Systems Science Approaches to the Opioid Crisis: Exploring its Multifaceted Nature through Agent-Based Model Simulations
Date
2023-11-23
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
ORCID
Type
Thesis
Degree Level
Doctoral
Abstract
Although opioids prescribed for medicinal purposes have shown temporary pain relief, they can lead to severe physical and psychological effects, addiction, and even death due to misuse, abuse, addiction, and overdose. The euphoric effects of opioids can drive individuals to seek street opioids, leading to a cascade of consequences that extend beyond personal health and are accompanied by a negative societal stigma, which can impede efforts to overcome the vicious cycle.
The central features of the opioid crisis reflect its complex nature, which can be effectively understood through the application of systems science methods. Purpose-specific simulation models can be created to replicate the key characteristics of the opioid crisis and used to analyze the behavior of the system, identify potential unintended consequences of different policy options, and evaluate alternative strategies. This work contributes three agent-based models, each addressing a different facet of the opioid crisis.
The first model examines the impact of COVID-19-related school closures on nonmedical prescription opioid use among youth. Grounded in social impact theory, this model explores the dynamics that may influence opioid use following school closures. By combining opinion dynamics and acute withdrawal intensity, the model simulates youth decision-making regarding opioids use. It suggests that lifting school closures could significantly increase non-medical prescription opioid use among youth. Effective interventions targeting risk factors at home can help prevent increased youth opioid use after school closures.
The second model evaluates the effectiveness of prescription regimes utilizing machine learning monitoring programs in identifying patients at risk of opioid abuse during treatment. It incorporates a hidden Markov model into an agent-based simulation to classify patients' underlying states of prescription opioid use. A synthetic data experiment was conducted using the calibrated agent-based simulation model to generate time series data for feature selection. Lowering prescription doses yields favorable results in terms of overdose rates, escalation to street opioids, and prescription legitimacy, emphasizing the need for comprehensive evaluation of public health interventions.
The third model focuses on modified opioid agonist therapy (OAT) guidelines during the COVID-19 pandemic. It simulates individuals receiving OAT, including those with increased take-home doses. The model assesses the impact of increased take-home doses on treatment retention and opioid-related harms. Model findings suggest that increasing take-home doses could enhance treatment retention. However, the increased opioid-related harms among certain groups of patients receiving higher take-home doses of OAT underscores the importance of expanding naloxone availability within the networks of OAT patients.
At a methodological level, this dissertation demonstrates the integration of opinion dynamics theories, AI-based health policies, and hierarchical state-charts to enhance the utility of agent-based models in addressing public health issues. It highlights the models' ability to generate time series data for machine learning techniques and evaluate the long-term impacts of AI-based healthcare policies. Furthermore, the modular design patterns used in the models facilitate comprehensive policy assessment while retaining generality.
Description
Keywords
Systems Science, Opioid Crisis, Agent-based Modeling, Simulation Models
Citation
Degree
Doctor of Philosophy (Ph.D.)
Department
Computer Science
Program
Computer Science