Key PointsQuestion
Are interventions to reduce pain-related opioid prescribing for patients who are discharged from the emergency department associated with variation in opioid prescription rate and/or prescribed opioid quantity?
Findings
In this meta-analysis of 63 unique studies, 34 of 51 studies assessing prescribing rates (67%) reported a significant reduction after intervention implementation, and 17 of the 39 studies assessing prescribed opioid quantity (44%) reported a significant intervention-related reduction. There were no data on patient-centered outcomes such as pain relief.
Meaning
This analysis found that specific interventions may be associated with reducing the opioid prescription rate; however, novel interventions are needed to reduce the quantity of opioids per prescription by emergency department physicians while evaluating their associations with patient-centered outcomes.
Importance
Limiting opioid overprescribing in the emergency department (ED) may be associated with decreases in diversion and misuse.
Objective
To review and analyze interventions designed to reduce the rate of opioid prescriptions or the quantity prescribed for pain in adults discharged from the ED.
Data Sources
MEDLINE, Embase, CINAHL, PsycINFO, and Cochrane Controlled Register of Trials databases and the gray literature were searched from inception to May 15, 2020, with an updated search performed March 6, 2021.
Study Selection
Intervention studies aimed at reducing opioid prescribing at ED discharge were first screened using titles and abstracts. The full text of the remaining citations was then evaluated against inclusion and exclusion criteria by 2 independent reviewers.
Data Extraction and Synthesis
Data were extracted independently by 2 reviewers who also assessed the risk of bias. Authors were contacted for missing data. The main meta-analysis was accompanied by intervention category subgroup analyses. All meta-analyses used random-effects models, and heterogeneity was quantified using I2 values.
Main Outcomes and Measures
The primary outcome was the variation in opioid prescription rate and/or prescribed quantity associated with the interventions. Effect sizes were computed separately for interrupted time series (ITS) studies.
Results
Sixty-three unique studies were included in the review, and 45 studies had sufficient data to be included in the meta-analysis. A statistically significant reduction in the opioid prescription rate was observed for both ITS (6-month step change, −22.61%; 95% CI, −30.70% to −14.52%) and other (odds ratio, 0.56; 95% CI, 0.45-0.70) study designs. No statistically significant reduction in prescribed opioid quantities was observed for ITS studies (6-month step change, −8.64%; 95% CI, −17.48% to 0.20%), but a small, statistically significant reduction was observed for other study designs (standardized mean difference, −0.30; 95% CI, −0.51 to −0.09). For ITS studies, education, policies, and guideline interventions (6-month step change, −33.31%; 95% CI, −39.67% to −26.94%) were better at reducing the opioid prescription rate compared with prescription drug monitoring programs and laws (6-month step change, −11.18%; 95% CI, −22.34% to −0.03%). Most intervention categories did not reduce prescribed opioid quantities. Insufficient data were available on patient-centered outcomes such as pain relief or patients’ satisfaction.
Conclusions and Relevance
This systematic review and meta-analysis found that most interventions reduced the opioid prescription rate but not the prescribed opioid quantity for ED-discharged patients. More studies on patient-centered outcomes and using novel approaches to reduce the opioid quantity per prescription are needed.
Trial Registration
PROSPERO Identifier: CRD42020187251
Rates of prescription opioid–related deaths have remained very high during the last 10 years in the US.1 Despite this ongoing crisis, opioids are still frequently prescribed for home pain management by physicians2-4 because some patients may need them for pain control when alternatives are not sufficient.5 Although a decrease in opioid prescribing has been observed in recent years from emergency department (ED) physicians,6-8 it can still be optimized because variability in opioid prescribing among ED physicians remains substantial, and overprescribing continues to be frequent.9-13 Opioid prescription to patients with acute pain after an ED visit can lead to long-term use14,15 and opioid use disorders.16 Furthermore, it has been shown that as much as 68% of the initial opioid prescription after an ED visit for acute pain is left unused.17,18 These extra pills can be diverted, leading to opioid misuse, dependence, and overdose in our communities.19-21
Limiting overprescribing could have a substantial effect on opioid diversion and misuse and therefore on the opioid crisis.22,23 Several opioid reduction strategies have been proposed by medical institutions, cities, states, and other stakeholders. These approaches vary extensively, ranging from prescription drug monitoring programs,24-27 new laws,28-30 policies, guidelines,31-34 prescriber education initiatives,35,36 or changing the default quantity of opioids in electronic medical record prescription orders.37-40 The efficacy of these interventions on the opioid prescribing rate or quantity at ED discharge remains uncertain and, to our knowledge, has never been examined systematically. The identification of approaches associated with greater efficacy could help policy makers elaborate more targeted programs to prevent opioid misuse and deaths. Our main objective was to review and analyze the evidence regarding interventions to reduce the opioid prescribing rate or quantity for treating pain in adults discharged from the ED.
This meta-analysis was registered before its initiation (PROSPERO identifier: CRD42020187251). The results are presented as per the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) reporting guideline (supporting checklist/diagram).41 We also followed the Synthesis Without Meta-analysis reporting rules42 to complement the PRISMA guideline.
We included all intervention studies designed to reduce the opioid prescription rate and/or the quantity of opioids per prescription given to adults discharged from the ED (≥18 years of age) for home pain management. Except for case reports and case series, all types of study designs were included.
Studies that exclusively included pediatric patients or populations with substance use disorder or only evaluated opioids given during the ED stay were excluded. Studies performed in settings other than the ED, that were without interventions, that pertained to opioid use unrelated to pain, or that did not report opioid prescribing rate or quantity were also excluded.
