ABSTRACT The widespread availability of synthetic opioids (fentanyl) has fueled the rapidly rising rates of unintentional over dose (OD) fatalities. Policy makers, state and local agencies, and investigators have focused on the Ohio experience as a bellwether for the experiences of other states because of the representativeness of Ohio’s demographics. The lack of timely geospatially-linked longitudinal data sources has impeded the ability of communities and state agencies to pivot allocation of resources to regions where they are needed most. Limited integration of environmental risk factors such as sociodemographic characteristics fail to support identification of new targets for intervention or new approaches to emerging threats from changes in local drug supply. We believe that agile data systems and informatics tools that can be used to demonstrate the utility of predictive analytics and machine intelligence approaches on how to enable data-driven decision making will ultimately prove translatable across the country and diverse localities. Our proposed Opioid and Substance Use Disorder Data Enclave (O-SUDDEn) will provide a novel and transformative approach to support rigorous and reproducible research on the opioid and substance use crises. O-SUDDEn will fill several existing gaps in data infrastructure and prediction models that include machine learning, geospatial analyses, and community context. The specific aims for O-SUDDEn are as follows: Aim 1: Data linkage with establishment of O-SUDDEn. We will develop a geospatially-sensitive, individual-level secure data lake that integrates multiple disparate data sources that meets the requirement of a coded-limited set under HIPAA. New data sources include real-time individual-level longitudinal data from urine drug testing (UDT) and community contextual data based on the Ohio Opportunity Index, an area-level social determinants of health developed and used by our team. We will develop query and use tools for data harmonization and integration, prepare and release data sets for dissemination and secondary data analyses; and facilitate secondary use of related administrative data to generate evidence that informs targeted opioid interventions. Aim 2: Develop Predictive Models and Surveillance algorithms. Geospatial and machine learning will be used to model the contribution of opioid, cocaine, and stimulant use on OD, OD death and opioid use disorder/substance use disorder (OUD/SUD); temporal relationship of real time data including UDT and demographic and contextual variables to identify high- risk populations and subpopulations or geographic locales. We will validate model performances and predictive power and then disseminate surveillance and forecasting algorithms through the O-SUDDEn portal for end-user access. Aim 3: Deploy a human-centered platform and actional informatics. Applying co-design principles with a human-centered work group consisting of end-user stakeholders (community, researchers, and state policy makers/agencies), we will develop and deploy an end-user tailored portal and effective user-friendly tools to provide actionable insights to inform opioid treatment and targeted harm reduction strategies.

PROJECT NARRATIVE The Opioid and Substance Use Disorders Data Enclave project (O-SUDDEn) will provide a novel and transformative approach to support actionable insights into the opioid and substance use crisis. It will fill several gaps in existing data infrastructure through a secure data lake that will provision geospatially-sensitive, individual and community-level data integrated with multiple, disparate data sources including real-time individual-level longitudinal data from urine drug testing (UDT). The curated data will capture the evolving drug environment and community contextual data based on the Ohio Opportunity Index, an area-level social determinants of health index, with geo-spatial and machine learning methods to explore the temporal relationship of real time data including UDT and demographic and contextual variables to identify high-risk populations and subpopulations or geographic locales.