4 - (SYM 11) Identifying Suicide Risk with Digital Biomarkers of Emotional States

Suicide risk is a major problem, especially in a range of vulnerable populations. A major challenge for managing suicide risk is that we have poor methods of identifying and predicting those who are likely to suicide. This study attempts to identify people who are at risk for suicide by using digital markers  that have been validated to measure different emotional states. We assessed 150 war veterans with and without posttraumatic stress disorder (PTSD), who completed the Suicidal Ideation Attributes Scale, and were videorecorded while talking about their traumatic events. Recorded data were analysed using AI-based software that integrates facial, acoustic, and speech data derived from recordings. Overall accuracy in identifying suicidality was 0.78. This report will also outline the capacity of the AI-generated assessment to predict suicidality 6 months after initial assessment. To provide independent validation of these results, comparable analyses were conducted with a sample of 140 bereaved individuals with a range of suicidal risk. Taken together, these findings point to the utility of AI-based analyses of facial, acoustic, and speech data to detect suicidal risk.