2 - SYM (105) Predictive Validity and Treatment Sensitivity of the Death-implicit Association Test in a Sample of Suicidal Inpatients

Background: The most widely used implicit measure of suicide risk is the Death-Implicit Association Test (D-IAT; Nock et al., 2010). However, findings on its validity are mixed and it is unclear if the D-IAT predicts suicidal behavior, particularly among the highest risk populations (Moreno et al., 2022; Sohn et al., 2021). Furthermore, it is not established whether the D-IAT is sensitive to treatment. Thus, it is unclear if associations with death are malleable or if the D-IAT is a suitable measure of treatment success. This study examined whether D-IAT scores changed over the course of inpatient treatment, and whether D-IAT scores predict suicide attempts over a 6-month follow-up period.

Method: 194 inpatients with recent suicide attempts provided valid D-IAT data at intake to a suicide prevention clinical trial (M age = 32.82 [SD = 12.66]; 52.1% female; 24.0% Hispanic/Latino; 64.9% White). Inpatients were randomly assigned to receive treatment as usual (TAU) or TAU plus Brief Cognitive Behavioral Therapy for Suicide Prevention. Participants were re-administered the D-IAT and an objective interview of suicidal behavior post-treatment and monthly during a 6-month follow-up period. Logistic regression was used to examine whether pre- or post-treatment D-IAT scores predicted presence of suicidal behavior during 6-month follow-up. Boot-strapped paired-sample t-tests were used to examine whether D-IAT scores changed from pre- to post-treatment. Repeated measures ANOVA was used to examine whether change in D-IAT scores was influenced by treatment condition.

Results: Stronger implicit associations with death at intake were associated with higher likelihood of aborted attempts over 6-month follow-up: Exp(b) = 10.80, Wald = 7.77, p = .005. Pre-treatment D-IAT scores were not associated with actual (χ2(df=1) = .45, p</em> = .503) or interrupted attempts (χ2(df=1) = .47, p = .491) or preparatory behavior. Post-treatment D-IAT scores were not associated with later suicidal behavior (χ2(df=1) = 1.27, p = .259). Additionally, D-IAT scores did not change from pre- to post-treatment for either treatment condition: M difference = .05 [-.01:.11], df = 171, SE = .03, p = .103.

Conclusion: Findings suggest that the D-IAT may not be sensitive to treatment effects or a useful standalone measure of risk for suicidal behavior in high-risk populations. It may be that implicit associations with death are trait-like and resistant to change in the period following a suicide crisis.