Data Sources and Search Strategy
In collaboration with an information specialist, we developed a search strategy based on the intersection of 3 search themes: opioid, emergency, and prescription (detailed search strategies are available in eMethods 1 in the Supplement). The following databases were searched from their inception to May 15, 2020: MEDLINE, Embase, CINAHL, PsycINFO, and Cochrane Controlled Register of Trials. An updated search was performed at the end of the process on March 6, 2021, to collect recently published reports. We also searched ClinicalTrials.gov and the International Standard Randomized Controlled Trial Number registry for ongoing studies. Gray literature was searched using Google Scholar, and the first 300 hits were screened to identify any relevant studies. References from studies meeting inclusion criteria were examined to identify additional relevant studies. There were no exclusions based on language.
The identified references from all databases were transferred to Covidence systematic review manager (Covidence systematic review software, Veritas Health Innovation [https://www.covidence.org]). Six reviewers (R.D., J.P., M.M., J.-M.C., D.W., and A.C.) participated in a 2-stage selection process of eligible studies. After duplicate removal, each citation was screened by 2 independent reviewers using the study title and abstract. Finally, 2 independent reviewers evaluated the full text of the remaining citations against inclusion and exclusion criteria. During both stages, a third reviewer (R.D. or A.C.) resolved discrepancies.
The data for all pertinent variables were extracted independently by 2 reviewers (J.P. and M.M.) using a standardized electronic Excel sheet (Microsoft Corporation), and the conflicts were resolved by consensus from the 2 reviewers. Study authors were contacted when outcomes were incomplete or when adult patients in the ED were mixed in with other populations. When the type of pain (acute or chronic) or the problem or diagnosis was not explicitly specified in the studies, we attributed both for type of pain and all for problem and diagnosis. Also, when time series data could be extracted from the study, we considered it as an interrupted time series (ITS) study; otherwise, we considered it a preintervention-postintervention design, even if the authors reported their study as an ITS.
The Cochrane Effective Practice and Organisation of Care (EPOC) Taxonomy of Implementation Strategies43 standard was used to categorize the different interventions into changes in health care organizations, in the clinician’s behaviors, and in the patient’s use of health services. A consensus was obtained across 6 reviewers (R.D., J.P., M.M., J.-M.C., D.W., and A.C.) to categorize the interventions as (1) education, policy, or guidelines (EPG); (2) prescription drug monitoring program or state law (PDMP); (3) clinician peer comparison (CPC); (4) electronic medical record quantity changes (EMR-QC); and (5) physical therapy (PT) (this category was originally planned as “other” but was finally composed of only physical therapy studies). Clinician peer comparison was categorized separately from EPG because of the additional motivational incentive induced by peer comparisons. A more detailed description of each intervention and the manner and rationale for how they were regrouped are presented in eMethods 2 in the Supplement. Each study was classified into 1 of the 5 intervention categories by 2 independent reviewers (R.D. and J.P.), and discrepancies were resolved by a third reviewer (A.C.). Studies using more than 1 intervention category were classified into the predominant category by reviewer consensus.
Primary and Secondary Outcomes
Our primary outcome was the variation in the opioid prescription rate and/or quantity generated by the intervention. Our secondary outcomes were the patients’ level of pain relief, patients’ satisfaction with their opioid prescription, and percentage of patients requiring additional opioid prescriptions.
Quality Assessment and Risk of Bias
The quality assessment of all retained articles was performed by 2 independent reviewers (J.P. and M.M.); conflicts were resolved either by consensus or by a third reviewer (R.D.). The risk of bias for preintervention-postintervention and ITS study designs were evaluated using the Risk of Bias in Nonrandomised Studies of Interventions (ROBINS-I) tool.44 The risk of bias for the cohort studies and randomized controlled trials were evaluated using the EPOC risk of bias tool.45 Abstracts were automatically considered at critical (ROBINS-I) or high (EPOC) risk of bias.
Data Synthesis and Analysis
Descriptive statistics were performed for intervention categories, study designs, country of origin, and type of pain (acute or chronic). For each study, we presented the absolute reduction in the opioid prescription rates and/or the absolute change in total amount of opioids per prescription. Rates were reported in percentage of discharged adults who were prescribed opioids. Quantities of opioids per prescription were presented as median (IQR) or mean (SD) number of pills or in total milligram morphine equivalent or morphine equivalent daily dosage. We also calculated the proportion of studies that showed a trend or a statistically significant (2-sided P < .05) reduction of opioid prescribing rate or quantity.
If the rate or quantity of opioid prescribed was reported in at least 3 studies, the results were pooled and included separately in a meta-analysis for each type of outcome and each type of study design (ITS vs other). For preintervention-postintervention, cohort, and randomized clinical trial (RCT) study designs, the opioid prescription rate was expressed as an odds ratio (OR). The number of events (opioid prescriptions) over the number of patients discharged from the ED for preintervention/control and postintervention/treatment groups was used to compute the OR. Opioid quantities per prescription were expressed as standardized mean difference (SMD). When means and SDs were not available, they were estimated based on medians or other statistics using the methods developed by Wan et al.46 Studies without sufficient data for these analyses were only described.
The ITS studies were analyzed according to Cochrane (EPOC) recommendations.47 Because ITS data were not examined uniformly across studies, we reanalyzed all available ITS study data using the same method, as recommended by Ramsay et al,48,49 and calculated the 6-month step change. Details of our approach to standardize ITS study results are available in eFigure 1 in the Supplement.
