Application of the ARIMA Model to Forecast Future Overdose Deaths

Current data for opioid policy in affected states is outdated by 1-2 years, highlighting the necessity for a predictive model to forecast fatal drug overdoses and aid in the implementation of evidence-based preventative policies. We utilized Auto-ARIMA, an automated version of the ARIMA linear regression model, to forecast overdose deaths in the U.S., each state, and Washington D.C. Our forecasts for 2021 and 2022 were based on data from 2015 to 2021, sourced from the CDC's Vital Statistics Rapid Release program. Our ARIMA model forecasts an increase in overdose deaths across the U.S. and in many states, with some states potentially seeing a doubling of deaths from 2020 to 2022. This research represents an initial step in using real-time forecasts to evaluate the effectiveness of state-level public health policies in combating the opioid crisis.