Three-Year Retention in Buprenorphine Treatment for Opioid Use Disorder Among Privately Insured Adults

As the “epidemic” of opioid use disorder (OUD) sweeps through the United States, causing death and disability, enrollment and retention of people with OUD in opioid agonist treatment (OAT) have emerged as an important clinical objective (1,2). OAT retention is associated with substantial improvements in medical morbidity and social functioning as well as reduced HIV transmission and criminal behavior (3–6) and mortality (7), such that OAT with either buprenorphine or methadone has become the standard medical treatment for OUD (4,8). Although only a small proportion of eligible patients with OUD are engaged in either OAT or other treatments (9), OAT utilization has steadily increased over the past decade, mainly driven by the increased accessibility to buprenorphine that can be provided in less restrictive office-based practice, unlike methadone maintenance (6,10–15). On the other hand, it has been suggested that long-term retention may be lower in buprenorphine treatment compared with methadone treatment (16,17), which poses a significant clinical challenge because treatment retention is crucial for therapeutic success (7,18).

There are limited data on long-term retention in buprenorphine treatment for OUD beyond six and 12 months (17,19), and the influence of diverse sociodemographic, clinical, and service use characteristics on OAT retention has been examined in few large studies (20). A recent national study based on data from the Veterans Health Administration (VHA), an integrated health system caring for U.S. veterans at little or no copayment expense, showed 61% retention a year after buprenorphine initiation among mostly male veterans with OUD and high rates of general medical and psychiatric comorbidity (21). Among numerous sociodemographic and diagnostic characteristics that were evaluated, only black race had a discernible (and negative) independent association with buprenorphine treatment retention.

Data on buprenorphine treatment retention among privately insured patients are far more limited. Predictors of buprenorphine treatment retention among privately insured patients with OUD may be quite different from those in the VHA population because privately insured patients tend to be much younger and to have higher income, lower rates of disability, and fewer psychiatric comorbidities (22,23). Recent insurance claims–based studies have suggested that buprenorphine treatment retention rates may be much lower in privately insured populations than in the VHA population, but these studies lacked detailed data on clinical correlates of better retention (24–27). Most important, the impact of insurance disenrollment on treatment retention has yet to be examined.

In this observational study, we used claims data from the MarketScan Commercial Claims and Encounters Database (IBM Watson Health) documenting sociodemographic and clinical characteristics, health service utilization, and prescription drug claims from a selection of insured individuals of large employers and private health plans. Among patients newly started on buprenorphine for OUD in federal fiscal year 2011 (FY 2011) and followed up through FY 2014, we examined the duration of buprenorphine treatment up to and beyond three years along with correlates, including sociodemographic and diagnostic characteristics, health service and psychotropic medication use, and insurance disenrollment. These data thus broaden the picture of the challenges facing efforts to increase buprenorphine treatment retention among patients with OUD.

Methods

Sample and Data Source

The MarketScan database includes claims from commercial insurance companies representing insured employees and their dependents, early retirees, individuals with coverage through the Consolidated Omnibus Budget Reconciliation Act (COBRA), and Medicare-eligible retirees with employer-provided Medicare supplemental plans. For comparability, we followed the methods of an earlier published study of buprenorphine treatment retention in the VHA (21). We first selected patients with a diagnosis of OUD (ICD-9-CM codes 304.0x, 305.5x, and 304.7x—either opioid abuse or opioid dependence) from October 1, 2010, to September 30, 2011 (FY 2011) in the MarketScan database. We then excluded patients who filled at least one prescription for buprenorphine during the first 60 days of FY 2011, thus leaving a sample of patients with OUD who were not receiving buprenorphine at the beginning of FY 2011. We included only patients who filled prescriptions for buprenorphine or buprenorphine-naloxone tablets and excluded those receiving a buprenorphine transdermal patch indicated for pain management. Because OUD is considered a relapsing chronic disease requiring long-term treatment, we elected to consider both uninterrupted and interrupted participation in treatment as continued treatment engagement.

Measures

We calculated the total duration of treatment as the number of days between the first day of buprenorphine fill in FY 2011 and the last date buprenorphine was filled through the end of FY 2014 (September 30, 2014). We defined four mutually exclusive groups on the basis of the duration of retention in treatment: 0–30 days, 31–365 days, one to three years, and more than three years, consistent with prior studies (28–33).

