Suicide is a major public health concern in the United States, accounting for over 45,000 deaths in 2020 (CDC, 2020). To support their vision for a “nation free from the tragic experience of suicide,” the National Action Alliance for Suicide Prevention (NAASP) established a research agenda to reduce the annual suicide rate 20 % by 2025 (NAASP, 2014). This agenda includes identifying improved capabilities to better detect individuals at highest risk for suicide across care settings. An existing limitation in our ability to effectively assess suicide risk is the almost exclusive reliance on self-report (Franklin et al., 2017; Kleiman et al., 2017). This can be especially problematic due to the transient nature and potential reluctance of individuals to express suicidal thoughts and behaviors (Carter et al., 2017; Franklin et al., 2017; Nock and Banaji, 2007). Detecting implicit associations with death and suicide has been offered as a more objective marker of suicide risk that can provide a complementary approach to patient self-report (Glenn et al., 2017a, b; Millner et al., 2018; Ellis et al., 2016).

The most widely studied measure of these implicit associations is the Death/Suicide Implicit Association Test (d/s-IAT) (Barnes et al., 2017; Nock et al., 2010; Sohn et al., 2021). The d/s-IAT examines implicit self-identification with death versus life by evaluating the relative difference in response times when life vs. death/suicide related stimuli are associated with the self (Nock et al., 2010; Nock and Banaji, 2007). Prior research has demonstrated that individuals at risk for suicide respond relatively more quickly than non-suicidal individuals to stimuli associating “death/suicide” and “me”, than when “life” and “me” are paired (Barnes et al., 2017; Glenn et al., 2017a,b; Nock et al., 2010; Tello et al., 2020). In a study by Nock et al. (2010), adults presenting to a psychiatric emergency department following a suicide attempt had relatively faster implicit associations pairing the self with death/suicide compared to those with other psychiatric emergencies. These patients also had a six-fold increase in the risk for a suicide attempt during the 6-month follow-up period (Nock et al., 2010). Similar findings were obtained in veterans in a psychiatric inpatient setting (Barnes et al., 2017), with veterans expressing relatively faster associations between self and death (compared to self and life) twice as likely to attempt suicide within 6 months of the baseline assessment (Barnes et al., 2017).

Evidence regarding the d/s-IAT and suicide risk has been inconsistent. In a study by Rath et al. (2021), d/s-IAT scores were not predictive of suicide attempts in adult psychiatric inpatients above and beyond depression, prior suicide attempts, and suicidal ideation. Glenn et al. (2019) found the d/-s-IAT was significantly associated with future suicide ideation in youth aged 12–19 years receiving psychiatric outpatient treatment, but not when controlling for baseline ideation. An additional study by Brent et al. (2021) among youth in pediatric emergency departments found the d/s-IAT to be only a modest predictor of future suicide attempts, with self-report of known suicidal risk factors showing greater predictive accuracy.

While research involving the d/s-IAT has been conducted with mostly younger populations, or populations within a relatively restricted age range (Barnes et al., 2017; Brent et al., 2021; Nock et al., 2010; Glenn et al., 2019), little is known about the discriminating power of the d/s-IAT in older adults. The current study examined whether implicit associations with death/suicide distinguished participants with depression and prior suicide attempts, participants with depression and no attempts, and healthy controls across the adult lifespan. Based on existing literature, we hypothesized that participants with prior suicide attempts would have higher difference scores on the d/s-IAT (faster relative association between self and death than to self and life) compared to participants with no attempts and healthy controls, and that participants with no attempts would have higher difference scores relative to healthy controls. We also sought to determine how age itself might affect the association between group status and d/s-IAT performance. The IAT outcome measure relies on reaction time differences between different conditions, and it is possible that age effects on reaction times may have an impact on the discriminating power of these scores. The present study examined participants aged 16 to 80 years, allowing for analysis of these aging effects, in addition to standard group differences.