Free
Pain Medicine  |   August 2019
Impact of an Opioid Safety Initiative on Patients Undergoing Total Knee Arthroplasty: A Time Series Analysis
Author Notes
  • From the Patient Safety Center of Inquiry, Veterans Affairs Boston Healthcare System, Boston, Massachusetts (Q.C.); the Patient Safety Center of Inquiry, Durham Veterans Affairs Healthcare System (H.-L.H., W.B., M.P., T.B., K.R.), the Department of Anesthesiology, Duke University Health System (H.-L.H., T.B., V.K., K.R.), and NoviSci, LLC. (R.O., M.A.B.), Durham, North Carolina; Anesthesiology and Perioperative Care Service, Veterans Affairs Palo Alto Healthcare System and the Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine, Palo Alto, California (E.R.M., S.C.M.); Veterans Affairs Pharmacy Benefits Management Services, Hines, Illinois (C.B.G.), and the Center for Value Based Pharmacy Initiatives, University of Pittsburgh Medical Center Health Plan, Pittsburgh, Pennsylvania (C.B.G.).
  • This article is featured in “This Month in Anesthesiology,” page 1A.
    This article is featured in “This Month in Anesthesiology,” page 1A.×
  • Supplemental Digital Content is available for this article. Direct URL citations appear in the printed text and are available in both the HTML and PDF versions of this article. Links to the digital files are provided in the HTML text of this article on the Journal’s Web site (www.anesthesiology.org).
    Supplemental Digital Content is available for this article. Direct URL citations appear in the printed text and are available in both the HTML and PDF versions of this article. Links to the digital files are provided in the HTML text of this article on the Journal’s Web site (www.anesthesiology.org).×
  • This article has a visual abstract available in the online version.
    This article has a visual abstract available in the online version.×
  • Submitted for publication September 22, 2018. Accepted for publication April 3, 2019.
    Submitted for publication September 22, 2018. Accepted for publication April 3, 2019.×
  • Address correspondence to Dr. Raghunathan: Patient Safety Center of Inquiry at the Durham Veterans Affairs, Department of Anesthesiology, Duke University Medical Center, DUMC 3094, Durham, North Carolina 27710. Karthik.Raghunathan@duke.edu. This article may be accessed for personal use at no charge through the Journal Web site, www.anesthesiology.org.
Article Information
Pain Medicine / Clinical Science / Pain Medicine / Opioid
Pain Medicine   |   August 2019
Impact of an Opioid Safety Initiative on Patients Undergoing Total Knee Arthroplasty: A Time Series Analysis
Anesthesiology 8 2019, Vol.131, 369-380. doi:https://doi.org/10.1097/ALN.0000000000002771
Anesthesiology 8 2019, Vol.131, 369-380. doi:https://doi.org/10.1097/ALN.0000000000002771
Abstract

Editor’s Perspective:

What We Already Know about This Topic:

  • Opioid overuse remains rampant even in hospitals, but whether administrative opioid safety initiatives reduce use remains unclear

What This Article Tells Us That Is New:

  • The authors evaluated the effects of a Veterans Administration national Opioid Safety Initiative using interrupted time series analysis to compare trends before and after starting the initiative

  • There was a trivial increase in pain scores, and a substantial reduction in patients with chronic preoperative and postoperative opioid prescriptions

Background: The Opioid Safety Initiative decreased high-dose prescriptions across the Veterans Health Administration. This study sought to examine the impact of this intervention (i.e., the Opioid Safety Initiative) on pain scores and opioid prescriptions in patients undergoing total knee arthroplasty.

Methods: This was an ecological study of group-level data among 700 to 850 patients per month over 72 consecutive months (January 2010 to December 2015). The authors examined characteristics of cohorts treated before versus after rollout of the Opioid Safety Initiative (October 2013). Each month, the authors aggregated at the group-level the differences between mean postoperative and preoperative pain scores for each patient (averaged over 6-month periods), and measured proportions of patients (per 1,000) with opioid (and nonopioid) prescriptions for more than 3 months in 6-month periods, preoperatively and postoperatively. The authors compared postintervention trends versus trends forecasted based on preintervention measures.

Results: After the Opioid Safety Initiative, patients were slightly older and sicker, but had lower mortality rates (postintervention n = 28,509 vs. preintervention n = 31,547). Postoperative pain scores were slightly higher and the decrease in opioid use was statistically significant, i.e., 871 (95% CI, 474 to 1,268) fewer patients with chronic postoperative prescriptions. In time series analyses, mean postoperative minus preoperative pain scores had increased from 0.65 to 0.81, by 0.16 points (95% CI, 0.05 to 0.27). Proportions of patients with chronic postoperative and chronic preoperative opioid prescriptions had declined by 20% (n = 3,355 vs. expected n = 4,226) and by 13% (n = 5,861 vs. expected n = 6,724), respectively. Nonopioid analgesia had increased. Sensitivity analyses confirmed all findings.

Conclusions: A system-wide initiative combining guideline dissemination with audit and feedback was effective in significantly decreasing opioid prescriptions in populations undergoing total knee arthroplasty, while minimally impacting pain scores.

