(PS8-27) Common Factors in Individuals at Risk for Suicide

Abstract Text: Suicide rates increased over the course of the COVID-19 pandemic, especially among youth (Bridge et al., 2023). This series of tragedies generated more widespread attention towards the issue of suicide which remains one of the most important and vaguely understood issues in psychology. Suicide is preventable, not predictable (Roza et al., 2023). Our research aims to identify common factors reported by individuals who report thoughts of suicide and previous suicidal behavior. We hypothesized that these individuals would share a set of common positive factors.

Our first set of analyses was conducted in an existing dataset consisting of self-report scales including the Penn State Worry Questionnaire (PSWQ), the Mood and Anxiety Symptoms Questionnaire (MASQ), and the Positive and Negative Affect Schedule (PANAS), which were administered to 580 undergraduate students on a secure website. A Spearman’s ranked correlation (suited to ordinal data) was performed on each scale to analyze individual items’ correlation with thoughts of death or suicide over the past two weeks, which was rated on a scale from 0 (not at all) to 4 (extremely). We employed correlational heat maps to identify which items in each scale and which subscales were common factors among participants who reported suicidal thought. Results indicated that four individual items (feelings of worthlessness, choking sensations, fear of death, and disgust with self) had moderate to strong positive correlations with thoughts of suicide or death, all r(580)>.45. Only one subscale (the PANAS guilt subscale), had a moderately strong correlation with suicidal thought, r(580) = 0.41.

Our second set of analyses uses the National Comorbidity Survey Replication (NCS-R) collected between 2001 and 2003 to assess the prevalence and correlates of mental disorders in the US (Kessler & Merikangas, 2004). We are using canonical correlation, a statistical machine-learning method used to analyze the statistically significant relationship(s) between sets of variables (Wang et al., 2023). Machine-learning algorithms with previous good performance have potential as preventative measures for suicide (Roza et al., 2023) and research using a machine-learning algorithm with medical records had moderate success in predicting suicidal behavior following discharge from inpatient treatment (Simon et al., 2024). Our analyses use canonical correlation to test for strong relationships between suicidal behavior items and other cognitive and behavioral variables in the NCS-R. We hope that this research will better equip clinicians to identify risk factors in their patients.