Sociodemographic characteristics included age, gender, and urban residence based on metropolitan statistical area codes. Race-ethnicity data were not available. Psychiatric and general medical diagnoses were identified by using ICD-9 codes that were assigned to each patient at least once during baseline year FY 2011. We calculated the Charlson Index (34), an aggregate measure of medical comorbidity that has been shown to predict ten-year mortality (35). Patients were coded as having “any pain” if they had ICD-9 codes indicating herpetic pain (053.12 and 729.2), fibromyalgia pain (729.1), musculoskeletal pain (338.xx, 719.4, and 780.96), muscolospasm pain (728.85 and 781.0), pain from diabetes (250.6, 337.1, and 357.2), migraine and headache (346.x and 784.0), and other sources of pain (350.1, 352.1, 357.2, 729.2, and 781.0).

Psychiatric disorders (ICD-9 codes 290.00 through 319.99) coded into 11 classes (Table 1) (35) were used to identify outpatient visits for mental and substance use disorders, including OUD, in the baseline year FY 2011. Inpatient and outpatient service utilization codes were used to identify medical and surgical outpatient and emergency room visits. Duration of insurance enrollment was calculated as the number of months between first buprenorphine prescription and last date of enrollment.

The receipt of any outpatient psychotherapy in the baseline year of FY 2011 was identified by the following CPT encounter codes: outpatient psychotherapy, 90804–90815 and 90845; family therapy, 90846, 90847, and 90849; and group therapy, 90853 and 90857).

From the prescription drug claims, we identified psychiatric prescriptions filled in FY 2011 and divided them into five classes: antidepressants, antipsychotics, sedatives-hypnotics-anxiolytics, mood stabilizers (antiepileptics), and lithium. Then we summed the numbers of prescriptions in all of these classes to get the total number of psychotropic prescriptions filled by each patient (35).

Analysis

In the initial bivariate analysis, sociodemographic and diagnostic characteristics, health service use, psychotropic medication fills, and psychotherapy use during the baseline year of FY 2011 were compared between four groups: patients who received buprenorphine for 0–30 days or less (reference group) and those who received buprenorphine for 31–365 days, one to three years, and more than three years. To identify meaningful group differences, we used effect sizes rather than p values (that is, risk ratios for dichotomous measures and Cohen’s d for continuous measures), because the sample was large and thus very small differences would be likely to be statistically significant but not of meaningfully different magnitude. Risk ratios >1.5 or <.67 were considered to represent substantial differences on dichotomous variables. The difference between means divided by the pooled standard deviation was used to calculate Cohen’s d for continuous variables. Values >.20 were considered to represent more than small differences (36). Variables were included in survival analyses only if they met these effect size criteria and if the base rate was more than 5%.

We first included variables found to demonstrate substantial differences between groups in bivariate analyses in subsequent multivariate analyses. We used Kaplan-Meier curves and Cox proportional hazards models to identify variables independently associated with the discontinuation of buprenorphine treatment. We repeated the analyses including a variable representing months of insurance enrollment.

Results

We identified 61,447 patients with an OUD diagnosis who did not receive a buprenorphine prescription during the first 60 days of FY 2011, of whom 16,190 (26.3%) filled at least one buprenorphine prescription later in FY 2011. Among those initiated on buprenorphine, the proportion who remained engaged in treatment steadily declined with time, with 85% engaged for 31–365 days, 45.0% for one to three years, and only 13.7% for more than three years (mean±SD duration of retention=1.23±1.16 years).

FY 2011 Characteristics and Subsequent Retention

Mean age was higher in the group with retention for more than three years, compared with the group engaged for 30 days or less. No substantial differences between groups were found in gender, urban residence, medical comorbidities, or pain diagnosis. Medical comorbidity burden as indicated by Charlson Index was low, reflecting the young age of the study sample (Table 1).

Substance use disorder diagnoses other than OUD were generally less frequent among the groups with longer retention, and especially uncommon in the group with retention for more than three years (Table 1). Other psychiatric diagnoses were also lower among those retained for more than three years (Table 1).

Emergency room visits were lower in the group with retention for more than three years, compared with group engaged for 30 days or less (Table 2). All outpatient visits and any psychiatric or substance use disorder outpatient visits did not vary much between the different retention groups. Over two-thirds (67.3%) of the entire sample received some psychotropic medications, with no substantial differences across groups (Table 2). About a third of all patients received some form of psychotherapy in FY 2011, with higher-retention groups receiving outpatient psychotherapy at a slightly higher rate and number of episodes.

