New Tool Aims to Identify Soldiers at Highest Suicide Risk

Deborah Brauser

November 13, 2014

A new "risk algorithm" may help identify US soldiers at highest risk for suicide following psychiatric hospitalization and offer an effective opportunity to curb this growing problem in the US military, new research suggests.

Dr Ronald Kessler

The latest results from the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS) showed that 5% of the soldiers classified as being at highest risk for suicide accounted for 53% of all suicides that occurred within 12 months of inpatient discharge.

Strongest predictors of suicide in this group included being male or at a late age at enlistment, access to a number of firearms, past criminal offenses, prior suicide attempts or ideation, and number of antidepressant prescriptions filled in the previous year.

The investigators noted in a release that the suicide rate in the US Army has steadily increased since 2004. And although posthospitalization suicides account for only 12% of all the suicides in this military branch, "the algorithm would presumably help target preventive interventions," they said.

"When we combed through the massive amount of administrative data and indicators that predict suicide, we were able to create a composite index that did a very good job of isolating those soldiers with a high proportion of the suicides in the months after hospitalization," lead author Ronald Kessler, PhD, McNeil Family Professor of Health Care Policy at Harvard Medical School in Boston, Massachusetts, told Medscape Medical News.

However, Dr Kessler pointed out that several key issues need to be addressed before clinical implementation. "The algorithm is still a work in progress."

The study was published online November 12 in JAMA Psychiatry.

A Better Way

Dr Kessler noted that previous research has shown that the strongest predictor of suicide is recent hospitalization for a mental disorder.

"The period in the weeks and months after leaving the hospital and getting reintegrated back into society or into the Army is a particularly high-risk time," he said.

However, the investigators write that suicide is still a relatively rare outcome. "Therefore, the benefits of providing intensive...interventions to all recently discharged inpatients are low."

They add that a better approach would be to combine inexpensive universal interventions with more intensely targeted high-risk interventions.

But this would require the use of a reliable and objective risk evaluation system based on administrative data, because "research has consistently revealed that health care professionals are not accurate in making such assessments," write the investigators.

Army STARRS was established with a mandate to assess factors that may help protect the mental health of military personnel and factors that could put them at risk. The study is scheduled to run until June 2015, but findings have been reported as they have become available.

For the current Army STARRS analysis, the investigators sought to develop an "actuarial risk algorithm" that could predict suicide in the year following inpatient psychiatric treatment.

They examined data for 53,769 active duty hospitalizations with psychiatric diagnoses at admission between 2004 and 2009, involving a total of 40,820 soldiers.

Machine learning methods were then used to evaluate predictive associations among a wide array of data systems.

Bad Outcomes

Results showed that 68 of the soldiers died by suicide within 12 months of psychiatric inpatient discharge, representing 12% of all Army suicides. This is equivalent to 263.9 suicides per 100,000 person-years vs 18.5 suicides per 100,000 person-years in the total US Army.

In addition, most of the posthospitalization suicides occurred in the "highest-risk stratum" (5% of the highest-risk hospitalizations; concentration of risk [CR], 52.9%; 3824.1 suicides per 100,000 person-years).

Strongest predictors of suicide included male sex (odds ratio [OR], 7.9; 95% confidence interval [CI], 1.9 - 32.6), nonviolent weapons offense (OR, 5.6; 95% CI, 1.7 - 18.3), prior suicide attempts (OR, 2.9; 95% CI, 1.7 - 4.9), verbal assault offenses (OR, 2.2; 95% CI, 1.2 - 4.0), age of 27 years or older at enlistment (OR, 1.9; 95% CI, 1.0 - 3.5), outpatient visits with suicidal ideation (OR, 1.6; 95% CI, 1.1 - 2.5), and number of registered pistols (OR, 1.3; 95% CI, 1.0 - 1.6).

Other strong risk predictors included aspects of prior psychiatric treatment and disorders diagnosed during hospitalization. For example, a higher number of filled antidepressant prescriptions had an OR of 1.3 (95% CI, 1.1 - 1.7), and a diagnosis of nonaffective psychosis while hospitalized had an OR of 2.9 (95% CI, 1.2 - 7.0).

Interestingly, the highest-risk hospitalization group also had significantly elevated risk for death by unintentional injury (CR, 10.1%; P = .008) within a year of hospital discharge.

"This was surprising. This was much more than the 5% you'd expect by chance," said Dr Kessler.

This group also had high rates of suicide attempts (CR, 9.1%; P < .001) and subsequent hospitalization (CR, 7.5%; P < .001).

"All of these things show that we're finding a risk composite that's not just risk for suicide but for a range of behaviors within that spectrum that might be considered 'flirting with death' or suicidal inclination," Dr Kessler said.

"What was surprising to me was that out of every 100 soldiers in that highest-risk group, close to half [46.3%] had one of those very bad outcomes. They died by suicide or by automobile accident, they made a suicide attempt, they had some nonfatal accident, or they were rehospitalized," he said.

The investigators note that this high concentration of risks could justify the expense of expanding posthospitalization interventions to this patient population, although "final determination requires careful consideration of intervention costs, comparative effectiveness, and possible adverse effects."

"With a group this small having so many bad things going on, some effort should be made by the Army to target them for special, expanded outreach efforts to try and prevent not just suicide but that whole ball of wax," said Dr Kessler.

He noted that the investigators are now trying to develop a predictive algorithm for outpatients, as well as for particularly high-risk periods for military personnel, such as during the first months after return from deployment.

"Suicide is a very complicated thing. In order to tackle it in a systematic way, we need to break the problem into parts and look at each of several different high-risk subgroups with a somewhat different model."

Need for Posthospitalization Services

"The key message of this study is related to the need for services post hospitalization," Timothy Lineberry, MD, chief medical officer for the Greater Green Bay Market of Aurora Health Care in Wisconsin and a national expert on suicide risk assessment, told Medscape Medical News.

Dr Timothy Lineberry

"The investigators looked at identifying a group at high risk and potentially targeting services to that specific population," he added.

Dr Lineberry, who was not involved with this research, is the former medical director of the Mayo Clinic Psychiatric Hospital in Rochester, Minnesota, and was in the US Air Force from 1991 to 2003.

"In terms of looking at their actuarial suicide risk algorithm, they found that about 53% of suicides occurred after the 5% of hospitalizations of highest predicted risk. So obviously it's not catching everybody," he noted.

"I think there's the question of, how do you manage to try to get the larger population? The hypothesis about actuarial methods being better than clinical predictive methods by an individual clinician has been shown quite a bit. So those methods do have a place," said Dr Lineberry.

"But from a translational standpoint, there are already a number of standards that address the need for follow-up after psychiatric hospitalization within 1 week."

He noted that, outside of the military, this information is collected through the Healthcare Effectiveness Data and Information Set (HEDIS). And although the current study reported about a particular high-risk group, it could be said that anyone who enters into a psychiatric hospital is already at high risk.

"My argument is based on the standards and recommendations that the UK has in place and that we have in the United States that you need to provide intensive-level services to everyone who is psychiatrically hospitalized, said Dr Lineberry.

"From a purely clinical standpoint, providing for a high-risk person who is hospitalized next to somebody who you felt wasn't quite as high risk would be challenging to implement," he added. "It's also important to keep that connection, making sure there's a seamless transition from inpatient care to outpatient care."

The study authors have reported several potential conflicts, which are listed in the original article. Dr Lineberry is no longer working for the Department of Defense as a consultant but did report seeing some earlier Army STARRS data, although not these findings.

JAMA Psychiatry. Published online November 12, 2014. Abstract


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.