Death from opioid overdose continues to rise in the United States and, rather than prescription drugs, the epidemic is currently being fueled by illicit synthetic opioids (fentanyl).1  As state and federal agencies mount responses to counter this threat, measures to decrease opioid prescriptions are being implemented across health systems.2–4  Within the Veterans Health Administration, high-dose opioid prescriptions and concurrent opioid-benzodiazepine prescriptions were substantially decreased after an Opioid Safety Initiative was implemented in 2011 within the Minneapolis Veterans Affairs Health Care System.5  The Opioid Safety Initiative was then rolled out nationwide in fiscal year 2013 (before October 2013) using an academic detailing approach (i.e., educational outreach that involved face-to-face tutorials with prescribers) that combined guideline dissemination with audit and feedback using dashboards.6,7  Specifically, the guideline was developed by a panel of multidisciplinary experts based on various clinical and epidemiologic evidence to provide healthcare providers with a framework by which to evaluate, treat, and manage the individual needs and preferences of patients with chronic pain. The dashboard is a computerized data display tool that aggregates electronic medical record data and visually tracks opioid prescription at national, regional, facility, and provider level of opioid prescription so that leadership at each facility could audit the data and provide feedback. A study examining the system-wide impact of the Opioid Safety Initiative found that patients with prescriptions for greater than 100 and greater than 200 mg/day in oral morphine equivalents, or concurrent opioid-benzodiazepine prescriptions, decreased by 16%, 24%, and 21%, respectively, over 1 yr after the Opioid Safety Initiative (vs. 1 yr before).7  Reductions in opioid prescriptions have also been reported nationally (outside the Veterans Health Administration), but with large variations at the county level.8  Whether, and how, pain scores are impacted by ongoing reductions in opioid prescriptions is unclear. It is also not known whether reductions in opioid prescriptions have occurred in populations receiving low daily doses of opioids within the Veterans Health Administration (not displayed in Opioid Safety Initiative dashboards).7 
There is concern that pain may be undertreated as opioid prescriptions decline.9  Such undertreatment may be particularly relevant in patients with osteoarthritis, and after total knee arthroplasty, since persistent postsurgical pain is common and total knee arthroplasty is the most frequent major operation in the nation.10,11  Although guidelines discourage long-term opioid therapy for chronic noncancer pain, new persistent opioid use is now described as the most common postoperative complication in the United States, especially after total knee arthroplasty.12–14  Most veterans receiving long-term opioid therapy after surgery are prescribed low daily doses, well below the thresholds cited in guidelines.14,15  The impact of the Opioid Safety Initiative on such populations is unclear, but relevant to understanding the effects of this intervention in populations that were not targeted in audit and feedback dashboards.7  The objective of this study was to determine the impact of the Opioid Safety Initiative in populations undergoing total knee arthroplasty within the Veterans Health Administration, the largest integrated healthcare system in the United States, by examining changes in pain scores and opioid prescriptions over extended timeframes. This an ecological study was an observational study at the population level, rather than individual level, using a quasi-experimental design, i.e., interrupted time series analysis to evaluate the longitudinal casual effects of the rollout of the Opioid Safety Initiative.16 We hypothesized that the rollout of the Opioid Safety Initiative would lead to a remarkable reduction of opioid prescriptions.
Materials and Methods
This study was approved by the Institutional Review Board at the Durham Veterans Affairs Healthcare System (Durham, North Carolina). Informed consent was waived since the data, contained in the Veterans Health Administration’s repository of electronic medical records (the Corporate Data Warehouse), were deidentified and analyses were conducted at the population level.
Study Design
This ecological study used retrospective group-level data to examine the impact of the Opioid Safety Initiative on pain scores and prescriptions for analgesics before and after total knee arthroplasty. Measurements were made at the group level every month (see details in the Variables and Measurement of Pain Scores, Analgesic Prescriptions, and Other Outcomes section). Time series analyses were conducted to examine trends throughout the course of a 27-month period after implementation (October 2013 to December 2015) versus trends forecasted using pre–Opioid Safety Initiative data over the 45-month period before implementation (January 2010 to September 2013).
Setting and Study Cohort
The Corporate Data Warehouse includes detailed information on patient characteristics, pain scores, and prescriptions.17–20  Using International Classification of Diseases, Ninth Revision, Clinical Modification procedure codes (81.54 or 00.80 to 00.84, 81.55, 81.59 for primary or revision total knee arthroplasty), we identified all veterans undergoing total knee arthroplasty within the Veterans Health Administration between January 2010 and December 2015. We excluded those that did not have a valid operative note and extracted the following data in cohorts before versus after October 2013: age at the time of total knee arthroplasty (in years), gender, race, comorbidities relevant to chronic opioid use (mental health conditions and substance use disorders), a composite comorbidity score,12,13,15  and type of total knee arthroplasty (primary vs. revision). Since we planned to include all eligible patients as previously described, we did not conduct a previous power calculation; the sample size was based on the available data. Due to small sample sizes in some centers, monthly rates had high variability within centers, but were evaluated and determined not to warrant exclusions as outliers. Basic demographics of the before and after populations were calculated as mean ± SD for continuous variables and percentage for categorical variables. We then computed standardized mean differences in the distributions of these characteristics treated after versus before implementation of the Opioid Safety Initiative.21  The standardized mean difference was calculated as the difference in mean between groups (before vs. after the Opioid Safety Initiative) divided by the pooled SD among all patients. We also examined differences in 30-day, 90-day, and 1-yr mortality rates (effectively following patients through to December 2016), and differences in various pain score-related and analgesic prescription–related measures after versus before the Opioid Safety Initiative (details in the Study Cohort, Descriptive, and Outcome Data section).
Variables and Measurement of Pain Scores, Analgesic Prescriptions, and Other Outcomes
Pain scores recorded in the Corporate Data Warehouse (Vital Signs domain) represent patient-reported values on the 0 to 10 numeric rating scale, mandated as the “fifth vital sign” in nursing encounters within the Veterans Health Administration after 1998.18  For each patient, we calculated the mean preoperative and the mean postoperative pain score based on all scores reported in outpatient settings (excluding the emergency department) over the 6-month periods before admission and after discharge, respectively. Scores were averaged to minimize variations reported over shorter time frame.19,20  To reflect the change in scores around total knee arthroplasty, we calculated the difference between mean scores for each patient (i.e., mean postoperative score minus mean preoperative scores). Differences were positive, zero, or negative values depending on whether mean postoperative pain scores were greater than, equal to, or less mean preoperative scores. We aggregated this difference as a group-level measure in populations that underwent total knee arthroplasty each month (the unit of analysis). We then conducted time series analyses (described in the Statistical Methods: Time Series Analyses section).
Medications containing opioids (including codeine, morphine, oxycodone, hydrocodone, oxymorphone, hydromorphone, fentanyl, meperidine, pentazocine, propoxyphene, butorphanol, levorphanol, nalbuphine, tapentadol, and methadone) were identified in the Corporate Data Warehouse Outpatient Pharmacy domain. Consistent with the Veterans Health Administration’s definition, “chronic use” was defined as continued prescriptions for more than 3 months in a 6-month window.6  In populations that underwent total knee arthroplasty each month, we estimated the proportions of patients with chronic preoperative opioid use and chronic postoperative opioid use as the numbers (per 1,000 patients) with prescriptions for greater than 3 months in a 6-month period before admission and after discharge, respectively. In line with previous studies, we allowed for postoperative prescriptions in the first 90 days after surgery,12,13,15  thus measuring chronicity beyond postoperative month 3. We examined other prescriptions to ensure that the absence of opioid prescriptions was not attributable to receipt from non–Veterans Affairs sources. Using similar methods, we computed proportions of patients with prescriptions for nonsteroidal antiinflammatory drugs (NSAIDs), gabapentinoids, and acetaminophen over the same time periods. Analyses (described in Statistical Methods: Analyses) examined whether observed trends after implementation of the Opioid Safety Initiative differed from forecasted trends.
