PROJECT SUMMARY/ABSTRACT (PROJECT LEADER CHAN) The over 90,000 U.S. drug overdose deaths in 2020 resulted primarily from people who started using drugs, and escalated their use, due to interacting factors of individual attributes and social network influence. Timely allocation of intervention resources has the potential to effectively reduce the harm and associated personal, healthcare, and societal costs of those impacted by substance use and misuse. There are currently no data- driven tools that allow stakeholders to predict an individual’s substance use, yet such tools would be valuable in identifying and aiding those most at risk of developing high rates of substance use. Current methodologies only rely on trend analyses, establishing correlations between risk factors such as friends that use drugs and substance use. A key challenge when modeling future substance use lies in the co-evolution of behaviors like drug use and peer networks of risk/support, which can change over time and may depend on each other. To untangle the co-evolution process of individual attributes and peer networks, the objective of this project is to develop separate probabilistic models that forecast individuals’ patterns of future substance use at short and long timescales and peer risk/support networks at long timescales. This will contribute to the long- term goal to develop stochastic simulations of realistic substance use disorder-related behavioral contagion in plausible dynamic networks to inform resource allocation and contingency planning. Using deidentified secondary data from the Rural Drug Addiction Research Center’s (RDAR) Regional Health Cohort, the project team will first model how individual attributes such as drug use and adverse childhood history, along with peer networks that include the extent of peer and confidant drug use, impact an individual’s substance use behavior. They will also model how individuals’ peer risk/support network links form or break in response to similarities or differences in attributes, including their drug use behaviors. This project is expected to result in the first stochastic forecasting models using data on covariates of individual attributes and peer network features. The models will be designed to predict future substance use and future peer risk/support networks at long timescales within months (Aim 1) and future substance use at short timescales within days (Aim 2). Successful machine-learning methods will be used to build these models and rigorously assess model generalizability/prediction performance (Aims 1 and 2). Project findings will provide a foundation for Project Leader Hau Chan to submit an innovative NIH R01 application to develop stochastic simulations of realistic substance use disorder-related behavioral contagion in plausible dynamic networks, which will ultimately inform resource allocation and contingency planning and lay the groundwork for just-in-time interventions to detect imminent increases in f...