No one can deny the severity of the opioid crisis in which we currently reside. But what is less clear is whether we are innocent bystanders, facing the consequences of someone else’s problem, or directly responsible for causing a portion of the addiction we are currently witnessing.
In a recent publication in the NEJM, Barnett et al attempted to define the role of the Emergency Physician in the current crisis. The authors performed a retrospective analysis of a random sample of 20% of Medicare beneficiaries who had an index Emergency Department visit from 2008 through 2011. Patients could not have received prescription opioids within the 6-months prior to this initial visit. Barnett et al limited their analyses to beneficiaries who had been continuously enrolled in Medicare Part D for at least 6 months prior to the index visit, to 12 months afterward. Beneficiaries with hospice claims or a cancer diagnosis between 2008 and 2012 were excluded from the analysis.
The authors identified Emergency Physicians, defined as physicians with an Emergency Medicine board specialty who billed 90% or more of claims with an Emergency Department place of service. Index Emergency Department visits were assigned to an individual physician and hospital. Physicians were stratified into high-intensity prescribers and low-intensity prescribers depending on the percent of patients visits in which they prescribed an opioid analgesic at discharge. The authors compared the rate of chronic opioid use over the following 12-months in patients who were seen by low-frequency and high-frequency prescribers.
The authors examined 377,629 patients, 215,678 which were treated by low-intensity opioid prescribers and 161,951who were prescribed by high-intensity opioid prescribers. The average rates of opioid prescriptions varied wildly between the groups (7.3% and 24.1% respectively).
The authors found that long term opioid use was higher in the patients who were seen by the high-intensity providers when compared to those seen by a low- intensity provider (1.51% and 1.16%, respectively). This 0.35% absolute difference remained statistically significant after the authors adjusted for potential confounders including age, sex, race, ethnic group, Medicaid eligibility, and disability status, and 11 different chronic conditions. With the help of some statistical wizardry, the authors calculate a number-needed-harm of 48 opioid prescriptions to create one additional long term-opioid user.
And while these authors went through impressive methodological hurdles to limit the potential biases that are common to a retrospective data-dredge of an administrative data-base, this is still a retrospective data-dredge of an administrative data-base. And despite its robust sample size and impressively low p-values, these shortcomings limit the conclusions that can be drawn from such a paper.
The concern voiced by some, is if Emergency Physicians prescribe a short course of opioid pain medication, we will place our patients at risk for the downstream harms of addiction and dependency. But this is not what this study found. At best Barnett et al demonstrated an association between patients seen by emergency providers who more frequently prescribe opioid pain medications at discharge, and an ever so slight increase in the use of long-term opioid pain medication over the following 12-months. This says nothing about addiction. In fact, it is only because of our negative connotation of opioid use, that we view this increase rate in long-term use as a negative outcome. While the authors’ proposed NNH of 48 is academically curious, it is frankly an inappropriate use of such an analysis. First it assumes a level of granularity to the data, that is just not present in a retrospective registry analysis such as this. More importantly, it is misleading, as it suggests a causative nature to a relationship where only an association has been demonstrated.
If anything this study highlights the safety of a short course of opioid pain medication. In a sample size of 377,629 patients the overall rate of long-term use was minimal (1.16% and 1.51% in the respective groups). And the increase in the rate of long-term use in patients seen by high-intensity providers was only 0.35%. Additionally, they were unable to demonstrate any clinically important signal of harm in the patients who were seen by high-intensity providers. The rates of opioid related hospital encounters over the following 12-months were 9.73% and 9.96% in the low-intensity and high-intensity groups, respectively. The rates of opioid poisoning were 0.07% and 0.10%. While many of these findings were statistically significant, the majority of the differences are clinically meaningless. In fact, the use of the p-value in this study is almost laughable, the sample size being so large, almost every measurement the authors report boasts a p-value less than 0.05.
As an explanation of the observed results, the authors suggest “therapeutic inertia”, or the concept that because of this initial prescription from the Emergency Department, the primary care providers were far more reticent about discontinuing an opioid pain medication once it was initiated. And while this hypothesis may explain these results, a far simpler explanation is equally likely. That the high-intensity providers more often prescribed opioid pain medication because the patients required them. And some of these patients went on to prescribe long-term opioid prescription because the very same patients required them. As with all non-randomized data, it is impossible to determine the vector of the line that bridges the data points in question.
