Objective To report trends and characteristics of long‐term opioid use for non‐cancer pain. Methods CONSORT (CONsortium to Study Opioid Risks and Trends) includes adult enrollees of two health plans serving over 1 per cent of the US population. Using automated data, we constructed episodes of opioid use between 1997 and 2005. We estimated age–sex standardized rates of opioid use episodes beginning in each year (incident) and on‐going in each year (prevalent), and the per cent change in rates annualized (PCA) over the 9‐year period. Long‐term episodes were defined as > 90 days with 120+ days supply or 10+ opioid prescriptions in a given year. Results Over the study period, incident long‐term use increased from 8.5 to 12.1 per 1000 at Group Health (GH) (6.0% PCA), and 6.3 to 8.6 per 1000 at Kaiser Permanente of Northern California (KPNC) (5.5% PCA). Prevalent long‐term use doubled from 23.9 to 46.8 per 1000 at GH (8.5% PCA), and 21.5 to 39.2 per 1000 at KPNC (8.1% PCA). Non‐Schedule II opioids were the most commonly used opioid among patients engaged in long‐term opioid therapy, particularly at KPNC. Long‐term use of Schedule II opioids also increased substantially at both health plans. Among prevalent long‐term users in 2005, 28.6% at GH and 30.2% at KPNC were also regular users of sedative hypnotics. Conclusion Long‐term opioid therapy for non‐cancer pain is increasingly prevalent, but the benefits and risks associated with such therapy are inadequately understood. Concurrent use of opioids and sedative‐hypnotics was unexpectedly common and deserves further study. Copyright © 2009 John Wiley & Sons, Ltd.