How simulation modeling can support the public health response to the opioid crisis in North America: Setting priorities and assessing value

Opioid-related harms are a major public health concern in North America. In the United States and Canada, respectively, over 72,000 and about 4100 people died of a drug overdose in 2017, the vast majority of which were opioid-related (Government of Canada, 2018; National Institute on Drug Abuse, 2019). Opioid-related overdose deaths in the United States have been increasing on an exponential trajectory since 1979, with the introduction of illicit synthetic opioids driving the most recent increases (Jalal et al., 2018). Overdose is one manifestation of the high prevalence of opioid use disorder (OUD, Bose et al., 2016) other outcomes include localized outbreaks of bloodborne infections, particularly HIV and hepatitis C virus (HCV), and rising incidence of endovascular infections (Bates, Annie, Jha & Kerns, 2019; Inocencio, Carroll, Read & Holdford, 2013; Schwetz, Calder, Rosenthal, Kattakuzhy & Fauci, 2019; Springer, Korthuis & Del Rio, 2019; Strassels, 2009; White, Birnbaum & Mareva, 2005). Questions about the long-term management of people with OUD – from availability of harm reduction services to use of medications to treat OUD (MOUD; terminology used to distinguish pharmacotherapies from psychosocial care often associated with treatment Blanco & Volkow, 2019) to increasing access to HCV treatment – have become priority topics in public health and healthcare delivery. Simulation models can guide and inform clinical and public health responses to the opioid crisis. Such models integrate understanding of OUD with the best available data, typically synthesizing information from multiple sources to simulate a population of interest, predict future trajectories of drug use and associated complications, and project both individual- and community-level impacts of interventions and policies. These outcomes frequently include health effects and economic costs, but can also include effects related to crime, housing, and other social sectors. Further, models can be used to simulate studies that may be unethical, controversial, infeasible, or prohibitively expensive to conduct using traditional designs. Finally, simulation models can also explore how evidence generated in one geographic or policy context can be applied to different contexts. Simulation models thus provide a structured, in silico environment for decision makers to compare the predicted effects of proposed policy alternatives over an extended time horizon within a causal inference framework before mak