Statistical Methods: Time Series Analyses
We assessed the impact of the Opioid Safety Initiative using time series analyses that applied autoregressive quasi-Poisson models, a flexible class of models to examine serial correlations in a parsimonious way, to the measures listed above. The use of counts from an over dispersed Poisson distribution has been described in recent analyses of opioid prescriptions for chronic pain.22  Autoregressive curves were fit to pre–Opioid Safety Initiative data (monthly measures during the 45-months before October 2013) with a mean and variance computed as follows: mean = exp(α1R(i-1) +…+ αkR(i-k) + lag 12 months + lag 1 month + lag 3 months) * PY and, variance = mean * scale parameter, where R(i) is the rate Y(i)/PY(i), with Y(i) representing the number of events occurring within a specific amount of patient time PY(i) at time i. The autoregressive model equations included terms to account for seasonality, i.e., monthly seasonality as well as 1- and 3-month autoregressive lags. Forecasted trends based on these autoregressive curves were then compared with observations over the 27-month period after October 2013. Primary parameters examined were: (1) aggregate differences in mean pain scores; (2) proportions (per 1,000) with chronic postoperative opioid use; and (3) proportions (per 1,000) with chronic preoperative opioid use.
Sensitivity Analyses
Since chronic preoperative use is the key predictor of chronic postoperative opioid use,23  we reanalyzed data separately in veterans with and without chronic preoperative use. We also examined the impact of the Opioid Safety Initiative on parameters described previously using conventional segmented regression methods.24  These methods evaluate the slope (or change in rate) before and after the interruption to determine if the policy had an effect. We evaluated the sensitivity of estimates to “interruptions” at different time periods (September or August rather than October 2013) (change points).25  This was done by first determining when the policy was put into place, and the visually evaluating the time series to determine when the effect took place. In terms of alternative pain score–related measures, we examined proportions of patients with mild (0 to 4), moderate (5 to 7), and severe (8 to 10) preoperative and postoperative pain in the cohorts treated before versus after the Opioid Safety Initiative. We set a standardized mean difference of ± 0.2 as the threshold for meaningful differences, since large sample sizes can yield statistically significant differences of limited clinical relevance.25  We also examined the proportions of patients that transitioned from moderate/severe preoperative pain to mild postoperative pain, and from mild preoperative pain to moderate/severe postoperative pain. In terms of analgesic prescriptions, we examined the proportions (per 1,000) with various nonopioid prescriptions preoperatively, and postoperatively. We also examined the proportions of patients that transitioned from chronic preoperative to no postoperative opioid prescriptions (cessation of long-term opioid therapy) and, from no preoperative to chronic postoperative opioid prescriptions (new initiation of long-term opioid therapy). Lastly, we examined whether opioid prescriptions changed at each of the 85 Veterans Affairs facilities that conducted total knee arthroplasties from 2010 to 2015. We computed rates of cessation of opioid use in chronic preoperative opioid users at these individual Veterans Affairs hospitals, during the 27 months after and during the 45 months before the initiative (defining cessation as the absence of prescriptions for at least 180 days, beyond 90 days after discharge). We also examined interfacility differences adjusted for patient demographics, comorbidities, type of total knee arthroplasty, and facility total knee arthroplasty volume. All analyses were performed using R Statistical Software (version 3.4.0), accounting for seasonality (R package version 1.6.1; https://www.r-project.org/. Accessed April 10, 2018.). Statistical significance was set a priori at P < 0.05; and all tests were conducted as two-tailed.
Results
Study Cohort, Descriptive, and Outcome Data
Between January 2010 and December 2015, 60,056 veterans underwent total knee arthroplasty within the Veterans Health Administration—about 700 to 850 per month (increasing at the rate of 2.41 patients per month, see Supplemental Digital Content, fig. 1, http://links.lww.com/ALN/B950). In table 1, which compares demographics, comorbidities, and mortality in 31,547 patients treated during 45 months before, versus 28,509 patients treated during 27 months after, October 2013, we found that the cohort treated after the Opioid Safety Initiative had lower 30-day, 90-day, and 1-yr mortality rates (P = 0.011, 0.012, 0.013, respectively), despite being slightly older (mean age: 65 vs. 64 yr; P < 0.0001) and significantly sicker (mean Charlson score: 2.79 vs. 2.29; standardized mean difference: ~0.2). Depression and alcohol abuse, factors predictive of increased chronic opioid use,12,13  were also increased in the post–Opioid Safety Initiative cohort.
Table 1.
Patient Characteristics and Mortality
Patient Characteristics and Mortality×
Patient Characteristics and Mortality
Table 1.
Patient Characteristics and Mortality
Patient Characteristics and Mortality×
×
In table 2, differences in pain score–related measures were minimal (standardized mean differences: less than 0.1) although statistically significant (P < 0.001). Approximately one-third of all patients, respectively, increased versus decreased versus had no change in mean postoperative pain scores versus mean preoperative pain scores. After the Opioid Safety Initiative, there was an increase in the proportion of patients with mean postoperative pain scores greater than mean preoperative pain scores (from 33.6% before to 35.4% after the Opioid Safety Initiative; standardized mean difference: 0.04; P < 0.0001). However, proportions with mild preoperative pain and proportions with mild postoperative pain were both increased after the Opioid Safety Initiative. Also, nearly one-half of all patients reported mild postoperative pain after the Opioid Safety Initiative. There was no significant difference in transitions from mild preoperative pain to severe postoperative pain, or from severe preoperative pain to mild postoperative pain. In table 3, differences in measures related to analgesic prescriptions are significant. There were significantly fewer patients with chronic postoperative opioid use (26.9% pre–Opioid Safety Initiative vs. 14.1% post–Opioid Safety Initiative; standardized mean difference: 0.41), and significantly fewer patients with chronic preoperative opioid use (32.2% pre–Opioid Safety Initiative vs. 25.5% post–Opioid Safety Initiative; standardized mean difference: 0.17). These large reductions were attributable to increased intermittent use of opioids, rather than to increased cessation of opioids (table 3). There was a significant increase in the proportion of patients with cessation of long-term opioid therapy after surgery (standardized mean difference: 0.15), reflecting the transition from chronic preoperative to no postoperative opioid use (from 4.9 to 8.7%; P < 0.0001) after the Opioid Safety Initiative Opioid Safety Initiative. Conversely, new initiation of chronic postoperative opioid use had decreased from 1.6 to 0.8% (P < 0.0001) after the Opioid Safety Initiative. Other notable changes included a large reduction in patient controlled analgesia during hospitalization (standardized mean difference: 0.32). Prescriptions for nonopioid analgesics increased, most notably for: gabapentinoids preoperatively (standardized mean difference: 0.12); aspirin and gabapentinoids in the hospital (standardized mean difference: 0.34 and 0.28, respectively); and, for acetaminophen and NSAIDs postoperatively (standardized mean differences: 0.23 and 0.2, respectively).