This post should not be viewed as a denial of the current opioid crisis. Nor is it permission for a more liberal prescribing strategy. Rather it is a simple reminder that the evidence presented to us is only as strong as its methodological constructs. A short course of opioid pain medication may very well lead to chronic long term use. Opioid prescriptions from Emergency Physicians may very well lead a small subset of patients to addiction and dependency. But the data presented by Barnett et al does not prove this relationship. I suspect the truth is far more complex, and likely cannot be defined using tools as blunt as administrative registry data.
Sources Cited:
- Barnett ML, Olenski AR, Jena AB. Opioid-Prescribing Patterns of Emergency Physicians and Risk of Long-Term Use. N Engl J Med. 2017;376(7):663-673.
- EM Nerd-The Case of the Partial Cohort - May 24, 2020
- EM Nerd: The Case of the Sour Remedy Continues - January 20, 2020
- EM Nerd-The Case of the Adjacent Contradictions - December 23, 2019
It is concerning to me the defensive tone of this post. I agree that there will be many limitations on a retrospective study of a large database, but taken in the context of a large amount of research coming out on the opioid epidemic, how can we be defensive about this data. If anything, as emergency physicians we should be concerned how our prescribing is affecting the long term use of opioids in our population. Particularly because the NEJM is discussing prescribing in the elderly, a high risk population to use opioids in. Here are some articles that support here… Read more »
Hi Chris, Thanks for writing. I am sorry you interpreted my post as “defensive”. Not my intention. I, in fact, have clinical equipoise regarding this topic. I didn’t at any point belittle or lessen the severity of the opioid crisis. The point of this post is whatever preconceived notions you may have regarding the safety of opioid prescriptions from the Emergency Department, this study did not demonstrate that short courses of opioids prescribed from the Emergency Department lead to an increased rate of opioid abuse. Data is data, especially given the “post-truth” era in which we currently exist, it is… Read more »
Hi Rory, I appreciate your criticism of the data and I agree with you that it is important and healthy in the scientific/medical community. As someone who works in both EM and has been closely studying the data on the opioid epidemic for many years, I think we need to put a study like this into the greater context of what we already know. I think this study supports two important concepts. First, there continues to be variation in opioid prescribing, up to 3 fold in this study, between EM providers in an elderly population. It would be nice to… Read more »
I agree with Chris. I also think labeling this study a “data-dredge” is unfair. Many health services researchers (including EM researchers) have used administrative datasets such as Medicare to evaluate important clinical and policy topics. Observational studies using administrative data have their limitations (as do all study designs/data sets) and I think Barnett et al were pretty transparent about the limitations of this particular study. A nice blog post by a colleague about correlation/causation in observational studies that I found helpful can be found here:
https://blogs.sph.harvard.edu/ashish-jha/
Hi Rory, While i agree that data mining can lead to spurious P values, their analysis accounts for as much of the variation as possible and is able to still show an association which i think actually speaks to how big the issue of opiod over prescribing is. Your conclusion “That the high-intensity providers more often prescribed opioid pain medication because the patients required them” is precisely what the study refuted and what we all keep telling ourselves. The exposure was the prescribing physician not the patient and hence to suggest that the patients somehow had higher pain needs doesnt… Read more »
Thanks for writing Sameer. My conclusion was not that “the high-intensity providers more often prescribed opioid pain medication because the patients required them”. My conclusion that because of the design of the study we are not able to determine the cause of the increase in chronic opioid use. I agree Barnett et al did a fantastic job at controlling for as much bias as possible but it is impossible to control for everything given the design of their study. Additionally the difference in chronic opioid use was minimal, with no additional harms identified. Is an absolute increase of 0.35% in… Read more »
New Data you should review: https://www.cdc.gov/mmwr/volumes/66/wr/mm6610a1.htm?s_cid=mm6610a1_e
Thanks Chris, I’ll take a look