An interrupted time series analysis of the Campaign to Reduce Opioid Prescribing primary care feedback intervention.
Problem
There is international concern over rising trends in opioid prescribing, largely attributed to prescribing for chronic non-cancer pain. Given accumulating evidence of harm, reversing the current trend in opioid prescribing would benefit a substantial at-risk population.The Campaign to Reduce Opioid Prescribing (CROP) intervention entailed sending 316 practices in West Yorkshire evidence-based, comparative and practice-individualised feedback on their overall prescribing and for patient groups at risk of long-term or strong opioid prescribing, bimonthly for 12 months from April 2016.We used population-level routine data to assess the effectiveness of an enhanced performance feedback intervention on opioid prescribing for non-cancer pain in general practice.
Approach
We assessed the change in the rate of opioid prescribing using a controlled interrupted time series (ITS) analysis. Routinely collected data were used to assess the trends in opioid prescribing across all 311 intervention practices and compared to 135 control practices. Data were collected for the four years prior to, the year during, and the year after the intervention using monthly epochs.We examined effects on different types of opioid prescribing and particular at-risk patient groups, and whether practice or patient characteristics are associated with any intervention effects.Trends in prescribing nationally at CCG level were compared for those CCGs targeted by CROP and those who were not to see if the underlying national trend was changed by the intervention.
Findings
In 2013, the opioid prescribing rate did not significantly differ between control and intervention practices (34.1 and 34.3 per 1000 patients per month respectively) and increased at a similar rate in both groups up until the intervention occurred.There was no statistically significant intervention effect during the first month of the intervention, nor a statistically significant reduction in the trend compared with that of controls of 0.03 prescription per 1,000 patients per month (95% CI -0.13, 0.08). However, monthly prescription rates fell significantly in intervention practices during the post intervention period by 0.05 prescriptions per 1,000 patients (95%CI -0.10, -0.01), whilst there was no significant change in control practices monthly prescription rates over the same period (-0.007; 95%CI -0.06, 0.05). Analysis of whether practice or patient characteristics are associated with any intervention effects are in progress.
Consequences
CROP represents an effective, population-level intervention to reduce opioid prescribing. The time lag for achieving effects is consistent with the notion that repeated feedback resulted in accumulating changes in clinical behaviour. Locally, CROP reversed the current rising national trend in opioid prescribing. This low cost intervention translates into a substantial population impact and reduction in patient harm. We are not aware of any other intervention that has achieved this at scale in UK primary care.