Violence and Suicide in a Nationally Representative Sample of Youth Aged 12-17: What Does It Mean to be at-Risk?

Background. Suicide attempts and school violence, including gun violence, have increased dramatically among youth aged 12-17 since 2019. Gun violence and suicide are now leading causes of death in this cohort. While coarse grouping variables like race, socioeconomic status, and rural/urban location are linked to violence and suicide risk, variable-centered analyses may elide complexity in data, including the presence of unique subgroups, or classes, with distinct patterns of behavior. To improve outcomes for youth vulnerable to violence engagement and suicide, this study applied latent class analysis (LCA) to investigate heterogeneity in patterns of violence and suicide risk behavior and modeled geographic, demographic, and family-level predictors of latent class membership.

Methods. Data were drawn from the 2021 National Survey on Drug Use and Health. The dataset included n=10,743 youth aged 12-17. Latent class indicators included previous year fighting at school, attacking a classmate, arrest, carrying a gun, suicide ideation, and suicide plan. First, an iterative mixture-modelling approach tested latent class models with an additional class specified at each iteration. Model fit was assessed with AIC, BIC, ssBIC, entropy, and substantive utility and interpretability. Second, following LCA model identification, latent class membership was modeled as a categorical dependent variable and regressed on geographic, demographic, and family-level predictors.

Results. A four-class solution best fit the data. A suicide risk class (n=1,172) had high probabilities of suicide ideation (p=0.98) and plan (p=0.58). A violence and suicide risk class (n=236) was had high probabilities of fighting at school (p=.072), attacking a classmate (p=0.58), suicide ideation (p=0.98), plan (p=0.78), and the second-highest probability of carrying a gun (p=0.19). A carceral risk class (n=483) had the highest probability of carrying a gun (p=.33), high probabilities of fighting at school (p=.73) and arrest (p=.90), and moderate probabilities of attacking a classmate (p=.43). A large, low-risk class (n=8,833), had the lowest (p<.09) probabilities of all indicators. Compared to the low-risk class, the suicide risk class had lower odds of being male (OR=0.32), Black (OR=0.66) or from a two-parent home (OR=0.73). The violence and suicide risk class had lower odds of being male (OR=0.60) and higher odds of rural location (OR=1.87). The carceral risk class had lower odds of a living in a two-parent home (OR=0.55) and higher odds of being male (OR=3.13), Black (OR=1.90) and below the poverty line (OR=1.57).

Conclusions: One low-risk class and three classes with heterogenous patterns of violence and suicide risk behavior were evident in these data. The violence and suicide risk class was more likely to be female and from a rural area, where mental health service availability may be limited. The carceral risk class, with the highest probabilities of arrest and carrying guns and greater odds of being male, Black, and in poverty, reflected the persistent racial disparities afflicting Black boys. The suicide risk class had low probabilities of violence engagement and were more likely to be girls from single parent homes. Results highlight the imperative for nuanced, finely targeted interventions to support youth with distinct vulnerabilities to violence and suicide risk behavior.