The rate and quantity of prescribed opioids are described according to the 5 intervention categories defined previously. All results are reported with 95% CIs. Because of the different natures of ITS effect size compared with the OR and SMD, these forest plots are presented separately. Heterogeneity was assessed statistically both overall and for each intervention category using I2 values. The τ2 and Cochran Q tests for heterogeneity are also reported. A χ2 test was used to determine whether there was a difference between intervention categories. When more than 2 intervention categories were statistically significant, pairwise subgroup χ2 tests were performed. All analyses were conducted using an inverse variance weighting method and a random-effects model, even if the I2 value was low, owing to the diversity of interventions.50
For each analysis of more than 10 studies, a funnel plot and Egger test51 were used to assess small samples publication bias.52 In 2 sets of sensitivity analyses, we used the 1-year step change for ITS studies when available and excluded studies at high risk of bias to evaluate differences in effect size. Segmented time-series regression analyses were performed using SPSS, version 26 (IBM Corporation); forest plots were executed using Revman, version 5.4 (The Nordic Cochrane Centre, The Cochrane Collaboration, 2014); and funnel plots and Egger tests were performed using Comprehensive Meta-Analysis program.53
Search Results and Study Characteristics
The initial search strategy generated 10 171 references after duplicate removal. Of these, 180 were kept for full-text review, and 63 unique studies were included in the review (Figure 1).12,27-32,35-37,39,40,54-104 Fifteen of 32 authors responded to our requests for more details on outcomes or population, which improved data availability for 10 studies.12,27,31,36,55,57,67,81,84,86 Most included studies were conducted in the US (55 [87%]),12,27-32,37,39,54-67,71-93,95-101,103,104 5 were from Australia,35,36,40,68,70 and 3 were from Canada.69,94,102 All studies were published within the last 10 years (2013 to 2021). Studies used mainly preintervention-postintervention (n = 39)12,28,30-32,35,39,40,58-70,75-85,92-94,99-102 and ITS (n = 21)27,29,36,37,54-57,71-74,87-91,95-98 designs; only 1 was an RCT,86 and 2 were cohort studies.103,104 The EPG intervention was used to help reduce opioid prescriptions in 21 studies31,32,35,36,54-70; PDMP, in 19 studies27-30,71-85; EMR-QC, in 11 studies37,39,40,95-102; CPC, in 10 studies12,86-94; and PT, in 2 studies.103,104 Fifty-two studies12,27-29,31,35-37,39,40,54,55,57,59-62,65,66,68-81,83-93,95-102 included a mix of pain problems, and 56 included both acute and chronic pain.12,27-32,35-37,39,40,54-57,60-62,65,66,68-81,84-104 The opioid prescription rate was reported for 25 studies,28,29,32,54,56,58-60,62,63,66,67,69,73-76,78-80,87,89,91,94,104 the prescribed opioid quantity for 13 studies,35,37,39,40,68,70,83,93,95-98,102 and both for 25 studies12,27,30,31,36,55,57,61,64,65,71,72,77,81,82,84-86,88,90,92,99-101,103 (Table 1).
Ten of 21 ITS studies27,36,54-57,88-91 demonstrated an overall moderate risk of bias (ROBINS-I); all other ITS and preintervention-postintervention studies12,28-32,35,37,39,40,58-85,87,92-102 were at serious or critical risk (eFigure 2 in the Supplement). The only included RCT86 was at low overall risk of bias, and both cohort studies103,104 were at high risk of bias (EPOC) (eFigure 3 in the Supplement).
Primary Descriptive Results
The absolute reduction of the prescribed opioid rate and quantity for included studies is presented in Table 2. Two studies37,54 reported and analyzed separately the results from their 2 sites (sites A and B). Of 51 studies assessing prescribing rates, 46 (90%) reported a reduction after intervention implementation (34 were statistically significant).12,27-29,31,32,54-59,62-67,69,76-80,82,84-86,88,90,94,100,101 Of 39 studies assessing prescribed opioid quantity, 32 (82%) reported an intervention-related reduction (17 were statistically significant).35,36,39,40,55,65,68,70,82-86,95,98,100,101
Primary Meta-Analysis: Interventions to Reduce Rate and/or Prescribed Opioid Quantity
Forty-five of 63 studies could be included in the meta-analysis for 1 or both components of our main outcome: 36 (80%) studies for the prescription rate11,26-31,35,53-58,62-66,70-74,78,81,83,85-90,100,102,103 and 23 (51%) for the prescribed opioid quantity11,26,29,30,35,36,38,39,54,56,63,64,70,71,81,83-85,94-98 at ED discharge. The interventions were associated with a significant reduction in the opioid prescription rate in both ITS (6-month step change, −22.61% [95% CI, −30.70% to −14.52%]; I2 = 77%) (Figure 2) and other study designs (OR, 0.56 [95% CI, 0.45 to 0.70]; I2 = 99%) (eFigure 4 in the Supplement). Interventions were not associated with a significant reduction in prescribed opioid quantity in ITS studies (6-month step change, −8.64% [95% CI, −17.48% to 0.20%]; I2 = 92%) (Figure 3). However, for other study designs, a small albeit significant reduction in prescribed quantity was found (SMD, −0.30 [95% CI, −0.51 to −0.09]; I2 = 100%) (eFigure 5 in the Supplement).
Intervention Category Analysis
Not all intervention categories had data for each outcome. For ITS design, no EMR-QC and PT studies reported data on prescription rate, and no CPC and PT studies reported data on prescription quantities. For other study designs, only PT studies did not report data on prescription quantities.
