Abstract
BACKGROUND Primary care clinicians write 45% of all opioid prescriptions in the United States, but little is known about the characteristics of patients who receive them and the clinicians who prescribe opioids in primary care settings. Our study aimed to describe the patient and clinician characteristics and clinicians’ perspectives of chronic opioid prescribing in primary care.
METHODS Using a mixed methods approach, we completed an analysis of 2016 electronic health records from 21 primary care practices to identify patients who had received chronic opioids, which we defined as in receipt of an opioid prescription for at least 3 consecutive months. We compared those receiving chronic opioids with those not in terms of their demographics, prescribing clinician characteristics, and risk factors for opioid-related harms, as identified by the Centers for Disease Control and Prevention Guideline on Opioid Prescribing for Chronic Pain. We then interviewed 16 primary care clinicians about their perspectives on chronic opioid prescribing.
RESULTS Of 84,029 patients, 1.1% (902/84,929) received chronic opioid prescriptions. Characteristics associated with being prescribed chronic opioids include being female, being of black or African American race, and having risks for opioid-related harms, such as mental health diagnoses, substance use disorder, and concurrent benzodiazepine use. Clinicians report multiple difficulties in weaning patients from chronic opioids, including medical contraindications of nonopioid alternatives and difficulty justifying weaning by stable long-term patients.
CONCLUSION Although patients prescribed opioids in primary care have higher risks of opioid-related harms, clinicians report multiple barriers in deprescribing chronic opioids. Future studies should examine strategies to mitigate these harms and engage patients in shared decision making about their chronic opioid use.
INTRODUCTION
Drug overdose deaths involving opioids continue to rise in the United States, with 42,249 people fatally overdosing in 2016, a 27.9% increase from 2015.1 Although illicit substances such as heroin and illicitly manufactured fentanyl contribute significantly to the problem, prescribed opioids are involved in approximately 40% of opioid overdose deaths.2 The National Survey on Drug Use and Health estimated that 1.8 million people had a prescription pain reliever use disorder and 11.5 million misused prescription pain relievers in 2016.3 An estimated 215 million opioid prescriptions were dispensed by retail pharmacies in 2016, reaching a rate of 66.5 dispensed opioid prescriptions per 100 persons in the United States.4 Of these opioid prescriptions, 45% were written by primary care clinicians.5,6
Despite high rates of opioid prescribing, the majority of primary care clinicians receive little or no training during medical school and residency in prescribing opioids or managing substance use disorders.7–9 As a result, few primary care clinicians feel prepared to screen for, diagnose, and treat prescription medication misuse.10–14 In recent years, a growing number of resources have been released to assist clinicians with managing opioid prescribing.15–18 In 2016, the Centers for Disease Control and Prevention (CDC) released its Guideline for Prescribing Opioids for Chronic Pain, which includes recommendations on assessing risk factors for opioid-related harms when prescribing opioids.15 Despite these guidelines, there remains substantial variation in the practice of prescribing opioids.19–21
Although other studies have examined large-scale variations,22–25 little is known about the patient- and clinician-specific factors associated with any opioid and chronic opioid prescribing in primary care. The reliance in these studies on claims data limited the examination of patient, clinician, and practice characteristics. This limitation could be overcome by the use of electronic health record (EHR) data to examine patient and clinician characteristics that could account for risky opioid prescribing. Linking clinicians’ perspectives to EHR data on prescribing practices can demonstrate the complexities of chronic opioid prescribing in primary care practice.
METHODS
Study Design
Using a mixed methods approach, we first completed a secondary analysis of 2016 EHR data from primary care practices in the Virginia Ambulatory Care Outcomes Research Network (ACORN) to describe patient and clinician characteristics associated with opioid prescribing and then interviewed 16 primary care clinicians from these practices to obtain their perspectives on chronic opioid prescribing. The Virginia Commonwealth University Institutional Review Board approved this study.
Setting
The Virginia Ambulatory Care Outcomes Research Network is a practice-based research network consisting of more than 150 primary care practices across the Commonwealth of Virginia. Practices represent a broad spectrum of settings ranging from academic to group practices to solo clinicians and are located in rural, suburban, and urban locations. For this study, we selected 21 primary care practices to represent more affluent, suburban, and more disadvantaged, urban settings and included EHR data from all clinicians from these practices (271 clinicians).
Quantitative Data Collection
We extracted EHR data for patients seen from January 1, 2016 to December 31, 2016 in each of the included primary care clinics. Extracted data included patient demographics, diagnoses, and prescriptions. Each patient was linked with their primary care clinician and cross-matched to clinician and practice characteristics, as found in the ACORN database. We queried all prescription data to extract opioid and benzodiazepine prescriptions (including benzodiazepine because it was identified by the CDC as a risk factor for opioid-related harm). We excluded patients aged <18 years, with sickle cell disease, in palliative care, receiving buprenorphine for opioid use disorder, or with cancer.
