Network analysis of risk factors for firearm assault.

Although individual risk and protective factors have been identified, the extent to which these factors interact within a broader risk network remains inadequately understood. We used network analysis to characterize the partial correlation structure among previously identified risk factors for firearm violence. We used baseline data from a prospective longitudinal study of 1506 participants aged 18-24 recruited from four Level 1 trauma centers in Philadelphia, Flint, and Seattle (2021-2023), 14.4% of which self-reported firearm assault (victimization or aggression, including threats) at baseline. Risk factors included substance use, mental health symptoms, firearm carrying, peer/family factors, and community factors. Regularized partial correlation network models estimated the partial correlation network. We characterized the network using various centrality indices, and discriminatory power of variable domains was quantified by the area under the receiver operator characteristic curve (AUROC). The network showed groupings of mental health symptoms (depression, anxiety, PTSD), substance use and negative behaviors (alcohol use, drug use, negative peer behaviors), and protective factors (resiliency, prosocial attitudes, peer support, positive peer behaviors, parental support). Centrality measures identified PTSD, community violence exposure, negative peer behaviors, and family conflict as strongly interconnected and influential. Nearly all predictive power of the full predictor set (AUROC = 0.86) was captured using these central factors; community violence was the strongest standalone factor (AUROC = 0.81). Firearm violence risk among young adults is shaped by an interconnected network of psychosocial, familial, and community factors. Central risk elements, particularly community violence and PTSD, emerge as high-impact leverage points for trauma-informed, multifaceted interventions addressing social and structural determinants.