Table 2.
Pain Score-related Measures: Before versus After the Opioid Safety Initiative
Pain Score-related Measures: Before versus After the Opioid Safety Initiative×
Pain Score-related Measures: Before versus After the Opioid Safety Initiative
Table 2.
Pain Score-related Measures: Before versus After the Opioid Safety Initiative
Pain Score-related Measures: Before versus After the Opioid Safety Initiative×
×
Table 3.
Analgesic Prescriptions: Before versus After the Opioid Safety Initiative
Analgesic Prescriptions: Before versus After the Opioid Safety Initiative×
Analgesic Prescriptions: Before versus After the Opioid Safety Initiative
Table 3.
Analgesic Prescriptions: Before versus After the Opioid Safety Initiative
Analgesic Prescriptions: Before versus After the Opioid Safety Initiative×
×
Main Results: Time Series Analyses
Time series analyses using autoregressive quasi-Poisson models confirmed a slight increase in mean postoperative pain scores relative to mean preoperative scores (fig. 1). Although patients continued to report lower scores after (vs. before) total knee arthroplasty, the extent of reduction was decreased. The Opioid Safety Initiative was associated with a relative increase (mean postoperative scores were greater than mean preoperative scores) from 0.65 to 0.81, by 0.16 points (95% CI, 0.05 to 0.27 points), with an observed measure of −0.65 points (95% CI, −0.54 to −0.76) after the Opioid Safety Initiative versus the forecasted measure of −0.81 points (based on pre–Opioid Safety Initiative trends). Conversely, the Opioid Safety Initiative had significantly accelerated an ongoing (pre–Opioid Safety Initiative) reduction in chronic postoperative opioid prescriptions (fig. 2). The Opioid Safety Initiative was associated with 871 (95% CI, 474 to 1,268) fewer patients with chronic postoperative prescriptions (fig. 2), a relative reduction of 20% (10 to 29%) versus the forecasted rate. The Opioid Safety Initiative was also associated with a smaller decrease in chronic preoperative prescriptions (fig. 3), with 863 (95% CI, 323 to 1,404) fewer patients with chronic preoperative opioid prescriptions, a relative reduction of ~13% (fig. 3).
Fig. 1.
Changes in pain at the population level. Each observation represents, at the population-level, mean postoperative minus mean preoperative pain scores every month. Over 72 consecutive months, on average, mean postoperative scores were less than mean preoperative scores. There was a significant change after October 2013, with an increase in mean postoperative scores relative to mean preoperative scores.
Changes in pain at the population level. Each observation represents, at the population-level, mean postoperative minus mean preoperative pain scores every month. Over 72 consecutive months, on average, mean postoperative scores were less than mean preoperative scores. There was a significant change after October 2013, with an increase in mean postoperative scores relative to mean preoperative scores.
Fig. 1.
Changes in pain at the population level. Each observation represents, at the population-level, mean postoperative minus mean preoperative pain scores every month. Over 72 consecutive months, on average, mean postoperative scores were less than mean preoperative scores. There was a significant change after October 2013, with an increase in mean postoperative scores relative to mean preoperative scores.
×
Fig. 2.
Changes in chronic postoperative opioid use. The proportion of patients with chronic postoperative opioid use (per 1,000 total knee arthroplasties) decreased every month, gradually, up to October 2013. However, there was a significant decline after roll out of the Opioid Safety Initiative (interrupted vertical line), with 20% fewer patients continuing chronic opioid use when compared to the number expected based on pre–Opioid Safety Initiative trends (observed n = 3,355 vs. expected n = 4,226). TKA, total knee arthroplasty.
Changes in chronic postoperative opioid use. The proportion of patients with chronic postoperative opioid use (per 1,000 total knee arthroplasties) decreased every month, gradually, up to October 2013. However, there was a significant decline after roll out of the Opioid Safety Initiative (interrupted vertical line), with 20% fewer patients continuing chronic opioid use when compared to the number expected based on pre–Opioid Safety Initiative trends (observed n = 3,355 vs. expected n = 4,226). TKA, total knee arthroplasty.
Fig. 2.
Changes in chronic postoperative opioid use. The proportion of patients with chronic postoperative opioid use (per 1,000 total knee arthroplasties) decreased every month, gradually, up to October 2013. However, there was a significant decline after roll out of the Opioid Safety Initiative (interrupted vertical line), with 20% fewer patients continuing chronic opioid use when compared to the number expected based on pre–Opioid Safety Initiative trends (observed n = 3,355 vs. expected n = 4,226). TKA, total knee arthroplasty.
×
Fig. 3.
Changes in chronic preoperative opioid use. The proportion of patients with chronic preoperative opioid use (per 1,000 total knee arthroplasties) decreased every month, gradually, up to October 2013, with a significant decline after roll out of the Opioid Safety Initiative (interrupted vertical line) ~13% fewer patients versus the number expected based on pre–Opioid Safety Initiative trends (observed n = 5,861 vs. expected n = 6,724). TKA, total knee arthroplasty.
Changes in chronic preoperative opioid use. The proportion of patients with chronic preoperative opioid use (per 1,000 total knee arthroplasties) decreased every month, gradually, up to October 2013, with a significant decline after roll out of the Opioid Safety Initiative (interrupted vertical line) ~13% fewer patients versus the number expected based on pre–Opioid Safety Initiative trends (observed n = 5,861 vs. expected n = 6,724). TKA, total knee arthroplasty.
