Predicting suicide risk remains a challenge. We examined whether neurocognitive performance on implicit associations toward suicide, motor speed, response inhibition, and executive functioning predicts suicide attempt and behavior in high-risk psychiatric patients.

Our sample (N = 298) consisted of inpatients (n = 161) and outpatients (n = 83) admitted for a suicide attempt (SA; n = 78), for suicidal ideation (SI; n = 76), or were non-suicidal psychiatric controls (PC; n = 90), and healthy controls (HC; n = 54). Participants were followed for 12 months, with follow-up assessments at 3-, 6-, and 12-months. Neurocognitive tasks were administered at baseline. Clinical symptom measures, suicidality, and electronic health record data were collected at each timepoint. ANCOVA was used to compare groups on neurocognitive performance, and logistic and Cox regressions examined whether neurocognitive performance predicted future actual suicide attempt and suicidal behaviors.

Participants had a mean age of 24.34 years (SD = 3.71). A total of 19 participants made an actual suicide attempt during the study. On neurocognitive tasks at baseline, the SA group had stronger implicit associations with death- and suicide-related words compared to the HC (d = 0.88, p < 0.001) and SI (d = 0.63, p = 0.005) groups and poorer executive functioning than the SI (d = 0.44, p = 0.043) group in multivariate models. Stronger implicit associations with death/suicide predicted higher risk of suicide attempts at the univariate (HR = 1.68 p = 000), but not multivariate level (HR = 1.17 p = 000), while slower motor speed predicted actual suicide attempts (HR = 1.81 p = 000) at the multivariate level.

Slower motor speed predicts actual suicide attempt and may help identify psychiatric patients who are at high risk for suicidal behavior.