System dynamics modeling to inform implementation of evidence-based prevention of opioid overdose and fatality: A state-level model from the New York HEALing Communities Study

Although opioid-related fatalities decreased in the United States (US) and New York State (NYS) from 2022 to 2023, fatality remains high (81,083 (US) and 5,308 (NYS) in 2023) after years of unprecedented increases of fatal and non-fatal overdoses (Centers for Disease Control and Prevention, National Center for Health Statistics, 2021, 2024; National Institute on Drug Abuse, 2023). A dramatic rise in the availability of illicitly manufactured fentanyl has also been documented in the US and NYS, resulting in a more potent opioid supply (Kilmer et al., 2022; New York State Department of Health, 2023a). Intentional and unintentional exposure to fentanyl among people who use drugs has been associated with increased risk of overdose and death (Hughto et al., 2022). Fentanyl co-involved with psychostimulants, benzodiazepines, and xylazine may characterize a new wave of the opioid epidemic (Ciccarone, 2021a; Friedman & Shover, 2023; Jenkins, 2021).

In 2019, the National Institute on Drug Abuse (NIDA) funded the HEALing (Helping to End Addiction Long-term®) Communities Study (HCS), a large implementation research project designed to reduce opioid fatalities, increase access to medications for opioid use disorder (MOUD), and reduce stigma toward people on MOUD (National Institutes of Health HEAL Initiative, n.d.; The HEALing Communities Study Consortium, 2020). The HCS employed a coalition-driven intervention to inform the deployment of evidence-based practices to rapidly reduce opioid-related overdoses and fatalities in 67 highly affected communities in NYS, Kentucky, Massachusetts, and Ohio. Through a data-driven approach to community-engaged planning and action, the HCS sought to learn how to increase the reach of evidence-based harm reduction and treatment interventions (Chandler et al., 2023; Chandler et al., 2020; El-Bassel et al., 2021).

System dynamics (SD) modeling was incorporated to support the HCS in NYS to engage community coalitions. SD models use feedback loops (i.e., closed sequences of time-dependent causal relationships) to hypothesize the endogenous drivers of a system’s behavior over time (Richardson, 2011). These feedback loops are able to capture accumulation processes, nonlinearities, and time delays to gain insight into the causal nature of complex problems (Yasarcan, 2023). SD models also serve as tools to help diverse community members build a shared appreciation of why systems problems manifest and persist, how such problems can be resolved, and what can be done to mitigate unintended consequences of policies and practices (Forrester & Senge, 1980; Senge & Sterman, 1992). Simulation analyses can then test policy interventions and assess possible intended and unintended consequences (Sterman, 2006).

Prior publications have described SD models of earlier waves of the US opioid and non-opioid drug epidemics (Levin et al., 1972, 1975; Homer, 1993, 1997; Wakeland et al., 2011, 2013, 2015, 2016; Homer & Wakeland, 2021; Lim et al., 2022; Stringfellow et al., 2022; Sabounchi et al., 2023). The earliest model examined the 1970s heroin epidemic in a New York City neighborhood characterized by high rates of youth heroin use (Levin et al., 1972, 1975). This model included feedback loops capturing the heroin supply, community education, policing, and incarceration, among others. Though not calibrated to historical data, the model suggested that a comprehensive set of policy interventions were needed to curb the epidemic. Another early illicit drug model studied the US cocaine epidemic of the 1970s and 1980s (Homer, 1993, 1997). A key feedback loop of this model showed how the popularity of cocaine drove an increase in its use. By highlighting time delays and gaps in data reporting of drug use, the model pushed back against the then-current idea that drug seizure policies were effective at reducing cocaine use prevalence.

More recently, Wakeland et al (2011, 2013, 2015, 2016) modeled excessive opioid prescribing practices in the US and the diversion of pharmaceutical opioids to the illicit market through 2011. An update extended the model’s boundary to include the effects of fentanyl in the illicit drug supply after 2013 (Homer & Wakeland, 2021). Another update incorporated additional structures for MOUD, naloxone use, supply-side changes on prescription opioids, and the perceived risk of overdose fatality (Lim et al., 2022; Stringfellow et al., 2022).

Building upon these earlier SD models and adding additional structures identified in our preparatory qualitative modeling of the opioid epidemic (Sabounchi et al., 2023), we present here an opioid SD model built to support implementation of the HCS in NYS and the short- and long-term effects of simulated strategies around opioid overdose education and naloxone distribution (OEND) and MOUD.