Abstract
BACKGROUND: The opioid crisis in the United States is a complex issue with interconnected factors that lead to opioid misuse and opioid-involved mortality. This study assessed the relative importance of different risk factor domains in predicting fatal opioid-involved mortality that occurred after hospital encounters involving opioids. METHODS: A machine learning model was developed by integrating multiple data sources, including hospital records, death records, and societal data. The model allowed simultaneous examination of risk factors across individual drug and non-drug related factors, hospital factors, and societal factors. RESULTS: 429,005 patients with opioid-related encounters in 2014 were assessed, where 56.6% were female and the mean age was 44.98. Among deaths that had specific drugs listed for both the hospital encounter and the death, 51.7% of hospital encounters progressed to a more potent opioid at death. Community factors cumulatively had similar importance as individual drug-related factors in predicting opioid-involved deaths and were relatively more important in predicting opioid-involved mortality compared to non-drug involved mortality. In predicting opioid-involved mortality, non-drug related individual-level predictors accounted for 45.1% of the importance. Community factors accounted for 27.9% of the importance and drug-related individual factors accounted for 22.5%. In contrast, community factors accounted for only 16.5% of the importance when predicting non-opioid-involved mortality. PRACTICE IMPLICATIONS: Rather than suggesting community factors outweigh individual factors, our results highlight individual vulnerability may be amplified or mitigated by broader environmental factors. Interventions targeting larger social determinants of health may be strongly influential in reducing drug-involved mortality. This study demonstrated a quantitative evaluation of the different domains of risk factors and highlighted the importance of considering societal and community factors in a holistic approach to preventing opioid-involved mortality.