Data for these analyses were pooled from three related studies of innovation in substance abuse treatment settings. All studies collected data via face-to-face interviews with program administrators between mid-2002 and mid-2004. The University of Georgia's Institutional Review Board approved the protocols for each of the studies.
Two of the studies involved representative samples of (a) publicly funded and (b) privately funded substance abuse treatment centers throughout the United States. Eligible centers were identified by enumerating the population of treatment facilities in sampled counties. To be eligible, treatment centers were required to provide treatment for alcohol and drug dependence at an intensity at least equivalent to structured outpatient programming, as defined by ASAM's patient placement criteria (Mee-Lee, Gartner, Miller, Schulman, & Wilford, 1996
). Programs were also required to be community-based (i.e., available to the general public). Together, these criteria excluded correctional facilities, Veteran's Health Administration programs, counselors in private practice, halfway houses, assessment programs, and driving-under-the-influence services. Programs exclusively providing methadone maintenance services and those exclusively providing psychiatric services were ineligible for the study. However, units offering either of these services, along with other substance abuse treatment services, were eligible and comprise a measurable proportion of the sample.
Centers were classified as “publicly funded” if they received a majority of their annual operating revenues from government grants or contracts, including block grant funds and criminal justice contracts. By contrast, centers were classified as “privately funded” if they received a majority of their annual operating revenues from private sources such as commercial insurance and clients’ out-of-pocket payments. Overall, centers in the public sample received an average of 81% of their annual operating revenues from government grants and contracts, whereas centers in the private sample received <20% of their revenues from such sources. During the study period, the administrators of 363 public and 403 private treatment centers were interviewed, representing response rates of 80% and 88%, respectively.
A third study collected similar data from 240 individual treatment units affiliated with NIDA's CTN. At the time of data collection, the CTN comprised 17 “nodes,” which were clusters of university-based research centers and community-based treatment programs. In total, 109 unique treatment organizations were affiliated with the CTN; these organizations operated 262 administrative units or “cost centers.” Generally speaking, units within a larger organization were defined by service population or modality; for example, an organization might operate three distinct “programs”: methadone services, adolescent residential services, and adult outpatient services. Each of these programs constituted a unit of analysis for this study. Administrative units that were dedicated to prevention/education/outreach services, correctional services, or assessment services were not interviewed, as they were unlikely to have direct involvement in any of the CTN research protocols. During the study period, administrators of 240 units within 104 organizations were interviewed, representing a response rate of 91.6% of all eligible CTN-affiliated treatment units.
In addition to the face-to-face interviews conducted with program administrators, brief telephone follow-ups were conducted 6 months later to identify any major changes in program operations, including the recent adoption of a number of evidence-based practices. Dependent variables for these analyses are drawn from follow-up contacts. Of the 1,006 programs completing face-to-face interviews, 904 (89.9%) completed 6-month follow-up interviews. There were no significant differences across samples in overall follow-up rates, and no significant differences between responding and nonresponding units on independent variables used in these analyses (data not shown).
Two dependent variables are modeled in these analyses: The first is organizational adoption of buprenorphine, which is measured as a dichotomous variable where 0 = no use of buprenorphine and 1 = buprenorphine was used in the program for opiate detoxification or as maintenance therapy; and the second is organizational adoption of voucher-based motivational incentives, which is measured similarly such that 0 = no adoption and 1 = use of voucher-based incentives. Both variables are measured on the 6-month follow-up interview. It should be noted that this definition of “adoption” is distinct from either implementation or institutionalization. Adoption refers to any use of the technique in the program, as distinct from the number of patients receiving the technique or the routineness with which the technique is employed. Adoption represents an early stage of diffusion of new treatment techniques and is thus an appropriate point of focus for these analyses.