There were significant differences in the reduction of the opioid prescription rates between intervention categories in ITS study designs (P = .003). The EPG intervention (6-month step change, −33.31% [95% CI, −39.67% to −26.94%]; I2 = 0%) provided a larger reduction rate (P < .001) compared with PDMP (6-month step change, −11.18% [95% CI, −22.34% to −0.03%]; I2 = 81%) (Figure 2). For the other study designs, EPG (OR, 0.47 [95% CI, 0.33-0.69]; I2 = 99%), CPC (OR, 0.46 [95% CI, 0.29-0.72]; I2 = 96%), and PDMP (OR, 0.61 [95% CI, 0.44-0.86]; I2 = 96%) showed a statistically significant reduction in the opioid prescription rate (P < .001) compared with PT (OR, 0.98 [95% CI, 0.49-1.95]; I2 = 75%) and EMR-QC (OR, 0.94 [95% CI, 0.88-0.99]; I2 = 99%), which did not show significant reduction (eFigure 4 in the Supplement).
In ITS studies, a statistically significant reduction in prescribed opioid quantities was demonstrated for EPG interventions (P < .001) (6-month step change, −15.38% [95% CI, −24.51% to −6.25%]; I2 = 18%) compared with PDMP (6-month step change, 3.62% [95% CI, 2.39% to 4.85%]; I2 = 0%) and EMR_QC (6-month step change, −11.65% [95% CI, −29.30% to 5.99%]; I2 = 87%) (Figure 3). In other study designs, PDMP (SMD, −0.37 [95% CI, −0.58 to −0.15]; I2 = 95%) showed a significant reduction in prescribed opioid quantities compared with EPG (SMD, −0.07 [95% CI, −0.15 to 0.02]; I2 = 33%), CPC (SMD, −0.51 [95% CI, −1.10 to 0.08]; I2 = 100%), and EMR_QC (SMD, −0.20 [95% CI, −0.47 to 0.07]; I2 = 100%) interventions (P = .03) (eFigure 5 in the Supplement).
Patients’ level of pain relief was not reported in any study. Patients’ satisfaction level with their opioid prescriptions was presented in 4 studies: 1 had a very low survey response rate (1.9%),57 2 reported no impact,87,89 and 1 found a slight gain (from 52% to 61%).88 One study82 reported on patients’ need for additional opioid prescriptions after the intervention and found no change for this outcome.
The 1-year step results of segmented ITS analysis for the opioid prescription rate and quantity are presented in eFigures 6 and 7 in the Supplement. The overall results of interventions and intervention categories were similar to those reported for the 6-month step change.
Removing the ITS studies with a high risk of bias (serious, critical, and high risk) left 10 studies27,36,54-57,88-91 with a moderate risk of bias and did not significantly alter our results (eFigures 8 and 9 in the Supplement). The RCT study (low risk of bias),86 with its CPC intervention, demonstrated a significant reduction in opioid prescription rate (−5.5%) and quantity (−8 milligram morphine equivalent).
Relative visual asymmetry was found in the 4 funnel plots for both outcomes and study designs (ITS vs others), suggesting a possible publication bias (eFigures 10-13 in the Supplement). However, Egger test results were nonsignificant for 6-month step change in the opioid prescription rate for ITS studies (Egger regression intercept: −1.65 [SE, 1.14]; P = .17), in the opioid prescription rate for RCT, preintervention-postintervention, and cohort studies (Egger regression intercept: 0.24 [SE, 3.4]; P = .94), in the prescribed opioid quantity for ITS studies (Egger regression intercept: −2.22 [SE, 1.22]; P = .10), and in the prescribed opioid quantity for RCT and preintervention-postintervention studies (Egger regression intercept: 7.08 [SE, 5.92]; P = .26).
This meta-analysis of recent studies originating mostly from the US (87%) showed that specific interventions were associated with a reduction of opioid prescription rates, but interventions in general were limited in reducing prescribed opioid quantities. In a subgroup analysis of the more robust ITS study designs, we showed that EPG interventions resulted in a larger prescription rate reduction compared with PDMP interventions. In addition, only EPG interventions were associated with a reduction in prescribed opioid quantities in ITS designs. Insufficient data were available on patient-centered secondary outcomes to reach any conclusion.
We included 10 ITS studies that presented a moderate risk of bias and a single RCT study that exhibited a low risk of bias. For the remaining studies, the risk of bias was generally high, mainly because of the nature of their preintervention-postintervention designs. Interrupted time series studies are among the strongest evaluative designs when randomization is not possible; they are considered a robust design commonly used to evaluate the impact of interventions and programs implemented in health care settings.105 Given that they are frequently undertaken in real-world settings, ITS studies may have stronger external validity.106
We excluded studies conducted in populations other than patients discharged from the ED with pain or in pediatric populations because the intervention impact might be different from that obtained in adults discharged from the ED. We also excluded case reports and case series because of the high risk of bias associated with these designs. We found no significant publication bias, because all Egger test results were nonsignificant. Overall, significant heterogeneity was expected because of the diverse intervention categories and study designs. However, the heterogeneity was generally lower for intervention categories in ITS design studies, particularly for EPG interventions (I2 = 0% and I2 = 18%).
In a sensitivity analysis, we excluded studies at high risk of bias and found that exclusion did not significantly change our main results (significant reduction of the prescription rate and a trend for a reduction in prescribed opioid quantities), which demonstrates their consistency. This association also persisted in the 1-year sensitivity analysis. Moreover, most studies (90%) corroborated the association between interventions and prescription rate reductions. This finding suggests that our results are consistent for this outcome.
Comparison With Previous Systematic Reviews
Systematic reviews of interventions to decrease opioid prescribing (rate or quantity) after surgery have demonstrated similar results.107,108 Although no PDMP or PT interventions were found in these reviews, they identified a type of intervention that relied exclusively on patient education. We have not encountered interventions of this nature in the present review.
