Objectives/Goals: Opioid analgesic misuse poses a critical public health problem in the United States. The 2015 National Survey on Drug Use and Health (SAMHSA) 33,091 people died due to opioid overdosing and 12.5 million people misused prescription opioids in 2015. In the state of Utah alone, there are around 24 prescription opioid deaths every month constituting half of the poisoning deaths in the state. Yet, there is a lack of understanding of the role of the underlying treatment, prescribing and dispensing systems and processes in these areas. There are multiple factors involved in misuse including those related to the patient, physician and pharmacy; knowledge guiding opioid prescribing; and population-level factors such urban, rural and frontier regions of residence of patients, their socioeconomic status, unemployment, a culture of utilizing prescription pain medications, and the scarcity of illegal drugs such as heroin and cocaine. Discovering patterns in misuse trajectories therefore requires flexible approaches to integrate their heterogeneous data and support knowledge discovery. Methods/Study Population: In order to understand the role of multiple factors involved, we are developing a framework for graph-based methods for discovering opioid misuse trajectories. In this approach, we elicited key concepts relevant to understand opioid misuse from literature review, domain experts and existing data sources. We then used data modeling methods to represent the data in sharable formats. Results/Anticipated Results: At a high level, these concepts are categorized into three domains and are represented as three hypergraphs: (i) patient-provider-pharmacy, (ii) knowledge guiding opioid prescribing, and (iii) population level information related to opioid misuse. We developed a logical model for representing these opioid misuse related metadata. We then transform and store various Utah opioid misuse related data in graphical data stores to support flexible methods of data integration and knowledge discovery over geographic and temporal dimensions using Big Data methods such as graph clustering, probabilistic graphical models and other machine learning methods for learning spatio-temporal trajectories of opioid misuse. Discussion/Significance of Impact: Graph-based methods provide a holistic and flexible approach to integrate diverse heterogeneous data associated with opioid misuse and supports different methods of knowledge discovery. We will our approach will provide generalizable and reproducible methods for understanding trajectories of opioid misuse. This knowledge discovery framework could support the next generation of translational research of opioid misuse by providing personal risk scores for managing patients in the clinic, population level strategies for reducing misuse, stratifying poison control center resources, and hypotheses for clinical research.

This work is partially supported by the Utah Center for Clinical and Translational Science funded by NCATS award UL1TR001067 and Utah PRISMS Informatics Center through NIH/NIBIB U54EB021973