Focal predictor variables measure direct and indirect exposures to these treatment techniques via involvement in clinical research. These and all other independent variables were measured at the time of the baseline interview. Two sets of variables about direct exposure, measuring the extent of organizational exposure to buprenorphine and voucher-based incentives via CTN protocols, were constructed for these analyses. Each set of variables categorizes treatment programs into one of three exclusive groups: Treatment center is outside the CTN (used as the reference category); center is in the CTN but its organization is not involved in the protocol in question; and center is in the CTN and its organization is involved in one of the protocols testing buprenorphine or motivational incentives. Because treatment programs have other opportunities for involvement in research outside the CTN, we also include a variable indicating whether the treatment center had previously been involved in a clinical research study involving its patients (1 = yes; 0 = no). Although this is not a direct measure of exposure to focal practices, it should provide some indication of whether “research-oriented” settings are differentially receptive to these two treatment techniques.
Exposure to innovations influences adoption decisions in the broader context of an organization's structure and resources. Thus, several additional predictor variables are included in these analyses. Because an organization's revenue streams can affect both willingness and ability to modify its treatment service offerings, a dummy variable is used to differentiate programs relying predominantly on public revenues (i.e., block grants and criminal justice contracts) from those relying predominantly on private revenues (i.e., commercial insurance and client fees). Similarly, an organization's profit orientation may influence decisions about innovation adoption as they seek more efficient or profitable service delivery methods. Profit orientation is a dummy variable measured such that 1 = for-profit organizations and 0 = not-for-profit organizations, including government-operated facilities.
Two indicators of program quality are included. Accreditation is measured as a dummy variable such that 1 = program is accredited by JCAHO or CARF and 0 = not accredited by either organization. Additionally, each program indicated whether it routinely surveys third-party payers and referral sources as to their satisfaction with the program's treatment services. Organizations more attuned to the satisfaction of their major “buyers and suppliers” may be more likely to adopt evidence-based treatment techniques. This variable is coded as 1 = treatment center routinely collects organizational satisfaction data and 0 = otherwise.
Each model includes one measure of the level of care most likely to be associated with adoption of the treatment technique. Because CTN protocols investigated the use of buprenorphine for detoxification, buprenorphine adoption models control for whether the treatment program offered detoxification services (1 = yes; 0 = no). Similarly, because CTN protocols tested motivational incentives in outpatient modalities, the contingency management models control for whether the treatment center operated on an outpatient-only basis (1 = outpatient only; 0 = inpatient/residential only or mixed inpatient/outpatient).
Three measures of program staff were also examined. First, because larger programs should have greater personnel resources to facilitate the implementation of new treatment approaches, both models control for program size, measured as the number of full-time equivalent (FTE) employees. For ease of interpretation, the absolute number of FTEs is provided in the descriptive statistics, but this measure is log-transformed in the multivariate analyses to adjust for skew. Second, the models control for the availability of physicians at the program. Physicians are necessary to prescribe the medication to clients, and their presence may also be viewed as an indicator of the overall professionalism of the program's staff. A dichotomous variable is used, where 0 = no physicians at the program and 1 = one or more physicians on staff or retained on contract. Third, because the credentials of program staff have repeatedly been associated with organizational-level innovation, both models include a measure of the percentage of counselors holding at least a master's degree.
Client needs may also impact organizational decisions to adopt alternative treatment approaches. Because buprenorphine is indicated for the treatment of opiate dependence and because much of the literature on contingency management focuses on its application to opiate-dependent populations, both models control for the percentage of primary opiate-dependent clients in the center's caseload on the interview date.
Finally, both models control for time. Because interview data for the pooled samples were collected over about a 24-month interval (late 2002 to mid-2004), a measure of time is needed to control for the natural diffusion process that may have occurred over the study period. Programs are grouped into five categories based on the date of the interview (1 = 2002; 2=first half of 2003; 3 = second half of 2003; 4 = first half of 2004; 5 = second half of 2004), and this variable is included in both models. (Modeling time as a set of five dummy variables or as a true continuous variable had no substantive impact on the findings reported here.)