Two systematic reviews109,110 specifically looked at US prescription drug monitoring programs in the postsurgical setting and found mixed results; they were not able to perform a meta-analysis of the reviewed studies. Similarly, other reviews not limited to the ED111-113 reported mixed results but also showed less effect in the ED and a lack of data on potential harms: no pain relief, increased ED revisits, and physicians changing their prescriptions to opioids not on the prescription drug monitoring programs (usually weaker). We included state laws as interventions in our PDMP category. Davis et al114 concluded that no data are available on the association between legislation and opioid-related morbidity or mortality and unintended negative outcomes. In our review, PDMP interventions demonstrated a small reduction of opioid prescription rate or quantity. However, in more robust ITS studies, PDMP interventions were associated with a small increase in the quantity of opioids prescribed. In addition, Moride et al111 found that prescription drug monitoring programs were mostly performed when abuse and diversions were suspected; these programs could thus reduce “drug-shopping” (obtaining multiple opioid prescriptions for the same problem). However, physicians being reassured that the patient is not seeking the drug might explain the increase in opioid quantities prescribed in PDMP studies. The ED’s rapid pace may prevent it from being an environment conducive to this intervention strategy.
Unexpectedly, the EMR_QC interventions that were specifically designed to reduce the quantity of opioids per prescription showed no significant reduction. Other reviews107,115 have reported a similar lack of an association between interventions and reduced prescriptions. However, this approach has been useful in postoperative settings116 and in ambulatory contexts when associated with state prescribing limits.117 The heterogeneity of the EMR_QC studies included in our review was high and may have contributed to the results; in some studies, the default EMR quantity of opioids to prescribe was reduced,37,40,95,98 whereas in others it was simply removed.39,97,99
The reduction observed for EPG interventions was somewhat unexpected considering that another review118 demonstrated that physicians’ adherence to guidelines was low for chronic noncancer pain treated with opioids. However, in that review, none of the included studies was mainly conducted in an ED setting, and EPG interventions led in an institutional context were found to be associated with a reduction in opioid prescriptions. Interestingly, some of these studies115,119 also included interventions focused on patients and other clinicians. Other reviews115,120 proposed that a multimodal educational strategy simultaneously targeting clinicians, patients, and other collaborators was the most promising approach to improve appropriate opioid use. However, they also showed numerous negative outcomes associated with strategies aiming to improve opioid use such as patients not receiving opioid prescriptions, overdose increase, naloxone-associated stigma, shifting the opioid crisis to a neighboring region, changing to another opioid class, and even increases in dose and proportion of prescribed opioids. These reviews also concluded that it would be ideal “to develop a method for reliably predicting the amount of opioid (if any) a patient may consider adequate for pain relief.”119(p70)
It is noteworthy to observe that there were almost no data on important patient-centered outcomes such as pain control level, satisfaction level, and additional opioid prescription needs across the 63 studies included in our systematic review. This finding can be explained primarily by the retrospective design of most of the studies. Considering the potential negative effect of these interventions on patient-centered outcomes, this outcome is a major knowledge gap. In the future, it is imperative to evaluate interventions designed to reduce opioid prescribing in a prospective manner and to integrate patient-centered outcomes.
Our review is limited by the quality of the included studies. However, our findings were consistent in the sensitivity analysis in which studies with a high risk of bias were excluded. Furthermore, 90% of included studies reported a reduction of prescription rate from almost all intervention categories. The intervention categories varied enormously on several parameters such as the type of intervention included, the way in which the intervention was implemented, the study design used, the type of outcomes measured, and the duration of the follow-up. Therefore, caution is warranted regarding the generalization of these categories. The heterogeneity (I2 value) was low for several categories in ITS studies except for PDMP studies on the rate of opioid prescribing. However, it was high for studies involving other study designs, suggesting high heterogeneity within preintervention-postintervention study designs. Some conclusions are limited by the small number of studies in the subgroup analyses. There were no ITS studies using EMR_QC for opioid prescription rate or using CPC for opioid prescribed quantity and no PT studies for either outcome. This lack of data limits our conclusions for these intervention categories. Furthermore, most studies were performed in the US and may not be generalizable to other health care systems. Physical therapy was the only type of intervention study in which participants were allocated at the physician discretion, leading to possible allocation bias. Except for ITS studies that account for secular tendency, the findings of most studies reflected the trend of decreased opioid prescribing in the ED within the past 10 years.6-8
The findings of this meta-analysis suggest that specific interventions may be better at reducing the rate (to a lesser extent in reducing the quantity) of prescribed opioids to patients discharged from the ED. Therefore, policy makers and clinicians should probably focus their efforts on these more promising approaches to reduce prescribing rates. However, researchers should address the important knowledge gap on the global effect of these interventions on patient-centered outcomes and use novel approaches to reduce the opioid quantity per prescription to patients discharged from the ED.
Accepted for Publication: November 9, 2021.
Published: January 13, 2022. doi:10.1001/jamanetworkopen.2021.43425
Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2022 Daoust R et al. JAMA Network Open.
Corresponding Author: Raoul Daoust, MD, MSc, Study Center in Emergency Medicine, Hôpital du Sacré-Coeur de Montréal, Le Centre Intégré Universitaire de Santé et de Services Sociaux du Nord-de-l’Île de-Montréal, 5400 Gouin Blvd W, Montreal QC H4J 1C5, Canada (raoul.daoust@umontreal.ca).
Author Contributions: Drs Daoust and Paquet had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Daoust, Paquet, Marquis, Chauny, Williamson, Émond, Cournoyer.
