Association of Hydrocodone Schedule Change With Opioid Prescriptions Following Surgery

Key Points

Question  Does an association exist between the October 2014 US Drug Enforcement Administration change for hydrocodone-containing medications from schedule III to II and the amount of opioid filled following surgery?

Findings  In this analysis of claims data of 21 955 privately insured surgery patients in Michigan, there was an immediate significant increase following the schedule change in the amount of opioids filled in the initial postoperative prescription, which was sustained for 1 year.

Meaning  Future efforts to curb opioid prescribing should be coupled with prescriber education and close follow-up to ensure commensurate reductions in opioid prescribing.

Importance  In 2014, the US Drug Enforcement Administration moved hydrocodone-containing analgesics from schedule III to the more restrictive schedule II to limit prescribing and decrease nonmedical opioid use. The association of this policy change with postoperative prescribing is not well understood.

Objective  To examine the hypothesis that the amount of opioids prescribed following surgery is associated with the rescheduling of hydrocodone.

Design, Setting, and Participants  An interrupted time series analysis of outpatient opioid prescriptions was conducted to examine the trends in the amount of postoperative opioids filled before and after the schedule change. Opioid prescriptions filled between January 2012 and October 2015 were analyzed using insurance claims data from the Michigan Value Collaborative, which includes data from 75 hospitals across Michigan. A total of 21 955 adult inpatients 18 to 64 years of age who underwent 1 of 19 common elective surgical procedures and filled an opioid prescription within 14 days of discharge to home were eligible for inclusion.

Main Outcomes and Measures  The primary outcome was the trends in the mean amount of opioids filled in oral morphine equivalents (OMEs) for the initial postoperative prescriptions before and after the schedule change date of October 6, 2014, compared using interrupted time series and multivariable regression analyses. Secondary outcomes included the total amount of opioids filled and the refill rate for the 30-day postoperative period. Subgroup analyses were performed by hydrocodone prescriptions, nonhydrocodone prescriptions, surgical procedure, and prior opioid use.

Results  Data from 21 955 patients undergoing surgical procedures across 75 hospitals and 5120 prescribers were analyzed. Cohorts before and after the schedule change were equivalent with respect to sex (10 197 of 15 791 [64.6%] vs 3966 of 6169 [64.3%] female; P = .69) and mean (SE) age (47.9 [11.2] vs 47.7 [11.3] years; P = .19). After the schedule change, the mean OMEs filled in the initial opioid prescription increased by approximately 35 OMEs (β = 35.1 [13.2]; P < .01), equivalent to 7 tablets of hydrocodone (5 mg). There were no significant differences in the total OMEs filled during the 30-day postoperative period before and after the schedule change (β = 18.3 [30.5]; P = .55), but there was a significant decrease in the refill rate (β = −5.2% [1.3%]; P < .001).

Conclusions and Relevance  Changing hydrocodone from schedule III to schedule II was associated with an increase in the amount of opioids filled in the initial prescription following surgery. Opioid-related policies require close follow-up to identify and address early unintended effects given the multitude of competing factors that influence health care professional prescribing behaviors.

Each day, more than 115 Americans die of an opioid-related overdose.¹ Prescription opioids sold to pharmacies, hospitals, and physicians have increased sharply from 1999 to 2010, and deaths due to prescription opioids have quadrupled.² Today, 2.6 million individuals in the United States live with opioid use disorders, including dependence, misuse, and addiction.^3,4 The opioid epidemic has cost the US health care system more than $1 trillion since 2001 and is projected to exceed another $500 billion in the next 3 years.⁵ Although substantial effort has been directed toward the treatment of opioid use disorders, effective prevention is critical to reduce the growing incidence of opioid-associated morbidity, mortality, and costs.

In response, there is strong interest in creating policies that mitigate the gravity of this public health crisis. In October 2014, the US Drug Enforcement Administration changed hydrocodone-containing products from schedule III of the Controlled Substances Act to schedule II. After this policy change, commonly prescribed formulations of hydrocodone were limited to a 90-day supply and could no longer be prescribed by telephone or facsimile transmission.⁶ Recent studies suggest that this policy shift resulted in a reduction of overall hydrocodone fills.^7,8 For example, Jones et al⁷ identified a 38% decrease in the number of opioid prescriptions and pills dispensed associated with surgical care following the schedule change of hydrocodone. However, postoperative opioid prescribing is also influenced by concerns for adequately treating pain, minimizing the need for additional refills, and optimizing patient experience and satisfaction.^9-11 Studies have found that roughly 80% of opioids prescribed after surgery remain unused and are provided in excess.^12,13 In addition, approximately 20% of patients undergoing surgery have recently been exposed to opioids and frequently require greater doses and refills following surgery.^14,15 Moreover, between 5% and 10% of opioid-naive patients persistently use opioids after surgery.^16-18 Given these factors, it is important to understand the effects of a policy that restricts opioid prescribing but may not be tailored to surgical care, to ensure safe prescribing following surgery.