Fig. 3.
Changes in chronic preoperative opioid use. The proportion of patients with chronic preoperative opioid use (per 1,000 total knee arthroplasties) decreased every month, gradually, up to October 2013, with a significant decline after roll out of the Opioid Safety Initiative (interrupted vertical line) ~13% fewer patients versus the number expected based on pre–Opioid Safety Initiative trends (observed n = 5,861 vs. expected n = 6,724). TKA, total knee arthroplasty.
×
Other Results: Sensitivity Analyses
The impact of the Opioid Safety Initiative on chronic postoperative prescriptions was sustained in the preoperatively naïve, but not in preoperatively opioid-using, populations (Supplemental Digital Content, fig. 2, top and bottom panel respectively, http://links.lww.com/ALN/B950). All findings were confirmed in conventional segmented regression analyses (Supplemental Digital Content, figs. 3 to 6, http://links.lww.com/ALN/B950). In segmented regression analyses using alternative interruptions in the time series (August 2013 and September 2013), we confirmed that October 2013 represented the ‘change point’ (Supplemental Digital Content, fig. 7, http://links.lww.com/ALN/B950).25  Increased prescriptions for acetaminophen, NSAIDs, and gabapentinoids were seen in both preoperative and postoperative periods (Supplemental Digital Content, figs. 8 to 13, http://links.lww.com/ALN/B950). With regard to the duration of postoperative opioid usage in opioid naïve patients, the proportion of patients who were on postoperative opioids for more than 180 days significantly decreased after the Opioid Safety Initiative (1.62% vs. 0.81%; P < 0.0001), as well as the proportion for patients who were on postoperative opioids for more than 90 days (5.82% vs. 2.6%; P < 0.0001). Lastly, greater proportions of patients ceased chronic opioid use over the 27 months after (when compared to the 45 months before) the initiative at every Veterans Affairs hospital. Both actual and adjusted rates of opioid cessation had decreased, with a greater than two-fold difference across Veterans Affairs hospitals, i.e., adjusted rates up to 12.8% in the top quartile versus 5.1% in the bottom quartile (Supplemental Digital Content, fig. 14, http://links.lww.com/ALN/B950).
Discussion
In this ecological study we sought to evaluate the impact of an opioid safety initiative, implemented across the Veterans Health Administration, on opioid prescriptions and pain scores in populations undergoing total knee arthroplasty—the most common major surgical procedure in the nation.10  Using detailed electronic health record data, we found that the post–Opioid Safety Initiative cohort was slightly older and sicker (on average), with decreased rates of death within 30 days, 90 days, and 1 yr after discharge (table 1). Mean postoperative pain scores were minimally increased relative to mean preoperative scores, after the Opioid Safety Initiative (table 2). Chronic preoperative and postoperative opioid prescriptions were significantly reduced, with larger proportions of patients stopping opioid therapy and lower proportions starting new persistent opioid use after the Opioid Safety Initiative (table 3). Nonopioid prescriptions were also increased after the Opioid Safety Initiative (table 3). In robust time series analyses, chronic postoperative and chronic preoperative opioid prescriptions were substantially decreased versus forecasted trends (reductions of 20% and 13%, respectively), while pain scores were minimally increased at the population-level (aggregate increase of 0.16 points on the 11-point numeric rating scale; figs. 1 to 3). Time series confirmed significant increases in preoperative and postoperative prescriptions for gabapentinoids, acetaminophen, and NSAIDs (Supplemental Digital Content, figs. 8 to 13, http://links.lww.com/ALN/B950).
These findings address several gaps in the literature and are of immediate relevance to policy makers and clinicians seeking strategies to reduce long-term opioid therapy. First, the Opioid Safety Initiative used an academic detailing approach to encourage safer prescribing.7  Guideline dissemination was combined with audit and feedback using dashboards, while also providing for technical and educational support from the local Chief Medical Officer (or designee) and one licensed prescribing physician at each facility.5,7  The large system-wide reductions in opioid prescriptions that we have observed (figs. 2 and 3, and Supplemental Digital Content, figs. 4 and 5, http://links.lww.com/ALN/B950) are consistent with a recent Veterans Affairs report.7  However, our study focused on populations receiving lower daily doses of opioids (~80% of patients in our study received daily oral morphine equivalents of less than 30 mg/day) versus thresholds cited in guidelines (50 or 90 mg/day).14,20  Reductions in the proportions of patients receiving opioids is desirable at all doses because there is no clear dose threshold distinguishing between overdose cases versus controls.26  Second, the management of persistent postsurgical pain with opioids is not specifically addressed in current prescribing guidelines,14  and such pain is of significant concern after total knee arthroplasty.11  Our finding that pain scores were only minimally increased at the population-level is reassuring given concerns about undertreated pain when opioid prescriptions were reduced.9  Thus, reductions in opioid use after total knee arthroplasty are desirable. Third, audit and feedback has been defined as the provision of clinical performance information to healthcare providers, teams, and organizations.27  This is based on the theory that providers may be better equipped to self-assess performance, and subsequently improve alignment of practice with recommendations, when comparing their practice to a target (e.g., a national or local measure).28,29  In previous literature, the effectiveness of audit and feedback varied widely and even effective feedback (resulting in increased adherence to guidelines) had not translated into better patient outcomes.30  Current evidence to support interventions reducing opioid prescriptions, as effective, is also of low quality.31  This study provides compelling evidence about the effectiveness of audit and feedback in reducing opioid prescribing in diverse populations (delivered in the Opioid Safety Initiative using multilevel dashboards, with technical and educational support). Lastly, no specific intervention was built into the Opioid Safety Initiative. Other than the provision of multilevel data in a standardized format (dashboard displays), Veterans Affairs clinicians were not directed toward a specific target.32  In line with other successful audit and feedback interventions,33–36  clinicians were free to interpret the data on their own practice and to choose a suitable target. As expected in academic detailing supported efforts, we observed increased prescriptions for other medications recommended in guidelines.37,38  This may have mitigated against higher pain scores, but implications deserve further study.
