State Agency Interviews
In mid-2006, we conducted brief telephone interviews with at least one staff member in the agency having primary oversight responsibility for the use of buprenorphine in each state. Our initial point of contact in every state was the single state agency (SSA), which is responsible for oversight of addiction services and disbursement of block grant funds. In 29 states, we were directed to speak with the State Methadone Authority (SMA) for information about buprenorphine. Information was obtained from 49 states; one state did not respond to repeated requests for an interview. Respondents were asked a number of questions about the state's policies related to the use of buprenorphine for addiction treatment, including the availability of Medicaid coverage, the presence of any state requirements above and beyond prevailing Federal requirements for the prescription of the medication, and the general disposition of the state office toward the use of buprenorphine in community-based treatment programs. The responses to several key questions were utilized as predictor variables in multivariate models examining program-level adoption of buprenorphine.
State agency staff were asked whether buprenorphine was a covered Medicaid benefit for addiction treatment. Although we collected some detail on the nature and extent of coverage provided (e.g., special populations, pre-approval requirements, and prescription limits), there was insufficient variation in these responses to meaningfully differentiate states based on this detailed information, as are often used in analyses of other, more common, medications. Thus, for these analyses, we use a simple dichotomous variable measuring whether Medicaid provides any coverage for buprenorphine in the state. In all, 28 states reported having Medicaid coverage for buprenorphine at the time of our interviews in 2006. (These states were AK, AZ, CA, DE, FL, HI, IL, MD, MA, MI, MN, NE, NJ, NM, NY, NC, ND, OH, PA, RI, SD, TN, VT, VA, WA, WV, WI and WY.)
A series of additional questions measured the extent to which each of the responding state agencies encouraged or actively facilitated the adoption of buprenorphine by community treatment programs. First, we asked whether the state agency had distributed to treatment providers two recent SAMHSA-authored clinical guidelines on the use of medication-assisted therapy for opioid dependence [40
]. In all, 23 of the responding agencies reported that they had distributed one or both of these publications either in hard copy or via internet links. Fifteen state agencies reported that they had not distributed these materials, while eleven state agency respondents reported being unfamiliar with these treatment improvement protocols.
The tenor of SSA/SMA guidance to OTPs regarding buprenorphine was also informative. There is variability among states in the extent to which buprenorphine is viewed as a preferable alternative to methadone, and whether mainstream treatment facilities are viewed as preferable loci of care to OTPs. Along these lines, 7 states had requirements that OTP advise their clients about the availability of buprenorphine; an additional 14 states actively encouraged OTPs to do so. The remaining states had provided no such guidance to OTPs. When asked a more general question about the state agency's orientation toward buprenorphine, 23 respondents characterized their state agency as "actively encouraging" the use of buprenorphine by addiction treatment providers; the remaining 26 indicated that their agency had taken "no real position" on buprenorphine. No respondents characterized their state as having a negative or discouraging position on this medication.
The responses to these questions were nearly perfectly correlated with reported distribution of the SAMHSA clinical guidelines; that is to say, states that actively encouraged the use of buprenorphine were also those that encouraged OTPs to inform clients about it, and distributed the SAMHSA treatment manuals. Thus, we reduced these three items to one measure for this analysis, namely, whether the state agency actively encouraged the use of buprenorphine.
All respondents were asked about any then-pending regulatory changes, in an effort to determine whether there was variability in the amount of activity surrounding this medication. Few states anticipated any forthcoming regulatory changes pertinent to the use of buprenorphine in community-based treatment settings. The few pending regulatory changes mentioned were procedural in nature, for example, clarifying payment and approval processes for prescribers and pharmacy boards. Due to lack of variation on these items, they are omitted from these analyses.
Treatment Facility Data
To examine the impact of regulatory and funding policies on the adoption of buprenorphine, data from a broad variety of treatment programs across the U.S. is essential. To ensure adequate representation of treatment programs across states, the 2006 National Survey of Substance Abuse Treatment Services (N-SSATS) was used for analysis. The N-SSATS is an annual survey of the population of facilities offering addiction treatment services in the U.S., conducted by SAMHSA. Although the N-SSATS is limited in terms of potential predictor variables, it is nevertheless the closest approximation to a census of the nation's treatment providers. The public use data file for the 2006 survey was obtained from the Substance Abuse and Mental Health Data Archive [42
]. The public use data file contains no facility identifiers; however, codes for state, county, and metropolitan area are included on each record.
The entire 2006 N-SSATS public use datafile includes 13,771 facility records. These records were filtered to exclude facilities that did not deliver substance abuse treatment services. Specifically, facilities were excluded if they exclusively provided intake, assessment and/or referral services, along with those focusing exclusively on the provision of services for persons arrested for driving under the influence. Because our interest was in the adoption of buprenorphine by organizations rather than individuals, solo practitioners were also excluded from these analyses. Finally, facilities operating in Puerto Rico and the other US territories were excluded, because no data were collected from their respective local government oversight agencies about relevant regulatory and funding policies for buprenorphine. These exclusions resulted in a final data set of 12,236 substance abuse treatment facilities.
