Refining the Evidence-Based Practice Attitude Scale (EBPAS): An Alternative Confirmatory Factor Analysis

Background: Evidence-based practice (EBP) is a dynamic process including clinical decision-making, research evidence, clinical expertise, and client characteristics (Sackett et al., 1996; Spring, 2007). Within this process, empirically supported treatments (EST) are an important component often considered the gold standard in human service organizations; however, they can come at a substantial cost. Therefore, it is important to consider different factors that may contribute to the effectiveness and successful implementation of ESTs, such as attitudes of service providers. Service providers' opinions and beliefs regarding ESTs can influence their utilization and implementation of new ESTs (e.g., Aarons, 2004). The goal of this study was to replicate the factor structure detailed by Aarons and colleagues for the Evidence-Based Practice Attitude Scale (EBPAS) and also test alternative models of this measure in a large child and family human services organization. The EBPAS is a widely used measure thus warranting additional examination of its structure and psychometrics.

Methods: Previous investigations of the EBPAS have surveyed participants from a variety of agencies. This study focused on one agency, Hillside Family of Agencies, and gathered information from a total of 1,273 service providers, over 80% of eligible participants. All participants were front-line employees (i.e., having direct contact with the children and families) representing different roles within the agency (e.g., direct care workers in residential settings, therapists, and mentors). The final sample (N= 1,260) had a mean age of 35 years (SD = 10.98; range: 19-73), 58.8% were women, and 74% were Caucasian. Service provider attitudes were measured using the EBPAS (Aarons, 2004; Aarons et al., 2007; Aarons et al., 2010), consisting of 15 items on a 5-point Likert scale ranging from 0 (not at all) to 4 (to a very great extent). First and second order confirmatory factor analyses were conducted. Models were estimated using Mplus 6.1 with MLR estimator to account for non-normal and nested data.

Results: The four-factor model tested in previous studies was replicated (Aarons, 2004; Aarons et al., 2007; Aarons et al., 2010), as well as the second order four-factor model examined by Aarons et al. (2010). Next, alternative first and second order five-factor models were performed, based on evidence from previous reports indicating that the Divergence subscale had low factor loadings and reliability, as well as our own reading of the items suggesting two distinct factors within this subscale. Results indicated that the alternative first order five-factor model was the best fit of the data (χ2[79] = 270.97, p < .001; CFI = .96; TLI = .95; SRMR = .038; and RMSEA = .044 [.04,.05]).

Conclusion: Results indicated better fit statistics for a five-factor model of evidence based practice attitudes among child and family service providers with Divergence split into two distinct factors (Research valueless and Superiority of clinical knowledge). Authors discuss these results with respect to the tradeoffs and utility of the Divergence scale as originally developed versus split into two subscales, as well as what this may mean for agencies hoping to effectively implement evidence based practice in their organizations.