Injuries/Violence
: Independent and Joint Impacts of Alcohol and Cannabis Outlet Densities on Self-Harm Injuries in California: A Bayesian Spatiotemporal Analysis Rafael Charris* Rafael Charris Ellicott Matthay
Background: With the rapid expansion of state recreational cannabis legalization in the US, it is vital to understand its impact on alcohol-related harms, including self-harm injuries. This study explores the interactive impacts of changes in the physical availability of recreational cannabis and alcohol on self-harm injuries in California.
Methods: We integrated statewide data on self-harm injuries (2017-2019) across 1604 ZIP Code Tabulation Areas (ZCTAs), incorporating death, emergency department visits, and hospitalization records alongside geocoded alcohol outlet listings and storefront recreational cannabis outlet listings. We used Bayesian spatiotemporal models to account for small-area spatial dependence, modeling quarterly ZCTA-level self-harm rates. The models examined alcohol and cannabis outlet densities and their interaction, adjusted for sociodemographic confounders. From the multiplicative differences that the model returns, we applied g-computation to calculate additive risk differences (RDs) corresponding to hypothetical interventions on alcohol and cannabis outlet densities.
Results: Adjusted models indicated that eliminating cannabis outlets was associated with a slight but imprecise reduction in nonfatal self-harm injuries (nonfatal RD: -0.37; 95% CI: -1.31, 0.60). Regarding alcohol outlets, a 30% reduction in these outlets’ densities was associated with a 2.5 per 100,000 decrease in nonfatal self-harm injuries (95% CI: -3.6, -1.44). These trends varied by demographic subgroup and injury means. Notably, no interaction was observed between alcohol and cannabis outlet reductions.
Conclusions: The introduction of storefront recreational cannabis outlets in California does not appear to have contributed to changes in rates of nonfatal self-harm injuries, but alcohol outlet densities were strongly correlated with rates of nonfatal self-harm injuries. Reducing alcohol outlet density is a candidate intervention for reducing self-harm. This analysis illustrates a novel approach for leveraging the posterior distributions of multiplicative-scale Bayesian spatiotemporal models to report policy-relevant additive-scale risk differences that account for spatial autocorrelation.