Barnett and his colleagues demonstrated that opioid-naive patients who presented to an emergency department and were treated by a high opioid prescriber were more likely to become long term opioid users than those who were treated by a low opioid prescriber.
The purpose of this study was to determine whether one opioid prescription can initiate long term opioid use. This is a difficult question to answer because though plenty of anecdotal evidence suggests yes, addiction has a complex genesis and the vast majority of patients who receive an opioid prescription for acute pain have no serious consequences. However, because developing opioid addiction is often a life-ruining event, and because we count opioid prescriptions in the hundreds of millions, if even a small fraction of patients started on opioids develop addiction as a result of the prescription, the harms to those affected outweigh the analgesic benefit offered to everyone else.
How do you correlate a single variable with an uncommon but extremely harmful, multifactorial adverse event? Prospectively randomizing 100,000 patients is not possible, so we’re confined to retrospective analysis. There are a number of researchers who have done just that [1 2 3 4 5 6 7 8 9 10], and though most of these studies came to a similar conclusion, their methodologies are comparatively weak. For example, linking a single opioid prescription to long-term use suffers from the confounder that patients who receive an opioid prescription are likely in more pain than patients who don’t, so it’s not surprising that they are more likely to go on to recurrent opioid use.
In the absence of a controlled study design, scientists look for a naturally occurring randomizing event, which, if you’re smart enough to identify and analyze it, is an experiment performed accidentally. This is exactly what happens when a patient shows up for emergency care: she is assigned to a provider randomly. Barnett and his group (none of whom are emergency physicians) brilliantly exploited this physician lottery by pairing it with the hugely variant opioid prescribing practice of 27,772 physicians in their sample. Thus the 215,678 patients seen by the highest-prescribing quartile of emergency docs differ from the 151,951 patients seen by the lowest-prescribing quartile of emergency docs only in that the former are more likely to be discharged with an opioid script. It’s as if 377,629 patients were randomized. Because, in effect, they were.
When you randomize 377,629 patients, you can identify small treatment effects, and they did. A patient who sees a high prescriber is slightly (0.35%) more likely to be using opioids one year from now. 0.35% is a small number, and this is concordant with our experience: most folks who get their first opioid script don’t run into trouble. But this study compellingly suggests that some small number–Barnett says 1 in 48, but maybe it’s 1 in 148 or maybe even 1 in 481–are set down the path to opioid misuse from a prescription for pain. This paper is not about implicating emergency medicine, as it has been spun, or even about judging high prescribers vs. low prescribers. As Dr. Barnett said in my correspondence with him, “Fundamentally our paper is about the concept that even one opioid prescription to a naive patient can be associated with long-term use.”
Apart from the strength of the correlation between one opioid prescription and long-term use, it’s hard to imagine that more than 1 in 7 patients discharged from the emergency department should walk out with vicodin, as was found in this study. If you’re looking for resources to help you prescribe more judiciously, readers can start here and listeners/watchers can start here.