Injuries/Violence
Importance of Violence Prevention Approach Type for Predicting Neighborhood Violence Rose Kagawa* Rose Kagawa Veronica Pear Colette Smirniotis Alex Kwong Hannah Laqueur
Background
Violence is a leading cause of death, disability, and despair, and the effectiveness of efforts to intervene on the social determinants of violence is not well understand. This study sought to determine what types of violence prevention efforts contribute most to the prediction of neighborhood violence rates.
Methods
Our study population included all census tracts in Cleveland, Ohio with complete data and >0 residents (N=169). The primary outcome of interest was the rate of major violent crime (homicide, rape, robbery, and aggravated assault) in 2019. Our main exposures were counts of 16 different violence prevention approaches (e.g. family relationship programs; anti-poverty policies, job readiness programs). To identify violence prevention approaches, we conducted an extensive review of public facing documents, resulting in 2,522 distinct intervention-locations that met our eligibility criteria. We grouped interventions by approach type following the CDC Technical Package for the Prevention of Youth Violence and other sources, then aggregated these approaches by year and census tract for the years 2015-2019. An additional 54 variables described the demographic, social, and physical features of Cleveland’s census tracts over the study period (e.g. education levels, walkability, building type, age distribution). We used random forest to identify the violence prevention approaches that were most predictive of the rate of violent crime in 2019 and the degree to which the inclusion of these approaches improved the predictive ability of the model.
Results (preliminary)
Including violence prevention approaches in the prediction model improved model performance (MSE with=50.6 vs MSE without=64.7). However, no violence prevention approaches were ranked in the top 10 for variable importance. Among violence prevention approaches, investments in the physical environment were ranked as most important, following by approaches that reduced community-level risk factors (e.g. anti-poverty policies), and career training opportunities.
Conclusions
This study identifies specific approaches to violence prevention that are most important for predicting future violence rates for a single city. This information can guide future evaluation efforts attempting to identify causal effects.