Acquisition, analysis, or interpretation of data: Daoust, Paquet, Marquis, Chauny, Williamson, Huard, Arbour, Cournoyer.
Drafting of the manuscript: Daoust, Paquet.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Paquet, Chauny, Williamson, Huard, Cournoyer.
Obtained funding: Daoust, Marquis, Cournoyer.
Administrative, technical, or material support: Daoust, Marquis, Chauny, Williamson, Cournoyer.
Supervision: Daoust, Paquet, Chauny, Huard.
Conflict of Interest Disclosures: Dr Daoust reported receiving grants from the Emergency Department fund from the Hôpital du Sacré-Coeur de Montréal and the Instituts de Recherche en Santé during the conduct of the study. Dr Chauny reported receiving grants from the Instituts de Recherche en Santé during the conduct of the study. No other disclosures were reported.
Funding/Support: This study was supported by the Fonds des Urgentistes de l’Hôpital du Sacré-Coeur de Montréal and the OPUM (Quantity of Opioids for Acute Pain and Limit Unused Medication Study) study group.
Role of the Funder/Sponsor: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: Monique Clar, BSc, Université de Montréal, helped in designing the search strategy. Gloria De Buyl, BSc, a private translator, and Dominique Petit, PhD, Université de Montréal, helped revise the manuscript, for which they were compensated.
2.Canadian Institute for Health Information. Opioid Prescribing in Canada: How Are Practices Changing? Canadian Institute for Health Information; 2019.
17.McCarthy
DM, Kim
HS, Hur
SI,
et al. Patient-reported opioid pill consumption after an ED visit: how many pills are people using?
Pain Med. 2021;22(2):292-302. doi:
10.1093/pm/pnaa048PubMedGoogle Scholar 18.Daoust
R, Paquet
J, Cournoyer
A,
et al. Quantity of opioids consumed following an emergency department visit for acute pain: a Canadian prospective cohort study.
BMJ Open. 2018;8(9):e022649. doi:
10.1136/bmjopen-2018-022649
PubMedGoogle Scholar 19.Lipari
RN, Hughes
A. How People Obtain the Prescription Pain Relievers They Misuse: The CBHSQ Report. Substance Abuse and Mental Health Services Administration; 2013:1-7.
24.Baehren
DF, Marco
CA, Droz
DE, Sinha
S, Callan
EM, Akpunonu
P. A statewide prescription monitoring program affects emergency department prescribing behaviors.
Ann Emerg Med. 2010;56(1):19-23.e1-3. doi:
10.1016/j.annemergmed.2009.12.011Google Scholar 25.Cantrill
SV, Brown
MD, Carlisle
RJ,
et al; American College of Emergency Physicians Opioid Guideline Writing Panel. Clinical policy: critical issues in the prescribing of opioids for adult patients in the emergency department.
Ann Emerg Med. 2012;60(4):499-525. doi:
10.1016/j.annemergmed.2012.06.013
PubMedGoogle ScholarCrossref 27.Suffoletto
B, Lynch
M, Pacella
CB, Yealy
DM, Callaway
CW. The effect of a statewide mandatory prescription drug monitoring program on opioid prescribing by emergency medicine providers across 15 hospitals in a single health system.
J Pain. 2018;19(4):430-438. doi:
10.1016/j.jpain.2017.11.010
PubMedGoogle ScholarCrossref 35.Donaldson
SR, Harding
AM, Taylor
SE, Vally
H, Greene
SL. Evaluation of a targeted prescriber education intervention on emergency department discharge oxycodone prescribing.
Emerg Med Australas. 2017;29(4):400-406. doi:
10.1111/1742-6723.12772
PubMedGoogle ScholarCrossref 37.Delgado
MK, Shofer
FS, Patel
MS,
et al. Association between electronic medical record implementation of default opioid prescription quantities and prescribing behavior in two emergency departments.
J Gen Intern Med. 2018;33(4):409-411. doi:
10.1007/s11606-017-4286-5PubMedGoogle ScholarCrossref 38.Montoy
JCC, Coralic
Z, Herring
AA, Clattenburg
EJ, Raven
MC. Association of default electronic medical record settings with health care professional patterns of opioid prescribing in emergency departments: a randomized quality improvement study.
JAMA Intern Med. 2020;180(4):487-493. doi:
10.1001/jamainternmed.2019.6544
PubMedGoogle ScholarCrossref 44.Higgins
JPT, Thomas
J, Chandler
J, et al, eds. Cochrane Handbook for Systematic Reviews of Interventions. 2nd ed. John Wiley & Sons; 2019.
45.Effective Practice and Organisation of Care (EPOC) Group. Suggested Risk of Bias Criteria for EPOC Reviews. Norwegian Knowledge Centre for the Health Services; 2014.
48.Ramsay
CR, Grimshaw
JM, Grilli
R. Robust methods for reanalysis of interrupted time series designs for inclusion in systematic reviews. In:
9th International Cochrane Colloquium, Lyon, France, 9–13 October 2001. BioMed Central; 2001. doi:
10.1186/2048-4623-1-S3-PA006 49.Ramsay
CR, Matowe
L, Grilli
R, Grimshaw
JM, Thomas
RE. Interrupted time series designs in health technology assessment: lessons from two systematic reviews of behavior change strategies.
Int J Technol Assess Health Care. 2003;19(4):613-623. doi:
10.1017/S0266462303000576
PubMedGoogle ScholarCrossref 50.Kontopantelis
E, Reeves
D. Performance of statistical methods for meta-analysis when true study effects are non-normally distributed: a comparison between DerSimonian-Laird and restricted maximum likelihood.
