Computational modeling to evaluate interventions for HIV and substance use

Background: People at risk for HIV often experience co-occurring substance use disorders. Shifting trends in the opioid epidemic have been accompanied by increases in methamphetamine and polysubstance use, which also increase risk for HIV. Evidence suggests that factors such as housing instability, incarceration, and unemployment may interfere with engagement in HIV prevention and care, and these factors are also associated with substance use. Because such interventions are resource intensive and logistically challenging, particularly for those who are less likely to engage in research in traditional settings, guidance is needed at the intervention development stage to determine the most impactful and efficient intervention strategies. Agent-based models (ABMs) can be used to virtually evaluate candidate interventions to facilitate more efficient and timely intervention development. Because they allow for the conduct of counterfactual experiments, ABMs can also facilitate identification of effects that would be difficult to identify using traditional statistical approaches and can provide valuable insights to understand causal mechanisms that give rise to complex systems. Objective: Building on an existing ABM platform, this proposal will utilize multiple existing data sources to characterize relationships among demographic and behavioral risk factors, substance use, mental health, and HIV prevention and care engagement among people at risk for HIV. We will combine methods from epidemiology, ABM, and robust decision making (RDM) to understand the potential impact of interventions for reducing substance use, overdose, and HIV transmission. Methods: We will apply statistical and computational methods to better understand how individual demographic, behavioral, and contextual factors, substance use, and mental health impact engagement in HIV prevention and care. We will then conduct a series of experiments to evaluate how these factors impact the uptake of existing biomedical interventions and compare outcomes under scenarios with different combinations of interventions using RDM. Significance: A better understanding of where and how to focus intervention efforts offers potential to improve substance use and HIV prevention and care outcomes among people at risk for HIV. Once developed, our methods and models can be adapted to other geographic areas to reflect local prevention priorities and can serve as an example application of epidemiology, ABM, and RDM methods to advance HIV and substance use prevention science.