In the present retrospective study analyzing data from Michigan, we sought to describe differences following the schedule change for hydrocodone based on the amount of postoperative opioids filled in the initial prescription following surgery and in the 30-day postoperative period. We also examined differences in postoperative opioid refill rates and the association of the schedule change with surgical procedure type and prior opioid exposure. We hypothesized that the initial amount of opioids filled following surgery would be altered following the schedule change given the absence of standardized guidelines to direct postoperative pain management and the concern for the need of additional refills to effectively manage pain.

We examined insurance claims from the Michigan Value Collaborative, a quality improvement collaborative funded by Blue Cross Blue Shield of Michigan that aggregates preferred provider organization insurance data from approximately 50% of Michigan’s commercially insured individuals.¹⁹ The present study was approved by the University of Michigan, Ann Arbor, institutional review board, which also waived the need for informed patient consent.

We included patients 18 to 64 years of age who underwent inpatient surgical procedures between January 1, 2012, and October 31, 2015. Using Current Procedural Terminology codes, we identified patients who underwent the following surgical procedures: general (open or laparoscopic appendectomy, open or laparoscopic cholecystectomy, or open or laparoscopic noncancer colectomy), bariatric (Roux-en-Y gastric bypass or sleeve gastrectomy), oncologic (open or thoracoscopic lung cancer resection, colorectal cancer resection, esophagectomy, pancreatectomy, or gastrectomy), cardiac (coronary artery bypass graft or valve surgery), vascular (carotid endarterectomy), orthopedic (knee replacement or hip replacement), gynecologic (hysterectomy), spine (disk herniation or other spine), and urological (prostatectomy) (eTable 1 in the Supplement). We excluded patients with a length of stay longer than 30 days or patients who were not discharged home or who did not fill an opioid prescription within 2 weeks of hospital discharge.

Demographic and clinical characteristics were obtained from claims, including sex, age, surgery type, preoperative opioid use, and comorbidities. We categorized prior opioid use according to preoperative opioid fills: naive (no fills between 1 year and 31 days prior to surgery), intermittent (filled <120 days’ supply between 1 year and 31 days prior to surgery during nonconsecutive months), and long term (≥3 fills in the 3 consecutive months prior to surgery or ≥120 days’ supply between 1 year and 31 days prior to surgery). Comorbid conditions were identified using the Centers for Medicare and Medicaid Services Hierarchical Condition Categories, which includes 79 conditions commonly used in risk-adjustment algorithms.²⁰

Our primary outcome was the change in the mean amount of opioids initially filled following surgery in association with the date of the schedule change. The initial prescription was defined as the first prescription claim for an opioid within 2 weeks following the index surgical procedure. We identified postoperative opioid prescriptions from pharmacy claims using National Drug Codes linked to generic drug names. Prescription information was converted to oral morphine equivalents (OMEs), and the OMEs for each prescription were multiplied by the total quantity filled to compute the OMEs per prescription.²¹ If more than 1 claim was filed on the same date, the OMEs were summed to represent the initial OMEs for the surgical case. Prescriptions were grouped by month and year, with each group mean serving as a data point. Secondary outcomes included the proportion of surgical cases for which additional opioid prescription claims were made following the initial postoperative prescription and the total OMEs filled in the 30 days following surgery.

We used χ² tests to compare patient attributes before and after the schedule change. Interrupted time series models were constructed for all opioid prescriptions and then hydrocodone prescriptions and nonhydrocodone prescriptions to assess trends in opioid fills before and after the schedule change. Data for time series regressions typically violate the assumptions of an ordinary least squares regression, in particular the assumption that the data are distributed evenly across time. For all models, Durbin-Watson statistics were used to evaluate autocorrelation. The presence of autocorrelation in the hydrocodone data required the use of autoregressive parameters. The absence of autocorrelation in the other 2 models indicated no clustering of the data across time, and an ordinary least squares regression or autoregression was used to describe changes in OMEs filled across time. Similar models were constructed for the rate of refill following the initial prescription and the total OMEs filled in the 30 days following surgery. To examine the association of the schedule change with surgical procedure type and preoperative opioid use, we constructed a multivariable hierarchical linear regression model to estimate the association of the schedule change with the amount of opioid provided (in OMEs) in the initial prescription following surgery, alongside patient-level factors (preoperative opioid use, surgical procedure, age, sex, and comorbid conditions), adjusting for the effect of patient clustering within surgeons and hospitals. Prescription amount was log transformed to ensure normally distributed residuals in the linear regression model. All analyses were conducted using SAS, version 9.4 (SAS Institute Inc), and 2-sided P < .05 was considered statistically significant.

