Although there are several evidence-based cost-effective interventions for people with opioid use disorder (OUD), they are underutilized. Questions remain regarding intervention selection, and cost of service delivery. Simulation modeling offers an opportunity to support decision making to address the syndemic of opioid overdose, HIV, and hepatitis C (HCV). In an article recently published in the International Journal of Drug Policy, the OUD Modeling writing group described the benefits of using simulation models for OUD-related clinical and public health decision making and the different types of simulation models available. They also identified the need to collect and synthesize local data to model sub-epidemics and calibrate models to support local decision making.
The authors outlined several benefits of simulation modeling that allow researchers to study scenarios that would not be possible to study in the real world at the necessary pace required for decision making. Simulation models can translate results from small-scale trials to the population level. They can also be used to explore uncertainty regarding current knowledge in order to highlight important individual-level, and system-level factors that drive outcomes. In doing so, they can explore the health effect and costs of different intervention combinations over long periods of time that exceed what would be feasible during a clinical trial. Additionally, models can examine benefits and costs from multiple perspectives such as the healthcare system, the criminal justice system, or the broader societal perspective. This can help inform the priorities of future research and provide real-time estimates for policymakers to make decisions.
Model selection depends on the research question, whether the model needs to account for interactions between individuals, the time horizon, and the granularity of the data available. Various types of models can be used to address these questions, including Markov models, micro-simulation models, population-based compartmental models, and agent-based models. The authors provide examples of how model selection can be “fit to purpose.” CHERISH currently offers consultation services to provide guidance for building the appropriate model.
Ultimately, model selection also depends data availability. The opioid crisis in North America is made up of several local epidemics that vary regarding drug supply, demographic characteristics and treatment capacity. These factors affect intervention selection, the effectiveness of the intervention and the financial capacity for implementation and scale up. In order to inform policymakers at a local level, the authors identify the need for local context-specific data. For example, a recent study by CHERISH Research Affiliate Bohdan Nosyk and colleagues modeled local HIV epidemics in six US cities to recommend cost-effective intervention combinations that met local needs.
To maximize the impact of opioid models, the authors emphasize that substantial efforts will be required to collect, analyze, and synthesize the data needed to support model development. The authors encourage the modeling community to collaborate to share model code and pool resources to generate critical inferences; produce, adapt and cross-validate models; and effectively disseminate findings. They argue that the opioid crisis is unprecedented and the modeling community should recognize the need for a collaborative approach in order to respond as effectively as possible, while acknowledging the challenge to manage collaborations so that they produce timely results and address the questions of greatest relevance to policymakers at the local and regional level.