(PS11-8) A Network Approach to Assessing Suicide Risk Detection Using Objective Tasks

Detection of suicidal thoughts and behaviors (STBs) is often limited by reliance on self-report. Objective tasks designed to measure suicide-specific cognitive biases can differentiate individuals with and without STBs (Moreno et al., 2022). However, objective tasks may not equally predict STBs (Cha et al., 2018), and may differ from self-report variables in their association with suicide-related risk factors (Tucker et al., 2017) The present study used network analysis to explore relationships between task performance, STBs, and self-report constructs.

Undergraduates (n = 88) were recruited for a larger study of suicide risk detection using objective computer tasks. Participants completed self-report measures assessing STBs and other mental health constructs including suicidal ideation (SI) intensity, frequency of suicidal mental imagery, wish to live, wish to die, fearlessness about death, practical capability for suicide, entrapment, and defeat, followed by the Suicide Affect Misattribution Procedure (S-AMP, Tucker et al., 2017), Death/Life Implicit Association Task (DS-IAT, Nock et al, 2010), and a Free-Viewing Eye Tracking Task (FVET). The DS-IAT provides a score of implicit association with suicide by subtracting reaction time of sorting words into “self or life” from “self or death.” The S-AMP measures self-concept rating scores of an unrelated stimulus after viewing one of four types of stimuli, and rating scores following suicide stimuli indicate strength of suicide association. The FVET collects time spent viewing four types of images. Longer fixation time on suicide images indicates stronger selective attention toward suicide.  Participants were randomized to task order.

We generated a network displaying partial correlations between all variables of interest, using LASSO (Least Absolute Shrinkage and Selection Operator) with Extended Bayesian Information Criterion (EBIC) model selection in R, which retains only correlations greater than 0.25 (Epskamp et al., 2018).

S-AMP and FVET scores were moderately associated with each other, while IAT scores were weakly associated with FVET scores and not related to S-AMP scores. FVET fixation time was moderately associated with frequency of suicidal mental imagery and practical capability and weakly associated with SI intensity and wish to die. Task outcomes were not associated with attempt frequency. IAT scores were weakly associated with entrapment. S-AMP scores were moderately associated with practical capability and fearlessness about death, and weakly associated with defeat, entrapment, wish to live, and wish to die. Self-report measures tended to be moderately to strongly associated with each other. The full network model will be presented in the poster.

The S-AMP and FVET were moderately related to some self-reported suicide constructs, particularly those related to capability. These tasks may be more informative of the capacity for suicidal behavior relative to current SI frequency. Future research could investigate how recency of suicidal behavior impacts S-AMP and FVET scores and their associations with suicidal capability to determine the best time to intervene in suicidal behavior.