In total, 21 955 patients undergoing surgical procedures across 75 hospitals and 5120 prescribers were examined. From 2012 to 2015 in Michigan, hydrocodone-containing medications were the most frequent opioids filled after surgery (72.2%), followed by oxycodone (18.1%) and tramadol (2.1%). We observed modest differences between time periods with respect to surgery type, patient characteristics, and comorbidities (Table 1). For example, the proportion of patients undergoing bariatric surgery increased significantly between the study time periods (2.7% vs 7.5%), and there were more patients with long-term opioid use (9.1% vs 8.4%) who filled opioid prescriptions following the schedule change, although cohorts before and after the schedule change were equivalent with respect to sex (64.6% vs 64.3% female; P = .69) and mean (SE) age (47.9 [11.2] vs 47.7 [11.3] years; P = .19).

Table 1 also describes the unadjusted differences in the mean OMEs filled for the initial prescription following surgery by patient characteristics and surgery type. We observed increases in OMEs provided in the initial prescription following the schedule change among patients undergoing general surgical (mean difference, 28.0 OMEs; P < .001), cancer surgery (mean difference, 9.9 OMEs; P < .004), cardiac surgery (mean difference, 123.3 OMEs; P < .001), orthopedic surgery (mean difference, 20.5 OMEs; P < .001), spine surgery (mean difference, 66.0 OMEs; P < .001), and vascular surgery (mean difference, 47.0 OMEs; P = .02). In addition, we observed increases in initial postoperative opioid amounts among opioid-naive patients (mean difference, 16.8 OMEs; P < .001) and intermittent opioid users (mean difference, 60.3 OMEs; P < .001), but not among long-term opioid users (mean difference, −6.6; P < .07). Finally, we observed greater increases among male patients (mean difference, 45.8; P < .001) and among patients with metabolic (mean difference, 14.1; P = .006), cardiorespiratory (mean difference, 21.7; P < .001), neurologic (mean difference, 46.2; P = .02), oncologic (mean difference, 8.9; P = .03), trauma (mean difference, 65.8; P = .002), and substance abuse (mean difference, 4.2; P = .005) comorbid conditions.

An interrupted time series analysis indicated an increase of 35 OMEs per initial opioid postoperative prescription (P < .011) immediately after October 6, 2014 (Figure 1). At the beginning of the study period, the mean (SE) OME provided in the initial prescription following surgery was 371.36 (6.89) and decreased prior to the schedule change by a mean (SE) of 1.15 (0.35) OMEs per month (Table 2). Immediately following the schedule change, we observed an increase in the mean (SE) OMEs of 35.11 (13.15) for the initial postoperative prescription, equivalent to 7 additional tablets of hydrocodone (5 mg) per prescription. Although there was a mean (SE) monthly increase of 2.19 (1.48) OMEs per prescription in the months following the schedule change, this increase was not significant (P = .15).

When stratified by type of opioid prescription, we observed an increase of 30 OMEs per prescription for hydrocodone prescriptions but not for other opioid types (Figure 2). The mean (SE) hydrocodone OME provided in the initial prescription following surgery was 273.30 (3.74) in January 2012, and this increased prior to the schedule change by 0.57 (0.20) OMEs per month (Table 2). Immediately following the schedule change, we observed an increase in mean (SE) OMEs of 29.98 (9.13) for the initial postoperative prescription for hydrocodone-based opioids, which did not decrease significantly following the schedule change.

We then examined differences in refill rates and the total OMEs filled in the 30-day period following surgery, stratified by preoperative opioid exposure. For all patients, there were no significant differences in the total OMEs filled during the 30-day postoperative period before and after the schedule change (Table 2). We did observe a decrease in refill rates across all patients and opioid-naive patients following the schedule change. In January 2012, the mean (SE) refill rates were 31.26% (0.67%) for all patients and 23.26% (0.98%) for opioid naive patients (Table 2). Following the schedule change, the refill rate decreased by 5.19% (SE, 1.29%; P < .001) for all patients by 4.35% (SE, 1.88%; P < .02) for opioid naive patients, which persisted throughout the study period. We did not observe a difference in the OMEs provided in the refill prior to and after the schedule change.