Strengths and Limitations
Time series analyses are commonly used to assess the impact of an intervention with clearly defined start points. We applied autoregressive quasi-Poisson models (with terms for seasonality) to study changes in pain scores and opioid use at the population-level in groups of patients undergoing total knee arthroplasty every month (700 to 850 per month) after a time point (October 2013) versus before it. We were able to account for ongoing reductions in opioid prescriptions (before the Opioid Safety Initiative), using data during 45 months before implementation.39  Comparing projected trends with observed trends (figs. 1 to 3), we found plausible changes in analgesic prescriptions across the Veterans Health Administration, a health system with effective services for pain management, and for the treatment of mental illnesses and substance use disorder.2,3  We confirmed all findings in additional distinct segmented regression analyses (Supplemental Digital Content, figs. 3 to 7, http://links.lww.com/ALN/B950). Confounding due to underlying changes in patient characteristics is unlikely given that populations treated after the Opioid Safety Initiative were likely to have increased (rather than reduced) opioid use. Depression and alcohol use disorder were slightly increased in the post–Opioid Safety Initiative period.12,13  We limited the study from January 2010 to December 2015, to avoid errors related to a transition to International Classification of Diseases, Tenth Revisions coding.
The study has many limitations. First, this study is ecological and patient-level trajectories in pain scores and opioid prescriptions were not examined. Results at the group-level may not translate into similar results at the patient-level. Second, the pain scores that we used, while readily available in electronic health records, do not provide biopsychosocial or functional information.19  We cannot determine how pain-related functioning was impacted when opioid use decreased. Third, time-varying confounding, e.g., changes in surgical and medical practice for total knee replacement patients over time remains a threat to validity.39  Changes in response to interventions other than the Opioid Safety Initiative may explain observed effects. Also, segmented regression analyses showed that the Opioid Safety Initiative was associated with sustained reductions in chronic postoperative opioid use in preoperatively naive patients, but not in preoperative opioid users (Supplemental Digital Content, figs. 3 to 7, http://links.lww.com/ALN/B950). This suggest differential impact of the Opioid Safety Initiative on subpopulations. Lastly, findings may not be generalizable to other populations undergoing surgery since we only included patients that underwent total knee arthroplasty.
In conclusion, this study used interrupted time series analysis of detailed electronic health record data to determine that long-term opioid therapy had significantly decreased after implementation of the Opioid Safety Initiative within the Veterans Health Administration, while pain scores had increased minimally at the population-level, in patients undergoing total knee arthroplasty. This quasi-experimental study supports the effectiveness of a modest provider-focused audit and feedback approach using dashboards. This is important for policy makers to consider, as a contrast to (more rigid) regulatory interventions aimed toward decreasing opioid prescriptions.40  Even in a population at-risk for persistent postsurgical pain, patients undergoing total knee arthroplasty, pain scores were only minimally impacted.11,41–43  Whether similar results are seen in patient-level analyses, and whether increased prescriptions for gabapentinoids, NSAIDs, and acetaminophen are safe require future study.
Research Support
The contents of this study represent the views of the authors and do not necessarily represent the views of the U.S. Department of Veterans Affairs or the United States Government. This study is funded by the National Center for Patient Safety, Field Office 10A4E, Office of Quality Safety and Value, Department of Veterans Affairs, through the Patient Safety Center of Inquiry at the Durham Veterans Affairs Medical Center, Durham, North Carolina. Dr. Raghunathan received salary support and the study’s design, conduct, and reporting was conceived independent of the funding office.
Competing Interests
The authors declare no competing interests.
References
Seth, P, Scholl, L, Rudd, RA, Bacon, S . Overdose deaths involving opioids, cocaine, and psychostimulants - United States, 2015-2016. MMWR Morb Mortal Wkly Rep 2018; 67:349–58 [Article] [PubMed]
Gellad, WF, Good, CB, Shulkin, DJ . Addressing the opioid epidemic in the United States: Lessons from the Department of Veterans Affairs. JAMA Intern Med 2017; 177:611–2 [Article] [PubMed]
Opioid Safety Initiative (OSI): VHA Pain Management [Internet]. Available at: https://www.va.gov/painmanagement/opioid_safety_initiative_osi.asp. Accessed April 6, 2018
Penm, J, MacKinnon, NJ, Boone, JM, Ciaccia, A, McNamee, C, Winstanley, EL . Strategies and policies to address the opioid epidemic: a case study of Ohio. J Am Pharm Assoc (2003) 2017; 572S 148–53 [Article]
Westanmo, A, Marshall, P, Jones, E, Burns, K, Krebs, EE . Opioid dose reduction in a VA Health Care System–implementation of a primary care population-level initiative. Pain Med 2015; 16:1019–26 [Article] [PubMed]
Management of Opioid Therapy (OT) for Chronic Pain (2017) [Internet]. Available at: https://www.healthquality.va.gov/guidelines/pain/cot/. Accessed June 6, 2018
Lin, LA, Bohnert, ASB, Kerns, RD, Clay, MA, Ganoczy, D, Ilgen, MA . Impact of the Opioid Safety Initiative on opioid-related prescribing in veterans. Pain 2017; 158:833–9 [Article] [PubMed]
Guy, JGP, Zhang, K, Bohm, MK, Losby, J. Vital signs: changes in opioid prescribing in the United States, 2006–2015. MMWR Morb Mortal Wkly Rep. 2017; Available at: https://europepmc.org/articles/pmc5726238. Accessed April 10, 2018
Comerci, GJr, Katzman, J, Duhigg, D . Controlling the swing of the opioid pendulum. N Engl J Med 2018; 378:691–3 [Article] [PubMed]
Fingar, KR, Stocks, C, Weiss, AJ, Steiner, CA . Most frequent operating room procedures performed in U.S. Hospitals, 2003–2012: statistical brief #186.. Healthcare Cost and Utilization Project (HCUP) Statistical Briefs. 2014. Rockville (MD): Agency for Healthcare Research and Quality (US)
Lewis, GN, Rice, DA, McNair, PJ, Kluger, M . Predictors of persistent pain after total knee arthroplasty: A systematic review and meta-analysis. Br J Anaesth 2015; 114:551–61 [Article] [PubMed]
Sun, EC, Darnall, BD, Baker, LC, Mackey, S . Incidence of and risk factors for chronic opioid use among opioid-naive patients in the postoperative period. JAMA Intern Med 2016; 176:1286–93 [Article] [PubMed]
