Analgesic methadone prescribing in community health centers among patients with chronic pain.

Electronic health record (EHR) data were extracted from 185 Oregon and 50 California primary care CHC clinics on OCHIN’s (not an acronym) network. OCHIN is a nonprofit health information technology organization that provides a single, linked instance of the Epic ® EHR across every clinic in the network. 6 Only clinics with an active primary care department in both periods were included. The Oregon Health & Science University’s Institutional Review Board approved this study. We included patients aged ≥18 years with ≥1 ambulatory visit to a study clinic in 2012–2013 (N = 45,424) or 2017–2018 (N = 112,815), and ≥1 of the following chronic pain conditions identified via International Classification of Diseases, 9th and 10th revisions (ICD-9/10), codes: back pain, chronic tension headache, migraine headache, joint pain, neuropathic pain, chronic fatigue syndrome, endometriosis, fibromyalgia, temporomandibular disorders, urologic chronic pelvic pain syndrome, and vulvodynia. We excluded patients with a cancer diagnosis and those with any record of liquid methadone, which is primarily used for opioid use disorder (OUD) treatment. Patients were randomly assigned to a single period if eligible for both (N = 33,281).

Our study sample included 158,239 patients; most were female (64 percent), middle-aged (38 percent between 30 and 49 years of age), non-Hispanic White (68 percent), and Medicaid insured (34 percent) ( Table 1 ). Opioid prescribing among patients with chronic pain decreased markedly (aRR = 0.51; 95 percent confidence interval (CI): 0.47–0.56) among all opioid categories over the time periods. The largest decrease was for methadone prescriptions (aRR = 0.19, 95 percent CI: 0.14–0.27) ( Table 2 ). Of patients prescribed any long-acting opioid, 24 percent (n = 405/1,683 patients; data not shown) received a methadone prescription in 2017–2018 compared with 34 percent (843/2,466) in 2012–2013. In 2012–2013, the characteristics associated with being prescribed methadone rather than an alternate long-acting opioid were having a prior OUD diagnosis (aRR = 1.67, 95 percent CI: 1.47–1.90) and being publicly insured (aRR = 1.70, 95 percent CI: 1.13–2.54 for Medicaid and aRR = 1.56, 95 percent CI: 1.04–2.27 for Medicare) or uninsured (aRR = 1.93, 95 percent CI: 1.22–3.06) vs privately insured. Female sex (aRR = 0.87, 95 percent CI: 0.77–0.98) and older age were associated with lower methadone prescribing rates, with those ≥65 years of age having significantly lower rates than those in all other age groups ( Table 3 ). Several factors continued to be associated with methadone prescribing in 2017–2018, including being aged 30–49 (aRR = 1.69, 95 percent CI: 1.28–2.23) or 50–64 (aRR = 1.43, 95 percent CI: 1.09–1.88), having Medicare vs private insurance (aRR = 1.37, 95 percent CI: 1.10–1.71), and having a prior OUD diagnosis (aRR = 2.10, 95 percent CI: 1.38–3.20). Male sex was no longer significantly associated with receiving a methadone prescription. In 2017–2018 only, patients seen in large rural clinics (aRR = 0.83, 95 percent CI: 0.70–0.99) and of a race/ethnicity other than non-Hispanic Black (vs non-Hispanic White) (aRR = 0.62, 95 percent CI: 0.42–0.94) had lower rates of having methadone prescribed.