Finally, we examined differences in the mean OMEs filled for the initial prescription by preoperative opioid exposure and surgical procedure, adjusted for patient age, sex, and comorbid conditions. In multivariable analysis, we observed an increase in OMEs from 263.8 to 284.9 OMEs (P < .001) (Figure 3; eTable 2 in the Supplement). Significant increases in OMEs were observed among patients undergoing spine surgery (474.3 vs 547.8; P < .001), general surgery (188.8 vs 209.5; P < .001), cancer surgery (221.1 vs 259.7; P < .01), cardiac surgery (311.4 vs 397.8; P < .001), and orthopedic surgery (458.9 vs 519.6; P < .001). We also observed significant increases among opioid-naive patients (250.4 vs 272.3; P < .001), intermittent opioid users (265.7 vs 285.4; P < .001), and long-term opioid users (369.6 vs 382.4; P < .008).

In this retrospective study, recent policy intended to curb opioid prescribing through the restriction of hydrocodone was associated with an increase in the amount of opioid filled after elective surgery in Michigan. Most of this increase was accounted for by a rise in hydrocodone, which was the most commonly prescribed opioid following surgery. The largest increases were noted among opioid-naïve and intermittent users as well as among patients undergoing cancer, cardiac, spine, orthopedic, and general surgical procedures. We observed significant declines in refill rates following the schedule change, but no difference in the overall amount of opioid filled in the 30-day postoperative period. Taken together, these findings suggest that the change in policy was associated with a curbing of refill rates but was not associated with a reduction in the overall amount filled in the 30-day postoperative period and was associated with an unintended increase in the amount of opioid prescribed in the initial postoperative opioid prescription across many procedures.

In contrast to the findings of the present study, prior studies have shown that hydrocodone fills have declined following the schedule change for hydrocodone across many types of care. For example, Tran et al⁸ observed declines in hydrocodone pharmaceutical claims among Medicaid patients across all types of care following the hydrocodone schedule change. In addition, Jones et al⁷ examined national trends in opioid fills following the schedule change for hydrocodone and observed a significant drop in the amount of hydrocodone dispensed following the schedule change. Moreover, the largest declines were observed among prescriptions associated with surgical care. Our findings build on this body of work by further examining initial and subsequent opioid fills at the patient level linked to episodes of surgical care and the potential risk factors for ongoing postoperative opioid use. Our analysis examined differences based on OMEs, rather than on number of pills prescribed, because prescribers may be inclined to select a stronger opioid rather than to provide more pills. In contrast to prior work, we observed an increase in the amount of opioid provided in the initial prescription following surgery but no change in the total amount of opioid filled in the 30-day postoperative period, despite declines in refill rates. It is plausible that policy that is broadly targeted to all types of care may not entirely address the nuances of surgical prescribing. Postoperative opioid prescribing varies widely and is often in excess given the historical lack of evidence-based prescribing guidelines.^12,13,22,23 In addition, for decades, opioids were thought to have no significant potential for addiction when used for acute postoperative pain.^24-29 Guidelines have been established in the setting of primary care for the management of chronic pain to decrease opioid dependence and curb opioid use disorders.²¹ However, opioids prescribed following surgery are provided for acute pain, which is expected to resolve quickly for most surgeries. Postoperative opioid prescribing has often been driven by concerns regarding the need for refills and ensuring patient satisfaction, despite recent studies suggesting opioid prescribing is not linked to the probability of requests for prescription refills nor to measures of patient satisfaction.^23,30-32

It is possible that, in the absence of clear prescribing guidelines for postoperative care, restricting opioid prescribing may inadvertently motivate surgeons to prescribe greater amounts to ensure adequate pain treatment. The schedule change for hydrocodone specifically restricted the mechanism for hydrocodone refill and likely created new barriers for patients to receive postoperative refills when pain is undertreated. Prescribers may offer more hydrocodone in the initial outpatient prescription to prevent the need for additional clinic or emergency department visits, especially given the low uptake of electronic prescribing of controlled substances by surgeons and the inability to call in refills under the schedule change.³³ Our findings may also have important implications for other prescribing policies. Recent measures have focused on restricting opioid prescribing by the number of days an opioid is supplied per prescription. However, for acute pain, opioid prescriptions are typically provided as needed, rather than on a scheduled basis. Therefore, current 5- and 7-day prescribing limits would still allow for overprescribing (eg, 2 tablets as needed every 4 hours for pain for 7 days would come to 84 tablets) given recent studies indicating that for many elective procedures 15 tablets or fewer is sufficient.^12,13,23 Our findings highlight the need to couple policy alongside robust evidence regarding average consumption and the prescriber factors that drive opioid prescribing.

