the number of doctor visits in the past year — the researchers were able to capture a more comprehensive snapshot of each soldier’s life, analyzing everything from criminal justice data to work performance reports to medical records.
What they found: Being male, enlisting late in life, having a history of criminal offenses, owning weapons, and taking antidepressants in the last year, among other mental-health factors, all emerged as strong predictors of suicide risk. “It’s sort of a lethal mix — there’s not one variable that predicts suicide,” said Kessler. “In most cases, it’s an adding up of many, many little pieces that put people over the threshold.”
Using these factors, “we came up with an algorithm that could rank each soldier who’d been hospitalized from the lowest to highest in risk [of suicide],” Kessler said. From there, the researchers isolated the 5 percent of soldiers with the highest predicted risk, and found that 53 percent of posthospitalization suicides occurred in this group.
“If you turn it around and say, ‘Of those highest-risk people, how many kill themselves?’ it’s between 3 and 4 percent, or about 3,500 per 100,000,” said Kessler. “To put that in perspective, 18 out of 100,000 people in the Army kill themselves a year.” In other words, the algorithm successfully determined which soldiers are at the very highest risk of suicide — a group that clinicians could then target for mental-health intervention.
And not just for suicide prevention, either. This group also faced an elevated risk of a number of other undesirable outcomes. “Another 3 to 5 percent died in an ‘accidental death,’” said Kessler. “That’s much, much higher than you’d expect by chance. So clearly, not all accidents are accidents.” The rates of suicide attempts and serious car accidents were also unusually high among these soldiers, suggesting that “they are kind of flirting with death,” he said. “Forty-six of a typical 100 people in this highest-risk group have at least one of these bad things happen to them.”
The researchers are now working with the Army to implement their algorithm in real time — that is, while at-risk soldiers are still in the hospital. “They’re thinking through what kinds of interventions they’re going to put in place,” Kessler said. Although it’s unclear what these programs will entail, suicide-prevention efforts that have worked in the U.K. could serve as a model: required outpatient visits within a week of discharge, 24-hour crisis teams, and aggressive outreach when patients miss follow-up appointments. “Our hope is now that we’ve isolated the group, we can do something to nip the problem in the bud,” said Kessler.
Your Next Read: How to Deal with Depression in Your 20s