No statistically significant differences in the mean dose of buprenorphine were found between the four retention groups (Table 3).

Insurance Continuation

During the first three years after buprenorphine initiation, 49.3% (N=7,988) disenrolled from their insurance plan. The mean enrollment periods and disenrollment rates in each retention groups are shown in Table 4.

Survival Analysis and Multivariate Cox Proportional Hazards Analysis

In a multivariate Cox proportional hazards analysis (Table 5), being older than the median age of 31 years (hazard ratio [HR]=.82) and receiving any outpatient psychotherapy in FY 2011 (HR=.86) were strongly associated with lower risk of disengagement from buprenorphine treatment, whereas receiving any inpatient mental health treatment in FY 2011 was associated with increased risk of disengagement (HR=1.20). Having any emergency room visits or a diagnosis of any substance use disorder in addition to OUD or having any psychiatric diagnoses in FY2011 were also associated with elevated risk of disengagement.

When we added “each 30 days of continuous insurance enrollment” as a variable, the effect of FY 2011 predictors of treatment disengagement were largely unchanged. However, for each 30 days of additional enrollment in the insurance plan, there was a 10% lower risk of treatment disengagement (Table 5). [Kaplan-Meier survival curves of retention in buprenorphine treatment based on receipt of inpatient mental health admission and outpatient psychotherapy in FY 2011 are included in an online supplement to this article.]

Discussion

In this study of a large privately insured cohort of patients initiated on buprenorphine for OUD in FY 2011, fewer than half were retained in treatment for more than one year and fewer than 15% for more than three years. Nearly half of the patients who initiated buprenorphine treatment disenrolled from their insurance plan in the next three years, and this was one of the stronger correlates of treatment disengagement. Mental health inpatient admissions and emergency room visits also predicted early treatment disengagement, whereas older age and receipt of psychotherapy in the year of buprenorphine initiation were associated with lower risk of treatment disengagement.

Buprenorphine treatment retention in this study (45.0% at one year) is similar to that reported in prior studies of commercially insured populations (24–27) but considerably lower than that observed at VHA facilities (61.6% with comparable methods) (21) and also considerably lower than that in a statewide public initiative in Massachusetts (65%) (37). Most important, VHA patients experienced no risk of disenrollment from their health care plan. In addition, compared with patients with commercial insurance, they were a decade older, had greater access to comprehensive care for substance use disorders, and were likely to have far lower copays (38,39). Higher age, a consistent predictor of better buprenorphine treatment retention (21,40–43), may also have contributed to better retention in the VHA. However, in the Massachusetts study that showed high retention (37), the age distribution was similar to that in the study reported here. In addition, patients with more severe mental health issues may seek more care, especially in the VHA system, where mental health care is relatively accessible at little or no cost (44).

Both the VHA and the Massachusetts studies followed well-supported comprehensive systemwide buprenorphine implementation efforts (37,45), which are often unavailable in outpatient practices funded by commercial insurance. Offering comprehensive psychosocial care for patients with OUD, along with buprenorphine treatment, may help providers funded by commercial insurance improve buprenorphine treatment retention (24,25).

We are not aware of prior studies that addressed continuity of insurance coverage as a factor in continued buprenorphine treatment in a commercially insured population (24–27). This is not an issue in the VHA, where entitlement is based on past military service, and this may largely explain the higher VHA retention rates. The association of continued insurance coverage with buprenorphine treatment retention is impressive and worrisome during this period of changes in health care laws and policies. OUD is a chronic disease that can have disastrous consequences without long-term treatment (46). Our data suggest that unavailability of buprenorphine treatment with loss of insurance coverage (which may itself result from continued opiate use) poses a major threat to effective treatment. Further studies are needed to identify the reasons for loss of insurance coverage among adults with OUD and whether OAT is continued in some cases through other types of financing.