Brummett, CM, Waljee, JF, Goesling, J, Moser, S, Lin, P, Englesbe, MJ, Bohnert, ASB, Kheterpal, S, Nallamothu, BK . New persistent opioid use after minor and major surgical procedures in US adults. JAMA Surg 2017; 152:e170504 [Article] [PubMed]
Dowell, D, Haegerich, TM, Chou, R . CDC Guideline for Prescribing Opioids for Chronic Pain–United States, 2016. JAMA 2016; 315:1624–45 [Article] [PubMed]
Mudumbai, SC, Oliva, EM, Lewis, ET, Trafton, J, Posner, D, Mariano, ER, Stafford, RS, Wagner, T, Clark, JD . Time-to-cessation of postoperative opioids: a population-level analysis of the Veterans Affairs Health Care System. Pain Med 2016; 17:1732–43 [Article] [PubMed]
Kontopantelis, E, Doran, T, Springate, DA, Buchan, I, Reeves, D . Regression based quasi-experimental approach when randomisation is not an option: Interrupted time series analysis. BMJ 2015; 350:h2750 [Article] [PubMed]
Noël, PH, Copeland, LA, Perrin, RA, Lancaster, AE, Pugh, MJ, Wang, CP, Bollinger, MJ, Hazuda, HP . VHA Corporate Data Warehouse height and weight data: Opportunities and challenges for health services research. J Rehabil Res Dev 2010; 47:739–50 [Article] [PubMed]
Pain Management VHA Directive 2009–053. U.S. Department of Veterans Affairs. Available at: https://www.va.gov/painmanagement/docs/vha09paindirective.pdf. Accessed June 15, 2018
Dobscha, SK, Morasco, BJ, Kovas, AE, Peters, DM, Hart, K, McFarland, BH . Short-term variability in outpatient pain intensity scores in a national sample of older veterans with chronic pain. Pain Med 2015; 16:855–65 [Article] [PubMed]
Hsia, H-L, Takemoto, S, van de Ven, T, Pyati, S, Buchheit, T, Ray, N, Wellman, S, Kuo, A, Wallace, A, Raghunathan, K. Acute pain is associated with chronic opioid use after total knee arthroplasty. Reg Anesth Pain Med [Internet]. Available at: http://dx.doi.org/10.1097/AAP.0000000000000831. Accessed July 2, 2018
Austin, SR, Wong, YN, Uzzo, RG, Beck, JR, Egleston, BL . Why summary comorbidity measures such as the Charlson Comorbidity Index and Elixhauser score work. Med Care 2015; 53:e65–72 [Article] [PubMed]
Tayeb, BO, Barreiro, AE, Bradshaw, YS, Chui, KK, Carr, DB . Durations of opioid, nonopioid drug, and behavioral clinical trials for chronic pain: Adequate or inadequate? Pain Med 2016; 17:2036–46 [Article] [PubMed]
Rozet, I, Nishio, I, Robbertze, R, Rotter, D, Chansky, H, Hernandez, AV . Prolonged opioid use after knee arthroscopy in military veterans. Anesth Analg 2014; 119:454–9 [Article] [PubMed]
Overman, RA, Farley, JF, Curtis, JR, Zhang, J, Gourlay, ML, Deal, CL . DXA Utilization between 2006 and 2012 in commercially insured younger postmenopausal women. J Clin Densitom 2015; 18:145–9 [Article] [PubMed]
Johnson, P, Moriarty, J, Peskir, G. Detecting changes in real-time data: A user’s guide to optimal detection. Philos Trans A Math Phys Eng Sci [Internet] 2017 Aug 13;375(2100). Available at: http://dx.doi.org/10.1098/rsta.2016.0298. Accessed April 15, 2018
Bohnert, AS, Logan, JE, Ganoczy, D, Dowell, D . A detailed exploration onto the association of prescribed opioid dosage and overdose deaths among patients with chronic pain. Med Care 2016; 54:435–41 [Article] [PubMed]
Brehaut, JC, Eva, KW . Building theories of knowledge translation interventions: Use the entire menu of constructs. Implement Sci 2012; 7:114 [Article] [PubMed]
Davis, DA, Mazmanian, PE, Fordis, M, Van Harrison, R, Thorpe, KE, Perrier, L . Accuracy of physician self-assessment compared with observed measures of competence: A systematic review. JAMA 2006; 296:1094–102 [Article] [PubMed]
Carver, CS, Scheier, MF . Control theory: A useful conceptual framework for personality-social, clinical, and health psychology. Psychol Bull 1982; 92:111–35 [Article] [PubMed]
Jamtvedt, G, Young, JM, Kristoffersen, DT, O’Brien, MA, Oxman, AD . Audit and feedback: Effects on professional practice and health care outcomes. Cochrane Database Syst Rev 20062 CD000259
Frank, JW, Lovejoy, TI, Becker, WC, Morasco, BJ, Koenig, CJ, Hoffecker, L, Dischinger, HR, Dobscha, SK, Krebs, EE . Patient outcomes in dose reduction or discontinuation of long-term opioid therapy: A systematic review. Ann Intern Med 2017; 167:181–91 [Article] [PubMed]
Statement of Dr. Carolyn Clancy, M.D. interim under Secretary for Health Veterans Health Administrations (VHA) Department of Veterans Affairs (VA) before the Committee on Veterans Affairs United States Senate [Internet]. Available at: https://www.veterans.senate.gov/imo/media/doc/VA%20Clancy%20Testimony%203.26.20151.pdf. Accessed October 6, 2018
Gude, WT, van Engen-Verheul, MM, van der Veer, SN, de Keizer, NF, Peek, N . How does audit and feedback influence intentions of health professionals to improve practice? A laboratory experiment and field study in cardiac rehabilitation. BMJ Qual Saf 2017; 26:279–87 [Article] [PubMed]
van der Veer, SN, de Vos, ML, Jager, KJ, van der Voort, PH, Peek, N, Westert, GP, Graafmans, WC, de Keizer, NF . Evaluating the effectiveness of a tailored multifaceted performance feedback intervention to improve the quality of care: Protocol for a cluster randomized trial in intensive care. Implement Sci 2011; 6:119 [Article] [PubMed]
Inga-Britt, L, Kerstin, B. Challenges faced in daily life by persons with type 2 diabetes: A meta-synthesis. Chronic Illn 2018 Jan 1;1742395318757852
Eva, KW, Armson, H, Holmboe, E, Lockyer, J, Loney, E, Mann, K, Sargeant, J . Factors influencing responsiveness to feedback: On the interplay between fear, confidence, and reasoning processes. Adv Health Sci Educ Theory Pract 2012; 17:15–26 [Article] [PubMed]
Ndefo, UA, Norman, R, Henry, A . Academic detailing has a positive effect on prescribing and decreasing prescription drug costs: A health plan’s perspective. Am Health Drug Benefits 2017; 10:129–33 [PubMed]
Rosenberg, JM, Bilka, BM, Wilson, SM, Spevak, C . Opioid therapy for chronic pain: Overview of the 2017 US Department of Veterans Affairs and US Department of Defense Clinical Practice Guideline. Pain Med 2018; 19:928–41 [Article] [PubMed]
Jandoc, R, Burden, AM, Mamdani, M, Lévesque, LE, Cadarette, SM . Interrupted time series analysis in drug utilization research is increasing: Systematic review and recommendations. J Clin Epidemiol 2015; 68:950–6 [Article] [PubMed]
Barlas, S . U.S. and states ramp up response to opioid crisis: Regulatory, legislative, and legal tools brought to bear. P T 2017; 42:569–92 [PubMed]
Hoffman, RS . Inadequate pain control versus opioid abuse: It is time for the pendulum to swing. Am J Health Syst Pharm 2014; 71:1537 [Article] [PubMed]
von Gunten, CF . The pendulum swings for opioid prescribing. J Palliat Med 2016; 19:348 [Article] [PubMed]
Schatman, ME, Darnall, BD . A pendulum swings awry: seeking the middle ground on opioid prescribing for chronic non-cancer pain. Pain Med 2013; 14:617 [PubMed]
Fig. 1.
