Spatial patterns of rural opioid-related hospital emergency department visits: A machine learning analysis

Abstract

As opioid-related overdose emergency department visits continue to rise in the United States, there is a need to understand the location and magnitude of the crisis, especially in at-risk rural areas. We analyzed sets of ZIP code level electronic health records for emergency department visits from 6 hospitals for two rural regions of Maryland with higher opioid-related overdose rates. Analysis of the demographics of visits found Black or African American emergency department visits in both rural regions were higher than the proportion of their population per region. We trained random forest models with socio-demographic factors and health risk factors on the visits data to understand drivers and risk factors for opioid misuse. The models ranked factors relating to opioid prescribing rates, race, housing, and poor mental health as highly important. Factors associated with opioid-related overdose emergency department visits were found to vary by race, gender, and location and may provide useful insights for designing mitigation initiatives.