Potentially Inappropriate Opioid Prescribing, Overdose, and Mortality in Massachusetts, 2011–2015

Abstract

Background: Potentially inappropriate prescribing (PIP) may contribute to opioid overdose. Objective: To examine the association between PIP and adverse events. Design: Cohort study. Participants: Three million seventy-eight thousand thirty-four individuals age ≥ 18, without disseminated cancer, who received prescription opioids between 2011 and 2015. Main Measures: We defined PIP as (a) morphine equivalent dose ≥ 100 mg/day in ≥ 3 months; (b) overlapping opioid and benzodiazepine prescriptions in ≥ 3 months; (c) ≥ 4 opioid prescribers in any quarter; (d) ≥ 4 opioid-dispensing pharmacies in any quarter; (e) cash purchase of prescription opioids on ≥ 3 occasions; and (f) receipt of opioids in 3 consecutive months without a documented pain diagnosis. We used Cox proportional hazards models to identify PIP practices associated with non-fatal opioid overdose, fatal opioid overdose, and all-cause mortality, controlling for covariates. Key Results: All six types of PIP were associated with higher adjusted hazard for all-cause mortality, four of six with non-fatal overdose, and five of six with fatal overdose. Lacking a documented pain diagnosis was associated with non-fatal overdose (adjusted hazard ratio [AHR] 2.21, 95% confidence interval [CI] 2.02–2.41), as was high-dose opioids (AHR 1.68, 95% CI 1.59–1.76). Co-prescription of benzodiazepines was associated with fatal overdose (AHR 4.23, 95% CI 3.85–4.65). High-dose opioids were associated with all-cause mortality (AHR 2.18, 95% CI 2.14–2.23), as was lacking a documented pain diagnosis (AHR 2.05, 95% CI 2.01–2.09). Compared to those who received opioids without PIP, the hazard for fatal opioid overdose with one, two, three, and ≥ four PIP subtypes were 4.24, 7.05, 10.28, and 12.99 (test of linear trend, p < 0.001). Conclusions: PIP was associated with higher hazard for all-cause mortality, fatal overdose, and non-fatal overdose. Our study implies the possibility of creating a risk score incorporating multiple PIP subtypes, which could be displayed to prescribers in real time.