Given the high rates of Veteran suicide each year, suicide prevention has become a VHA strategic priority. While progress has been made in screening Veterans for suicide risk, high rates of false positives remain a challenge, potentially diverting clinical resources away from those Veterans who need them most. Innovative new approaches to assess suicide risk are needed, that accurately identify short-term (e.g. coming weeks) risk of suicidal behavior, and that are based on objective markers rather than relying primarily on self- report. Recent work has identified several cognitive domains, including impulse control and inhibition, increased distractibility, implicit semantic content, and processing of rewards/punishments, that may all be altered in suicidal individuals. The current project will use a battery of computer-based tests to assess these cognitive domains, followed by computational modeling to extract additional meta-variables representing latent cognitive processes, to prospectively predict short-term risk of suicidal behavior in high-risk Veterans. Veterans will be recruited from acute in-patient psychiatric units following a suicidal event, and tested at 3-month intervals for a year; we will also record suicidal behavior (SB), defined as an actual, interrupted, or aborted attempt or a behavior preparatory to suicide.
In Aim 1, we will use neurocognitive task scores collected shortly before the SB, to determine which tasks (and therefore, alterations in which cognitive domains) can prospectively predict short-term risk of SB (within the next 3 months).
In Aim 2, we will conduct computational modeling on the behavioral data, to extract additional meta-variables, describing latent cognitive processes such as response caution (impulsivity); subjective value of rewarding, punishing, and neutral feedback; perseveration; and tendency to explore new responses, to determine which of these meta-variables can prospectively predict SB.
In Aim 3, we will combine these neurocognitive task scores and meta-variables, along with standard indicators of suicide risk such as demographic, self-report, and clinical assessment, in a statistical prediction model, to determine whether inclusion of these cognitive variables and meta-variables can significantly improve prediction of short-term risk for SB in this high-risk Veteran sample. The results of this study will provide much-needed information to improve identification of Veterans likely to make a suicide attempt in the upcoming weeks, so that clinical resources can be targeted to help them. The computer-based neurocognitive tests represent relatively objective behavioral markers that could be deployed fairly simply in a clinical setting or even via telehealth. Additionally, our focus on underlying cognitive processes, as modifiable factors, will improve the ability to optimize intervention by suggesting specific therapeutic approaches based on an individual Veteran's cognitive profile.

Public Health Relevance

/Health Relevance About 5,000 Veterans take their own lives each year. While the VA has currently mandated a multifaceted prevention approach, including suicide risk screening of all Veterans seen for mental health issues, the continued large numbers of Veterans who attempt or die of suicide points to the desperate need for improved prediction of suicide risk. Existing prediction methods are subject to high rates of false alarm, so that only about 10% of Veterans flagged as at-risk actually attempt suicide; this high false alarm rate potentially diffuses critical clinical resources away from those who need them most. The development of more accurate predictive tools has been identified as one of the most important areas in suicide research, and would also improve the evidence base for clinical therapy, by more accurately identifying at-risk clients to participate in research trials and evaluating treatment effectiveness across time.

National Institute of Health (NIH)
Veterans Affairs (VA)
Non-HHS Research Projects (I01)
Project #
Application #
Study Section
Special Emphasis Panel (ZRD1)
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
VA New Jersey Health Care System
East Orange
United States
Zip Code