Several variables available on the N-SSATS file were included as control or predictor variables in examining patterns of buprenorphine adoption. These can be grouped into three broad categories: ownership, facility characteristics, and funding sources. Ownership is defined in terms of three mutually-exclusive dichotomous variables: government-owned, private non-profit, and private for-profit. Because studies have repeatedly documented higher rates of pharmacotherapy adoption in for-profit facilities, these are used as the reference category in the multivariate model.
Facility characteristics include several variables that may be associated with an organization's willingness or ability to adopt medications. First, because hospital-based programs have the infrastructure to support the use of pharmacotherapies, they may be more likely to adopt buprenorphine. Facilities self-reported whether they were located in a hospital (1 = yes, 0 = no). Next, because buprenorphine is indicated for the treatment of opioid addiction, opioid treatment programs (OTPs) may be more likely to adopt it. Programs are coded as OTPs if they self-reported this status to N-SSATS (1 = yes, 0 = no). Third, the N-SSATS includes organizations that may vary in the proportion of total activity devoted to addiction treatment, ranging from self-contained specialty addiction treatment programs and OTPs to units within general hospitals and mental health centers. Organizations having substance abuse treatment as their central focus are expected to be more likely to adopt addiction pharmacotherapies such as buprenorphine. This dichotomous variable is coded 1 if the facility's primary focus is substance abuse treatment, and 0 otherwise.
Because much of the early clinical research on buprenorphine has focused on its use in detoxification protocols, treatment programs offering detox services may be more inclined to adopt it. Programs self-reported whether they offered detoxification among their array of services (1 = yes, 0 = no). Buprenorphine provides community treatment programs an option for treating opioid-dependent clients without investing in the infrastructure required to offer methadone maintenance services. While hospital inpatient programs already have the resources necessary to utilize medications, outpatient programs may be less equipped to do so. To examine whether adoption rates vary significantly by primary treatment modality, we control for whether the program operates on an outpatient-only basis (i.e., does not offer hospital inpatient or residential treatment services). At the time of the survey, the use of buprenorphine with adolescent clients had not been thoroughly studied; as a result, its adoption in adolescent treatment programs as of 2006 is expected to be lower than in programs serving predominantly adult clients. For the purpose of these analyses, N-SSATS facilities are classified as adolescent programs if at least 75% of their clients were under age 18 on the survey reference date.
The use of pharmacotherapies and other evidence-based practices are generally viewed as indicators of higher-quality treatment programs. Because accreditation is also a widely accepted indicator of program quality, it is controlled for in these analyses. Programs self-reported their accreditation status to N-SSATS, and a dichotomous variable is used. Analyses also include a measure of the total number of admissions to the treatment facility in the past year, with the expectation that larger facilities will have both the resources and customer base needed to support the adoption of buprenorphine.
Upon its approval by the FDA, buprenorphine was widely advertised as a viable option for the delivery of opioid dependence treatment in rural and remote areas, where patients lacked access to opioid treatment programs. To examine its early diffusion in this regard, we include a measure of geographic location. Specifically, N-SSATS indicates the metropolitan or micropolitan area (if any) where each facility is located. These have been reduced to a dichotomous variable, where 1 = located in metro/micropolitan area, and 0 = located outside these areas (i.e., rural counties).
Three measures of potential program funding sources are coded from the N-SSATS data file. The first is whether the program receives any funds from federal, state, or local government sources. Programs receiving funding from public sources are likely to be more influenced by state policies. Receipt of such funds is reported as a dichotomous variable in N-SSATS, and is coded as such in these analyses. A second N-SSATS variable indicates whether the treatment program accepts Medicaid as a form of payment for treatment services (1 = yes, 0 = no). A third funding-related variable indicates whether the program has any written arrangements with managed care companies (1 = yes, 0 = no). This is included as an indicator of reliance on private insurance, which is expected to be positively associated with buprenorphine adoption.
Finally, the dependent variable in these analyses is a single N-SSATS item indicating whether the program uses buprenorphine (1 = yes, 0 = no). In separate questions, facilities indicated whether they used Subutex and/or Suboxone. An affirmative response to either item resulted in a "yes" response on the buprenorphine adoption variable. N-SSATS does not provide sufficient detail to measure the extent of use within programs (e.g., the number of clients receiving the medication, or the frequency with which it is prescribed). In 2006, 11.4% of eligible N-SSATS respondents indicated that they prescribed buprenorphine. Table provides descriptive statistics for each of these N-SSATS variables.
Characteristics of N-SSATS facilities in analyses (N = 10,410)
Logistic regression models were estimated to test the effects of facility characteristics and state environments on the likelihood of adoption of buprenorphine by community-based treatment programs. Responses from the SSA/SMA interviews were attached to the record of each treatment program in the data set. As a result, programs within the same state have non-independent observations on the state environment variables; to account for this, the multivariate analyses were run using the survey ("svy") set of commands available in Stata v8.0, with state as the primary sampling unit. This approach produces robust standard errors and accounts for the effect of clustering and stratification in survey sample designs when calculating variance, standard errors, and confidence intervals [43
Listwise deletion reduced the available N to 10,410 for the regression models. The majority of missing values were on the "past year admissions" variable. Facilities excluded from the analyses were not significantly different than those retained, either in their structural characteristics or in their adoption of buprenorphine (data not shown).