A0645
Title: Bayesian evaluation of local policy effects on overdose outcomes
Authors: Samrachana Adhikari - NYU School of Medicine (United States) [presenting]
Abstract: The United States drug overdose epidemic is a public health emergency resulting in a record number of overdose deaths. In response to the epidemic and to prevent and reduce harms from problematic drug use, many states have implemented overdose prevention and harm-reduction policies. In addition to the state laws, many cities have also created their own set of local policies specific to the conditions of the epidemic, unique to them. Despite the potential key role of local policies in reducing overdose outcomes, evidence of their population-level effectiveness is scant and contradictory, and there remain critical unanswered questions. A number of methodological challenges, due to small area data, co-occurring policies, heterogeneous policy environment, and exposure-confounder feedback, among others, have been major barriers to their evaluation. A Bayesian causal inference framework is presented to evaluate the impacts of local harm-reduction policies, adjusting for co-occurring policies and treatment-confounder feedback. The proposed longitudinal G-computation method with flexible Bayesian models incorporates random effects and adjusts for co-occurring policies through shrinkage priors to precisely estimate effects.
