• Kelly V Ruggles and Sonali Rajan.
• Center for Bioinformatics and Health Informatics, New York University Langone Medical Center, New York, New York, United States of America.
• Plos One. 2014 Jan 1; 9 (11): e111893.

ObjectiveTo apply discovery-based computational methods to nationally representative data from the Centers for Disease Control and Preventions' Youth Risk Behavior Surveillance System to better understand and visualize the behavioral factors associated with gun possession among adolescent youth.ResultsOur study uncovered the multidimensional nature of gun possession across nearly five million unique data points over a ten year period (2001-2011). Specifically, we automated odds ratio calculations for 55 risk behaviors to assemble a comprehensive table of associations for every behavior combination. Downstream analyses included the hierarchical clustering of risk behaviors based on their association "fingerprint" to 1) visualize and assess which behaviors frequently co-occur and 2) evaluate which risk behaviors are consistently found to be associated with gun possession. From these analyses, we identified more than 40 behavioral factors, including heroin use, using snuff on school property, having been injured in a fight, and having been a victim of sexual violence, that have and continue to be strongly associated with gun possession. Additionally, we identified six behavioral clusters based on association similarities: 1) physical activity and nutrition; 2) disordered eating, suicide and sexual violence; 3) weapon carrying and physical safety; 4) alcohol, marijuana and cigarette use; 5) drug use on school property and 6) overall drug use.ConclusionsUse of computational methodologies identified multiple risk behaviors, beyond more commonly discussed indicators of poor mental health, that are associated with gun possession among youth. Implications for prevention efforts and future interdisciplinary work applying computational methods to behavioral science data are described.

11 keywords...

hide…