In addition, insurance plans typically impose significant out-of-pocket costs and burdensome requirements for prior authorizations for buprenorphine, posing significant barriers to continued engagement. Although commercial insurance plans were not excluding buprenorphine by 2010, most plans still included buprenorphine in the high cost-sharing tiers, and 38% of the plans required preapproval, substantially higher than in 2003 (7%) (47,48). Insurance plans also restrict the duration of buprenorphine treatment, especially in managed Medicaid plans, and demand onerous counseling requirements (49,50). The effects of the Affordable Care Act that were intended to increase accessibility to substance abuse treatments are still evolving and implementation of the ACA, while currently (2017) in jeopardy, differs across participating states (51). These data suggest that although access to buprenorphine has increased over the past decade (6,10–15), there are still multiple financial barriers. Despite such barriers, patients with high motivation and financial means seem to be accessing buprenorphine treatment (50). The highly restrictive response to the opioid crisis among payers is of concern and deserves more detailed study and documentation.

Altogether, 26.3% of privately insured patients with a diagnosis of OUD initiated buprenorphine in the index year, compared with only 6% initiation of OAT, including buprenorphine (3.4%), in VHA in FY 2012 (unpublished data available on request). The reasons for differences in initiation rates likely involve major differences in case finding, eligibility, enrollment, and diagnostic documentation procedures. In addition, while the vast majority of buprenorphine in the VHA is provided through specialized substance abuse treatment clinics (45,52), buprenorphine provision among privately insured patients is mostly through office-based, nonpsychiatrist physicians (27), who are generally more permissive of buprenorphine prescription than psychiatrists (53). Also, VHA providers do not have the financial incentives that play a prominent part in increasing buprenorphine prescription in non-VHA settings (54,55). VHA is currently implementing programs to increase access to buprenorphine in primary care settings.

This is the first major, large-scale observational study to specifically examine the association of receipt of psychotherapy with buprenorphine treatment retention; both psychotherapy and buprenorphine are perceived as essential components of comprehensive OUD treatment (4). However, the need for psychotherapy or formal counseling with buprenorphine treatment is unclear (56). Psychotherapy or formal counseling yielded no additional benefits over standard physician counseling with regard to illicit opioid use, abstinence, or treatment retention in many recent buprenorphine clinical trials (57–60), whereas other studies have demonstrated added benefit (61–64). Our observational study suggests that receipt of psychotherapy in the first year of treatment was associated with greater retention. This might be a direct effect of psychotherapy, although it may also reflect a selection bias that is not accounted for by the variables available in the study (that is, patients with a better chance of sustained retention might have been selected for or chosen to receive psychotherapy).

Although the proportion of patients with psychiatric disorders was lower in the commercially insured population than in the VHA population (21), psychiatric disorders and their severity, as reflected in emergency room visits and mental health inpatient admissions, were associated with lower retention. Unlike the VHA study, which involved an older population (21), this analysis of data from a younger sample of patients showed that older age reduced the risk of dropout.

The data in this study are observational, which limits causal conclusions about determinants of buprenorphine treatment retention. We lacked data on several factors that have been shown to effect treatment retention, including duration and severity of OUD, type of opioid abused, certain psychosocial factors, nature and duration of treatment interruptions, adherence, and loss of patients because of death (20,24,25). As in the previous VHA-based study (21), we did not have information on the reasons for discontinuation. It is possible that a proportion of patients successfully tapered off buprenorphine with complete resolution of their OUD, although prior studies have suggested that most discontinuations likely represent premature treatment dropout, with a high risk of relapse (33,65,66). Although only outpatient prescriptions were included in the study, we cannot completely exclude the possibility that a proportion of buprenorphine disengagement in the first 30 days occurred after short-term “detox.” In addition, buprenorphine treatment can be influenced by physician characteristics, such as attitudes toward buprenorphine, and by organizational support (67–70). We also lacked information about whether patients disengaged from treatment first and then disenrolled from insurance (perhaps because of relapse and job loss), lost insurance first and discontinued treatment as a result, or received treatment from other sources after disengagement.

Conclusions

This national study of privately insured adults showed that among young patients with OUD who had a low burden of comorbid psychiatric disorders, retention appeared to be low, specifically compared with the VHA system, which offers unbroken enrollment, requires lower out-of-pocket costs, provides integrated care, and serves an older patient population with a high level of psychiatric comorbidity. Our data suggest that loss of insurance coverage is one of the major impediments to buprenorphine treatment retention in the privately insured population. Increasing comprehensive care of substance use disorders and psychiatric disorders at lower cost and without risk of coverage loss is likely to increase treatment retention, improving survival and general well-being.