We examined sex, age, race/ethnicity, insurance type at the majority of visits during each study period, OUD diagnosis prior to the study period, and the state (California and Oregon) and the rurality/urbanicity 7 of the patient’s primary clinic. We categorized opioid prescriptions as follows: none (no opioid prescriptions), short-acting opioids only (no long-acting prescription), long-acting opioids except methadone (long-acting but no methadone prescription), and methadone. Long-acting opioids included extended-release formulations of two US Federal Food & Drug Administration-approved buprenorphine formulations for pain (buccal film and transdermal form), 8 fentanyl patches, hydrocodone, morphine, oxycodone, oxymorphone, tapentadol, and tramadol. We described the percentage of patients in each opioid prescription category in both time periods, overall and by patient characteristics. Data were analyzed in SAS Enterprise GuideVersion 8.3. Poisson’s generalized estimating equations (GEEs) 9 estimated the adjusted relative rate (aRR) to reflect the relative differences in the likelihood of receiving an analgesic methadone prescription overall, comparing 2017–2018 to 2012–2013, and across different patient characteristics. Models were adjusted for correlation of patients within health systems using a sandwich estimator. Separate models were run for each prescription group outcome. We analyzed patient characteristics associated with receiving ≥1 methadone prescription versus ≥1 other long-acting opioid prescription for each time period using the same GEE approach, with separate models for each time period.

While prior studies note decreased opioid prescribing in general 10 , 11 and methadone prescribing specifically over time, 5 , 12 , 13 this study examined changes in both overall opioid and methadone prescribing rates within CHCs that provide care for a large proportion of high-risk patients. Methadone represented a small percentage of opioid prescriptions, and the proportion of patients who received methadone declined significantly between the two time periods. Our findings provide evidence that CHC clinicians in our study clinics are following policy guidelines to both decrease opioid prescriptions in general and methadone prescribing in particular for chronic pain. We found that certain characteristics associated with being prescribed methadone rather than other long-acting opioids did not change over the two periods. In both periods, those aged 30–64 years (vs ≥65), those with a prior OUD diagnosis (vs without), and those with Medicare (vs private insurance) were more likely to be prescribed analgesic methadone. When comparing the study periods, the elevated methadone prescribing seen in 2012–2013 among younger (vs ≥65) and nonprivately insured individuals was no longer uniformly apparent in 2017–2018. Higher prescribing only persisted for those aged 30–64 years and those with Medicare. In the later study period, other characteristics associated with methadone prescribing emerged, including lower rates among those of a race/ethnicity other than non-Hispanic Black (vs non-Hispanic White). The differences in prescribing among these subgroups are concerning, especially given that nonelderly adults and those who are non-Hispanic White have been shown to have higher rates of methadone-related overdose. 12 , 14 We are unaware of other studies that have examined the associations between patient characteristics and methadone prescribing for chronic pain over time. The characteristic most strongly associated with receiving a methadone prescription was a prior OUD diagnosis. This cannot be explained by assuming that methadone was prescribed for OUD treatment as dispensing of methadone for this purpose is limited to Substance Abuse and Mental Health Services Administration-certified Opioid Treatment Programs, and we excluded patients who had liquid methadone orders from our analysis. It is possible that patients with OUD might have experience using methadone for OUD treatment, and providers might be more comfortable prescribing methadone versus another long-acting opioid to this population. Given that this finding was significant in both study periods and increased over time, more research is needed to understand the etiology of this relationship. We included patients seen in Oregon and California (primarily Northern California) CHCs who were primarily publicly insured; thus, results might not generalize to other settings. The relatively small number of patients prescribed methadone required collapsing/removing some variable categories. Although we identified several characteristics associated with methadone prescribing in 2017–2018, the low number of patients prescribed long-acting opioids in this period limited our ability to evaluate changes in characteristics associated with prescribing over time. We do not know if ordered medications were filled or if methadone was used as first-line chronic pain medication, as we only have access to CHC prescriptions. Finally, we do not know if decreased prescribing was solely attributable to the updated CDC guidelines or if other potential initiatives/trainings led to this change.

The finding that prescribing of analgesic methadone among patients with chronic pain was low and decreased over time suggests that CHC clinicians are following policy guidelines related to methadone prescribing for pain. While some differences in methadone prescribing were reduced or eliminated over time, it is unclear why some subgroups of patients with chronic pain continue to be prescribed methadone at a higher rate than others despite known risks. If our findings are replicated in other studies, targeting interventions to decrease methadone prescribing among these populations might be warranted.