Changes in pain at the population level. Each observation represents, at the population-level, mean postoperative minus mean preoperative pain scores every month. Over 72 consecutive months, on average, mean postoperative scores were less than mean preoperative scores. There was a significant change after October 2013, with an increase in mean postoperative scores relative to mean preoperative scores.
Changes in pain at the population level. Each observation represents, at the population-level, mean postoperative minus mean preoperative pain scores every month. Over 72 consecutive months, on average, mean postoperative scores were less than mean preoperative scores. There was a significant change after October 2013, with an increase in mean postoperative scores relative to mean preoperative scores.
Fig. 1.
Changes in pain at the population level. Each observation represents, at the population-level, mean postoperative minus mean preoperative pain scores every month. Over 72 consecutive months, on average, mean postoperative scores were less than mean preoperative scores. There was a significant change after October 2013, with an increase in mean postoperative scores relative to mean preoperative scores.
×
Fig. 2.
Changes in chronic postoperative opioid use. The proportion of patients with chronic postoperative opioid use (per 1,000 total knee arthroplasties) decreased every month, gradually, up to October 2013. However, there was a significant decline after roll out of the Opioid Safety Initiative (interrupted vertical line), with 20% fewer patients continuing chronic opioid use when compared to the number expected based on pre–Opioid Safety Initiative trends (observed n = 3,355 vs. expected n = 4,226). TKA, total knee arthroplasty.
Changes in chronic postoperative opioid use. The proportion of patients with chronic postoperative opioid use (per 1,000 total knee arthroplasties) decreased every month, gradually, up to October 2013. However, there was a significant decline after roll out of the Opioid Safety Initiative (interrupted vertical line), with 20% fewer patients continuing chronic opioid use when compared to the number expected based on pre–Opioid Safety Initiative trends (observed n = 3,355 vs. expected n = 4,226). TKA, total knee arthroplasty.
Fig. 2.
Changes in chronic postoperative opioid use. The proportion of patients with chronic postoperative opioid use (per 1,000 total knee arthroplasties) decreased every month, gradually, up to October 2013. However, there was a significant decline after roll out of the Opioid Safety Initiative (interrupted vertical line), with 20% fewer patients continuing chronic opioid use when compared to the number expected based on pre–Opioid Safety Initiative trends (observed n = 3,355 vs. expected n = 4,226). TKA, total knee arthroplasty.
×
Fig. 3.
Changes in chronic preoperative opioid use. The proportion of patients with chronic preoperative opioid use (per 1,000 total knee arthroplasties) decreased every month, gradually, up to October 2013, with a significant decline after roll out of the Opioid Safety Initiative (interrupted vertical line) ~13% fewer patients versus the number expected based on pre–Opioid Safety Initiative trends (observed n = 5,861 vs. expected n = 6,724). TKA, total knee arthroplasty.
Changes in chronic preoperative opioid use. The proportion of patients with chronic preoperative opioid use (per 1,000 total knee arthroplasties) decreased every month, gradually, up to October 2013, with a significant decline after roll out of the Opioid Safety Initiative (interrupted vertical line) ~13% fewer patients versus the number expected based on pre–Opioid Safety Initiative trends (observed n = 5,861 vs. expected n = 6,724). TKA, total knee arthroplasty.
Fig. 3.
Changes in chronic preoperative opioid use. The proportion of patients with chronic preoperative opioid use (per 1,000 total knee arthroplasties) decreased every month, gradually, up to October 2013, with a significant decline after roll out of the Opioid Safety Initiative (interrupted vertical line) ~13% fewer patients versus the number expected based on pre–Opioid Safety Initiative trends (observed n = 5,861 vs. expected n = 6,724). TKA, total knee arthroplasty.
×
Table 1.
Patient Characteristics and Mortality
Patient Characteristics and Mortality×
Patient Characteristics and Mortality
Table 1.
Patient Characteristics and Mortality
Patient Characteristics and Mortality×
×
Table 2.
Pain Score-related Measures: Before versus After the Opioid Safety Initiative
Pain Score-related Measures: Before versus After the Opioid Safety Initiative×
Pain Score-related Measures: Before versus After the Opioid Safety Initiative
Table 2.
Pain Score-related Measures: Before versus After the Opioid Safety Initiative
Pain Score-related Measures: Before versus After the Opioid Safety Initiative×
×
Table 3.
Analgesic Prescriptions: Before versus After the Opioid Safety Initiative
Analgesic Prescriptions: Before versus After the Opioid Safety Initiative×
Analgesic Prescriptions: Before versus After the Opioid Safety Initiative
Table 3.
Analgesic Prescriptions: Before versus After the Opioid Safety Initiative
Analgesic Prescriptions: Before versus After the Opioid Safety Initiative×
×