The Prescription Paradox: A Multidimensional ARIMAX Ensemble Forecast of Opioid Mortality (2006–2034)
Faculty Mentor
Jing Kersey
Location
Russell Union Room 2052
Type of Research
On-going
Session Format
Oral Presentation
College
Jiann-Ping Hsu College of Public Health
Department
Biostatistics
Abstract
Public health policy currently faces a deadly paradox: while legal opioid dispensing rates have plummeted by nearly 50% since 2010, opioid-related mortality continues to surge. This research investigates this structural decoupling, challenging the outdated epidemiological assumption that reducing legal prescription volume will naturally resolve the overdose crisis.
To capture the complex, compounding drivers of this epidemic, this study constructed a multidimensional ARIMAX (Auto-Regressive Integrated Moving Average with Exogenous Regressors) ensemble model to forecast state-level opioid mortality through 2034. The methodology harmonized 19 years of historical data (2006–2024), integrating provisional mortality counts, CDC dispensing rates, and four critical Social Determinants of Health (SDOH): poverty, disability, transit access, and vehicle ownership. Crucially, a "Fentanyl Era" structural break variable was engineered to mathematically account for the supply-chain transition from diverted pharmaceuticals to highly toxic illicit synthetics. To measure the "echo effect" of social and supply-chain shifts, a temporal sensitivity analysis was conducted, testing 0-year, 1-year, and 2-year lags across an automated tournament of 54 distinct modeling scenarios.
The results demonstrate that models failing to account for the illicit fentanyl transition dangerously underestimate future mortality. The immediate-impact model (Lag 0) achieved the highest historical accuracy (RMSE: 58.3). Furthermore, while provisional 2024 data shows a temporary dip in deaths, the ensemble forecast predicts a "sticky baseline." The model projects that mortality will rebound and plateau near 2,176 annual deaths by 2034, driven by compounding social vulnerabilities and a toxic street supply. Ultimately, this study mathematically proves that the historical correlation between legal dispensing and overdose mortality is broken, signaling that intervention strategies must urgently pivot from prescription monitoring toward comprehensive harm reduction.
Program Description
.
Start Date
4-23-2026 1:45 PM
End Date
4-23-2026 2:00 PM
Recommended Citation
Rabari, Parthkumar Rameshbhai, "The Prescription Paradox: A Multidimensional ARIMAX Ensemble Forecast of Opioid Mortality (2006–2034)" (2026). GS4 Student Scholars Symposium. 135.
https://digitalcommons.georgiasouthern.edu/research_symposium/2026/2026/135
The Prescription Paradox: A Multidimensional ARIMAX Ensemble Forecast of Opioid Mortality (2006–2034)
Russell Union Room 2052
Public health policy currently faces a deadly paradox: while legal opioid dispensing rates have plummeted by nearly 50% since 2010, opioid-related mortality continues to surge. This research investigates this structural decoupling, challenging the outdated epidemiological assumption that reducing legal prescription volume will naturally resolve the overdose crisis.
To capture the complex, compounding drivers of this epidemic, this study constructed a multidimensional ARIMAX (Auto-Regressive Integrated Moving Average with Exogenous Regressors) ensemble model to forecast state-level opioid mortality through 2034. The methodology harmonized 19 years of historical data (2006–2024), integrating provisional mortality counts, CDC dispensing rates, and four critical Social Determinants of Health (SDOH): poverty, disability, transit access, and vehicle ownership. Crucially, a "Fentanyl Era" structural break variable was engineered to mathematically account for the supply-chain transition from diverted pharmaceuticals to highly toxic illicit synthetics. To measure the "echo effect" of social and supply-chain shifts, a temporal sensitivity analysis was conducted, testing 0-year, 1-year, and 2-year lags across an automated tournament of 54 distinct modeling scenarios.
The results demonstrate that models failing to account for the illicit fentanyl transition dangerously underestimate future mortality. The immediate-impact model (Lag 0) achieved the highest historical accuracy (RMSE: 58.3). Furthermore, while provisional 2024 data shows a temporary dip in deaths, the ensemble forecast predicts a "sticky baseline." The model projects that mortality will rebound and plateau near 2,176 annual deaths by 2034, driven by compounding social vulnerabilities and a toxic street supply. Ultimately, this study mathematically proves that the historical correlation between legal dispensing and overdose mortality is broken, signaling that intervention strategies must urgently pivot from prescription monitoring toward comprehensive harm reduction.