Stat Methods Med Res. 2012;21(6):657-659. doi:
10.1177/0962280211413451
PubMedGoogle ScholarCrossref 53.Borenstein
M, Rothstein
D, Cohen
D. Comprehensive Meta-analysis: A Computer Program for Research Synthesis. Biostat; 2005.
57.Acquisto
NM, Schult
RF, Sarnoski-Roberts
S,
et al. Effect of pharmacist-led task force to reduce opioid prescribing in the emergency department.
Am J Health Syst Pharm. 2019;76(22):1853-1861. doi:
10.1093/ajhp/zxz204PubMedGoogle ScholarCrossref 59.Jacobs
D, Vearrier
D. Implementation of a voluntary opioid prescribing guideline is effective in altering the prescribing habits of emergency department doctors in an urban setting.
Clin Toxicol (Phila). 2015;53:659.
Google Scholar 60.McGhee
JD, Bounds
RB, Papas
MA, Coletti
C. Implementation of a statewide opiate prescribing policy is not associated with a significant decrease in opiate prescriptions from the emergency department.
Value in Health. 2015:A306. doi:
10.1016/j.jval.2015.03.1782Google Scholar 62.Divino
V, Cepeda
MS, Coplan
P, Maziere
JY, Yuan
Y, Wade
RL. Assessing the impact of the extended-release/long-acting opioid analgesics risk evaluation and mitigation strategies on opioid prescription volume.
J Opioid Manag. 2017;13(3):157-168. doi:
10.5055/jom.2017.0383
PubMedGoogle ScholarCrossref 63.Motov
S, Drapkin
J, Butt
M,
et al. Analgesic administration for patients with renal colic in the emergency department before and after implementation of an opioid reduction initiative.
West J Emerg Med. 2018;19(6):1028-1035. doi:
10.5811/westjem.2018.9.38875
PubMedGoogle ScholarCrossref 66.Gordon
BD, Westgard
BC, Nelson
J, Zwank
MD, Chang
A. Electronic prescribing implementation decreases opiate prescriptions in an academic emergency department.
Acad Emerg Med. 2019:S74-S75.
Google Scholar 68.Pattullo
C, Donovan
P, Thomson
J, Suckling
B, Taylor
S, Bell
A. Towards state-wide opioid stewardship: the adaptation of the opioid prescribing toolkit in multiple emergency departments.
Emerg Med Australasia. 2020:16.
Google Scholar 69.Hartmann
RJ, Elder
JD, Terrett
LA. Impact of an emergency department opioid prescribing guideline on emergency physician behaviour and incidence of overdose in the Saskatoon Health Region: a retrospective pre-post implementation analysis.
CMAJ Open. 2021;9(1):E79-E86. doi:
10.9778/cmajo.20200071
PubMedGoogle ScholarCrossref 70.Pattullo
C, Suckling
B, Taylor
S,
et al. Developing and piloting an adaptable oxycodone quality improvement strategy: steps towards opioid stewardship.
Aust Health Rev. 2021;45(3):353-360. doi:
10.1071/AH20262PubMedGoogle ScholarCrossref 71.Sun
BC, Lupulescu-Mann
N, Charlesworth
CJ,
et al. Impact of hospital “best practice” mandates on prescription opioid dispensing after an emergency department visit.
Acad Emerg Med. 2017;24(8):905-913. doi:
10.1111/acem.13230PubMedGoogle ScholarCrossref 73.Bornstein
K, Supino
M, De Melo Panakos
P, Freeman
C. Effect of prescription drug monitoring program databases on opiate prescribing at a large county safety-net hospital emergency department.
Ann Emerg Med. 2019:S123-S124. doi:
10.1016/j.annemergmed.2019.08.274Google Scholar 77.Antkowiak
P, Strout
TD, Haydar
S. Effect of a controlled substances law on prescribing patterns of emergency providers.
Acad Emerg Med. 2018;S131.
Google Scholar 78.Love
J, Mudan
A, Shofer
F, Perrone
J. Emergency department discharge prescribing in association with a new prescription drug monitoring program.
J Med Toxicol. 2018;25.
Google Scholar 81.Weiner
SG, Kobayashi
K, Reynolds
J,
et al. The effect of prescription drug monitoring program electronic medical record integration on prescription drug monitoring program utilization and opioid prescribing.
Ann Emerg Med. 2019;74(4):S51. doi:
10.1016/j.annemergmed.2019.08.131Google ScholarCrossref 84.Sigal
A, Shah
A, Onderdonk
A, Deaner
T, Schlappy
D, Barbera
C. Alternatives to opioid education and a prescription drug monitoring program cumulatively decreased outpatient opioid prescriptions.
Pain Med. 2021;22(2):499-505. doi:
10.1093/pm/pnaa278
PubMedGoogle ScholarCrossref 85.Zeiner
AL, Burak
MA, O’Sullivan
DM, Laskey
D. Effect of a law requiring prescription drug monitoring program use on emergency department opioid prescribing: a single-center analysis.
J Pharm Pract. 2021;34(5):774-779. doi:
10.1177/0897190020918096
PubMedGoogle Scholar 86.Michael
SS, Babu
KM, Androski
C
Jr, Reznek
MA. Effect of a data-driven intervention on opioid prescribing intensity among emergency department providers: a randomized controlled trial.
Acad Emerg Med. 2018;25(5):482-493. doi:
10.1111/acem.13400PubMedGoogle ScholarCrossref 87.Guarisco
J, Salup
A. Reducing opioid prescribing rates in emergency medicine.
Ochsner J. 2018;18(1):42-45.
