Optimal Methods for Estimating Policy Effect Heterogeneity in Opioid Policy Research

点击翻译按钮获取中文摘要

Methods Research Project Abstract The nation’s opioid crisis continues unabated. The number of fatal overdoses increased nearly 30% in 2020, reflecting the continued spread of fentanyl as well as the increased stress, social isolation, job loss, and reduced access to treatment imposed by the COVID-19 pandemic. States continue to enact a diverse array of opioid-related policies, producing a complex and dynamic policy landscape. A core objective in opioid policy research is to accurately identify policies that will achieve the intended outcomes, but state-level evaluation studies must grapple with the limited statistical power afforded by the constrained sample size. Furthermore, there is heightened interest in understanding the extent to which the intended gains from specific policies are experienced equitably by all population subgroups and whether there are characteristics of states for which specific policies are more or less effective. This study seeks to provide opioid policy researchers with clear statistical guidance and novel statistical methods to conduct complex, high-quality policy evaluations. Opioid policy effectiveness may vary along multiple dimensions including time and state-, community-, and individual- level factors. To date, new econometric methods have focused on time-varying effects. In this context, traditional difference-in-differences (DID) estimates have been shown to be biased towards zero. In contrast, there has been limited attention to methods for examining differential policy effectiveness across state-, community-, or individual-level factors, yet these types of policy effect heterogeneity and have important ramifications for achieving equitable opioid-related outcomes. Obtaining robust effect estimates will require methods that extend easily to multilevel settings and borrow information across observations and levels. In this project, we will provide a comprehensive summary of the state of the statistical science in the opioid policy space with regard to estimation of policy effect heterogeneity. We will also develop novel statistical methods to address the complexities and sample size constraints inherent in examining policy effect heterogeneity. Finally, we will create a series of simulation tools that will inform and improve the methods opioid policy researchers and policy researchers more broadly utilize to determine which policies are most effective for whom within the opioid crises. The work will make both short- and long-term contributions by shifting and improving the statistical practice of opioid policy researchers, thereby improving the accuracy of opioid policy studies and thus, the quality of evidence regarding which policies are most effective and for whom. Our overall goals are to give applied policy researchers insights regarding the relative advantages and limitations of alternative methods for estimating heterogeneous policy effects and to develop novel statistical methods and accessible statistical guidance for conducting opioid policy evaluations, thereby providing the field with better tools for implementing such research.