A review of unintended consequences of health-related policy concluded that responsibility at the local level is more effective than a “one size fits all” national approach.³⁴ Furthermore, policy supported by data and informed by small pilot programs can mitigate unintended consequences. Going forward, tailoring policy toward the unique aspects of care that influence prescribing and pain management could enhance the effectiveness of measures to improve the safety of opioid prescribing. Pairing policy measures with specialty-specific health care professional–level education and patient-level education could also improve the potential success of future regulatory measures, particularly given the differences in the context of opioid prescribing for surgery compared with other episodes of care. Moreover, we identified significant differences in prescribing by procedure type. Although normative opioid consumption is not yet defined for many procedures, and the risk of new long-term use is similar across procedure types, prescribing behaviors may be influenced by other important factors, such as the prevalence of chronic pain or long-term opioid use in patient populations, the availability of standardized pain management protocols, and the acceptability of opioid alternatives. Future studies that elicit prescriber preferences, knowledge, and expectations regarding pain management and attitudes toward opioid analgesics could provide a deeper understanding of how prescribers may respond to policy measures to improve the safety of opioid prescribing. Finally, electronic prescribing has been proposed to improve appropriate opioid prescribing, decrease fraud, and identify abuse. Although uptake remains low (15% in Michigan), such measures can facilitate regular reporting and inform the development of evidence-based guidelines.³³

Our study has several limitations. First, our data were drawn from a single state; thus, our observations may not be generalizable to other regions. We examined only adults with employer-based insurance and their dependents, and the associations with the policy change could be different among individuals insured by other payers or the uninsured and among those in other health systems, such as Veterans Administration medical centers. We also observed differences between each cohort with respect to surgery type and prior opioid use. Although these differences were small, they may have contributed to the amount of opioid prescribed. In addition, the proportion of long-term opioid users increased between study periods, underscoring the prevalence of long-term opioid use in Michigan. Our study may also lack sufficient power to determine differences in long-term trends following the schedule change given the study period. Within claims data, we do not have measures of patient-reported pain or opioid consumption, and the effect of this policy from the patient perspective cannot be discerned from this analysis. Finally, our analysis evaluates only population-level effects, and further research is necessary to understand more granular factors that influence surgical prescribing, such as surgeon or patient preferences, experiences, or financial aspects.

Tighter restrictions on hydrocodone prescribing was associated with an immediate increase in the amount of opioid filled following surgery but no change in the overall opioid amount filled in the 30-day postoperative period. Opioid-related policy is likely influenced by clinical context, and identifying the factors that motivate prescriber behaviors will be important to curb opioid prescribing and encourage opioid alternatives for surgical care.

Accepted for Publication: May 13, 2018.

Published Online: August 22, 2018. doi:10.1001/jamasurg.2018.2651

Correction: This article was corrected on December 19, 2018, to fix an error in Figure 2.

Corresponding Author: Jennifer Waljee, MD, MPH, Michigan Opioid Prescribing Engagement Network, 2800 Plymouth Rd, North Campus Research Complex, Bldg 16, Ann Arbor, MI 48109 (filip@med.umich.edu).

Author Contributions: Dr Waljee had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Habbouche, Lee, Khalsa, Englesbe, Brummett, Waljee.

Acquisition, analysis, or interpretation of data: Habbouche, Steiger, Dupree, Waljee.

Drafting of the manuscript: Habbouche, Steiger, Khalsa, Brummett, Waljee.

Critical revision of the manuscript for important intellectual content: Habbouche, Lee, Steiger, Dupree, Englesbe, Brummett, Waljee.

Statistical analysis: Steiger.

Obtained funding: Dupree, Brummett, Waljee.

Administrative, technical, or material support: Habbouche, Dupree, Englesbe, Brummett, Waljee.

Supervision: Lee, Dupree, Englesbe, Brummett, Waljee.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported by a Research Project Grant (R01DA042859) from the National Institute on Drug Abuse to Drs Dupree, Englesbe, Brummett, and Waljee.

Role of the Funder/Sponsor: The National Institute on Drug Abuse 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; or the decision to submit the manuscript for publication.