Qualitative Data Collection
We conducted semistructured interviews with 16 primary care clinicians purposefully selected from the suburban and rural samples. Authors E.M.B. and V.J. conducted the interviews over the phone, which were audio recorded and lasted 30 to 40 minutes each. Using an interview guide, we asked clinicians about (1) their individual and clinic-based policies regarding appropriate opioid prescribing and treatment of chronic pain, (2) experiences with prescribing opioids to patients with chronic pain, (3) understanding and application of red flags for opioid prescribing, and (4) perceived challenges of weaning patients off chronic opioids.
Statistical Analysis
We calculated counts and frequencies of opioid medications prescribed by clinicians in 2016 to determine the most frequently prescribed opioids. We identified patients on chronic opioids based on the CDC definition of receipt of an opioid prescription for 3 consecutive months.15 Based on how medications were presented in the EHRs, we considered this to be 3 consecutive calendar months. We then used χ2 tests to determine associations between chronic opioid use and patient demographic characteristics and medical comorbidities that were identified by the CDC guideline as high risk for opioid-related harms.15 The risks include depression and anxiety, renal insufficiency, hepatic insufficiency, concurrent benzodiazepine use, substance use disorder, and sleep apnea. Next, we calculated the daily morphine milligram equivalent (MME) per patient on chronic opioids by multiplying the MME conversion factor (based on opioid formulation) by the pill count and then dividing by 30 (for the average number of days per month). We then determined MME per clinician with at least 1 patient on chronic opioids and per clinic serving at least 1 patient on chronic opioids as the total MME divided by the total number of prescriptions, then dividing by 30 ([total MME/total prescriptions]/30). Because of the right-skewed nature of MME, we applied the Kruskal-Wallis nonparametric rank test to compare the distribution of daily MME by patient, clinician, and practice characteristics. Resulting P-values, medians, and interquartile ranges were reported. We used SAS statistical software version 9.4 (SAS Institute).
Qualitative Data Analysis
Interview transcripts were analyzed via a combination of template and emergent coding processes.26 Template-based codes were derived from the interview guide, including themes related to the initiation, inheritance, and management of patients on opioids for chronic pain, and clinicians’ experiences attempting to wean patients off opioids. Authors S.T.T., E.M.B., and V.J. then read through the transcripts separately before meeting to discuss emergent themes and findings.
RESULTS
Overall Opioid Prescribing
Of 545,872 total prescriptions written in 2016 by a primary care clinician, 25,450 (4.7%) were opioid prescriptions. Of the 84,929 patients seen in 2016, 9,462 (11.1%) received an opioid prescription. Overall, oxycodone-acetaminophen was the most commonly prescribed opioid, followed by oxycodone. In urban, underserved clinics in our sample, 9.7% of prescriptions written were opioid prescriptions, and oxycodone, which was the most frequently prescribed opioid, was the 6th most commonly written prescription (Table 1). In suburban clinics in our sample, 3.0% of prescriptions were opioids, and the top prescribed opioid was codeine-guaifenesin, which was the 47th most commonly written prescription.
Chronic Opioid Prescribing
Of 84,929 patients, 902 (1.1%) patients received chronic opioid prescriptions in 2016. Women were more likely than men to receive chronic opioid prescriptions (65.5% of patients receiving chronic opioids and 60.2% of patients not receiving chronic opioids, P <.01). Furthermore, blacks or African Americans constituted 14.7% of those not on chronic opioids and 43.1% of those on chronic opioids seen in primary care (P <.01), whereas 57.4% of those not on chronic opioids and 48.2% of those on chronic opioids were white (Table 2). Of the risk factors identified by the CDC guideline for opioid-related harms, comorbidities such as sleep apnea, depression or anxiety, substance use disorder, hepatic insufficiency, renal insufficiency, and concurrent benzodiazepine use were associated with receiving chronic opioids (Table 2). The median MME per day also varied based on demographics and comorbidities of patients on chronic opioids. Patients with depression or anxiety, substance use disorder, hepatic insufficiency, or renal insufficiency were more likely than those without these comorbidities to be on a higher overall MME per day (Table 3). Those of black or African American and white race received similar MMEs per day (P = 0.14).
Clinicians who are MD or NP (vs DO or PA), female (vs male), internists (vs family physicians), and residents (vs attendings) were more likely to prescribe higher daily MMEs to patients on chronic opioids (Table 3). Although large differences in median MME were observed for clinic type and the number of clinicians at the clinic, none of the clinic-level characteristics were significantly associated with MME.
