Symposia
Suicide and Self-Injury
Amanda M. Raines, Ph.D.
Clinical Investigator
Southeast Louisiana Veterans Health Care System
New Orleans, Louisiana
Jamie Tock, PhD
Health Scientist Specialist
San Francisco VA Health Care System
San Francisco, California
Claire Houtsma, Ph.D.
Clinical Investigator and Suicide Prevention Coordinator
Southeast Louisiana Veterans Health Care System
New Orleans, Louisiana
Kathryn Macia, Ph.D.
Postdoctoral Fellow
VA Palo Alto Health Care System
Livermore, California
Jane Herwehe, MPH
Data Action Team Lead
Louisiana Office of Public Health
New Orleans, Louisiana
Joseph Constans, Ph.D.
Senior Manager for Suicide Prevention
VA Office of Research and Development
New Orleans, Louisiana
Despite the well-established link between firearm access and suicide, little is known about other variables that may influence risk for death by self-inflicted gunshot. As individual factors have demonstrated limited predictive ability, scholars have called for studies that consider the multifaceted relations between myriad variables. One alternative to the typical cause and effect approach for investigating various forms of psychopathology is network analysis. Based on the premise that maladaptive outcomes are the result of multifaceted associations between biological, psychological, and social factors, network analysis is a process that allows the visualization and quantification of interrelations between numerous factors associated with a problematic outcome. For these reasons, network analysis has become increasingly popular in the field of psychological science. However, little research has applied this method to suicidal outcomes, and none have done so in the context of a veteran population. Data from 33,010 male suicide decedents (92% White; Mage = 59.02, SD = 18.53) were acquired from the National Violent Death Reporting System (NVDRS) to investigate characteristics of veteran suicide decedents who died by self-inflicted gunshot (71%) versus alternative methods (29% e.g. poisoning, hanging). Firearm and Non-Firearm networks were estimated separately using the Ising model for binary data. Network stability, edge-weight, and node centrality (expected influence) metrics were computed for each network with bootstrapped means and standard deviations to facilitate network comparison. Network comparison tests identified significant between-network differences for 56 edge-weights. Between-network correlations (r = .66 to .95) suggested moderate to strong correspondence in the relative strength of edges across networks. Problems with alcohol was significantly more central to the Firearm network and had the strongest direct influence in exacerbating other risk factors within this network, whereas depressed mood was significantly more influential in exacerbating risk factors in the Non-Firearm network. Problems with an intimate partner had a large total influence across both networks in exacerbating risk, although notably these problems interacted more strongly with alcohol-related problems in the Firearm network and depressed mood in the Non-Firearm network. Findings will be discussed with regard to previous research and clinical implications.