PubMedGoogle Scholar 88.Meisenberg
BR, Grover
J, Campbell
C, Korpon
D. Assessment of opioid prescribing practices before and after implementation of a health system intervention to reduce opioid overprescribing.
JAMA Netw Open. 2018;1(5):e182908. doi:
10.1001/jamanetworkopen.2018.2908
PubMedGoogle Scholar 89.Andereck
JW, Reuter
QR, Allen
KC,
et al. A quality improvement initiative featuring peer-comparison prescribing feedback reduces emergency department opioid prescribing.
Jt Comm J Qual Patient Saf. 2019;45(10):669-679. doi:
10.1016/j.jcjq.2019.07.008
PubMedGoogle Scholar 91.Dieujuste
N, Johnson-Koenke
R, Christopher
M,
et al. Feasibility study of a quasi-experimental regional opioid safety prescribing program in Veterans Health Administration emergency departments.
Acad Emerg Med. 2020;27(8):734-741. doi:
10.1111/acem.13980PubMedGoogle ScholarCrossref 92.Schaefer
T, Wolford
R, Bowman
C,
et al. Simple interventions reduce emergency department opioid prescriptions: a quality improvement project.
Acad Emerg Med. 2018;S111-S112.
Google Scholar 93.Anhalt
M, Tippery
A, Bidad
R, Anhalt
D, Blohm
E. Comprehensive approach to sustainable reduction in emergency department opioid prescribing.
West J Emerg Med. 2019;20(5.1):S17.
Google Scholar 94.Yang
F, Dreyer
J, Van Aarsen
K. Assessing opioid-prescribing patterns for low back pain patients before and after the implementation of clinician performance indicators in the emergency department.
CJEM. 2020;22(suppl S1):S31. doi:
10.1017/cem.2020.120Google ScholarCrossref 97.Villwock
JA, Villwock
MR, New
J, Ator
GA. EMR quantity autopopulation removal on hospital discharge prescribing patterns: implications for opioid stewardship.
J Clin Pharm Ther. 2020;45(1):160-168. doi:
10.1111/jcpt.13049
PubMedGoogle ScholarCrossref 98.Carlson
A, Nelson
ME, Patel
H. Longitudinal impact of a pre-populated default quantity on emergency department opioid prescriptions.
J Am Coll Emerg Physicians Open. 2020;2(1):e12337. doi:
10.1002/emp2.12337PubMedGoogle Scholar 100.Smalley
CM, Willner
MA, Muir
MR,
et al. Electronic medical record-based interventions to encourage opioid prescribing best practices in the emergency department.
Am J Emerg Med. 2020;38(8):1647-1651. doi:
10.1016/j.ajem.2019.158500PubMedGoogle Scholar 104.Pugh
A, Roper
K, Magel
J,
et al. Dedicated emergency department physical therapy is associated with reduced imaging, opioid administration, and length of stay: a prospective observational study.
PLoS One. 2020;15(4):e0231476. doi:
10.1371/journal.pone.0231476
PubMedGoogle Scholar 105.Ewusie
JE, Soobiah
C, Blondal
E, Beyene
J, Thabane
L, Hamid
JS. Methods, applications and challenges in the analysis of interrupted time series data: a scoping review.
J Multidiscip Healthc. 2020;13:411-423. doi:
10.2147/JMDH.S241085
PubMedGoogle ScholarCrossref 106.Bernal
JL, Cummins
S, Gasparrini
A. Interrupted time series regression for the evaluation of public health interventions: a tutorial.
Int J Epidemiol. 2017;46(1):348-355. doi:
10.1093/ije/dyw098PubMedGoogle Scholar 111.Moride
Y, Lemieux-Uresandi
D, Castillon
G,
et al. A systematic review of interventions and programs targeting appropriate prescribing of opioids.
Pain Physician. 2019;22(3):229-240. doi:
10.36076/ppj/2019.22.229
PubMedGoogle Scholar 113.Rhodes
E, Wilson
M, Robinson
A, Hayden
JA, Asbridge
M. The effectiveness of prescription drug monitoring programs at reducing opioid-related harms and consequences: a systematic review.
BMC Health Serv Res. 2019;19(1):784. doi:
10.1186/s12913-019-4642-8
PubMedGoogle ScholarCrossref 115.Hopkins
RE, Bui
T, Magliano
D, Arnold
C, Dooley
M. Prescriber education interventions to optimize opioid prescribing in acute care: a systematic review.
Pain Physician. 2019;22(6):E551-E562. doi:
10.36076/ppj/2019.22.E551
PubMedGoogle Scholar 116.Chiu
AS, Jean
RA, Hoag
JR, Freedman-Weiss
M, Healy
JM, Pei
KY. Association of lowering default pill counts in electronic medical record systems with postoperative opioid prescribing.
JAMA Surg. 2018;153(11):1012-1019. doi:
10.1001/jamasurg.2018.2083
PubMedGoogle ScholarCrossref 118.Hossain
MA, Asamoah-Boaheng
M, Badejo
OA,
et al. Prescriber adherence to guidelines for chronic noncancer pain management with opioids: systematic review and meta-analysis.
Health Psychol. 2020;39(5):430-451. doi:
10.1037/hea0000830
PubMedGoogle ScholarCrossref 120.Furlan
AD, Carnide
N, Irvin
E,
et al. A systematic review of strategies to improve appropriate use of opioids and to reduce opioid use disorder and deaths from prescription opioids.
Canad J Pain. 2018;2(1):218-235. doi:
10.1080/24740527.2018.1479842
Google ScholarCrossref