Clinician Perspectives
Clinician interviews revealed multiple factors informing the use of prescription opioids for chronic pain in primary care (Table 4). Clinicians largely saw the use of opioids to manage chronic pain as appropriate when caring for patients with extensive medical comorbidities or patients for whom nonopioid pain medications were contraindicated. Although all the clinicians we spoke to were aware of the CDC guidelines and risk factors for misuse, they maintained that the benefits of managing chronic pain with opioids at times outweighed the risks, particularly when it restored functional capacity and quality of life. However, despite the perceived benefits of opioids, most clinicians were reluctant to initiate patients on opioids for chronic pain. Instead, most clinicians reported having inherited the bulk of their patients with chronic pain from colleagues. Many felt frustrated at having to act as a pain specialist for patients taking large amounts of opioids that were sometimes in amounts or for conditions they thought were inappropriate. Frustrations were equally compounded by a perceived lack of time to appropriately manage patients’ chronic pain and a lack of control over patients’ access to other sources of opioids, such as in the hospital or from specialists.
DISCUSSION
Substantially higher rates of opioid prescribing were demonstrated in urban, underserved settings compared with suburban settings. Uniquely, our study found that being of black or African American race was associated with receipt of chronic opioid prescriptions. Previous studies examining opioid prescribing have demonstrated that white patients are more likely than black or African American patients to receive opioids in emergency department settings27 and overall.28,29 Our study, which focuses on primary care settings, suggests that chronic opioid prescribing may follow different racial trends. There may be other mitigating factors for opioid prescribing among patients who are on chronic opioids instead of opioids for acute purposes. Future studies are needed to elucidate why black and African American patients receive more chronic opioid prescriptions than white patients in primary care.
Our study also showed that primary care patients with higher comorbidities were more likely not only to receive chronic opioid prescriptions but also to receive these prescriptions at higher dosages (in MME per day). This finding is particularly concerning given the recent CDC guideline recommending caution in prescribing opioids for those with the comorbidities examined in our study. This result may not be unexpected because many patients with high-risk comorbidities may have contraindications to nonopioid medication alternatives. Nevertheless, patients with these comorbidities are at significantly higher risk of opioid-related harms, including overdose and death.
Although guidelines suggest engaging patients in shared decision making, increasing the availability of nonopioid treatment modalities, and increasing the availability of medication-assisted treatment for those with concurrent opioid use disorders,15,17 primary care clinicians identified multiple challenges in trying to reduce risk. These included contraindications to nonopioid treatment alternatives, lack of access to adjunctive management strategies, limited time, and the difficulty of weaning in patients on long-term chronic opioids. New interventions to help primary care clinicians overcome these barriers must be developed and tested in order for primary care clinicians to successfully wean patients from chronic opioids.
Our study has several limitations. First, our study examined prescribed, not filled, opioid prescriptions because we relied on EHR data. Second, our study examined data from only 21 clinics in Virginia, which limited the power to detect associations based on clinic characteristics. Although this sample included 84,929 primary care patients, it may not be generalizable to other primary care settings. Third, opioids can be prescribed only at monthly intervals without refills, whereas other medications may have refills that can last for up to 1 year. This difference may result in over-counting of discrete opioid prescriptions compared with other medications in our prescription data. It would not affect the comparisons between patients who received and prescribers who wrote for opioids, however.
Chronic opioid prescribing in primary care varies significantly by patient and clinician characteristics. Particularly, there is a higher likelihood for blacks or African Americans, women, and those with CDC-identified comorbidities to receive chronic opioid prescriptions in primary care settings. Although primary care clinicians realize the importance of limiting chronic opioid prescribing, multiple barriers exist in weaning patients off chronic opioids. Future studies should elucidate why specific trends in opioids prescribing exist, compare the differences in opioid prescribing in various settings, and explore possible interventions to help primary care clinicians overcome barriers in weaning patients with high risks of opioid-related harms.
Footnotes
Conflicts of interest: authors report none.
To read or post commentaries in response to this article, see it online at http://www.AnnFamMed.org/content/17/3/200.
Funding support: American Academy of Family Physicians Foundation (G1604JG).
Prior presentations: American Society of Addiction Medicine Annual Conference; April 12-15, 2018; San Diego, CA; and North American Primary Care Research Group PBRN Conference; June 22-23, 2017; Bethesda, MD.
- Received for publication July 27, 2018.
- Revision received October 30, 2019.
- Accepted for publication November 30, 2018.
- © 2019 Annals of Family Medicine, Inc.