Assessing Intervention Adaptations Made during SEL Implementation

Background and Purpose: Preventive interventions can promote social emotional competence and protect against emotional and behavioral disorders (Durlak et al., 2011). The diffusion of effective preventive interventions, however, can be compromised in implementation (Durlak & DuPre, 2008). Adaptation of an intervention may be necessary to accommodate diverse providers, populations, and service environments (Castro, Barrera, & Steiker, 2010). Scholars suggest that adaptations be made proactively and systematically, but in practice, adaptations are often ad-hoc.  This creates concerns about the erosion of fidelity and the dilution of outcomes (Elliott & Mihalic, 2004).

In order to study intervention adaptations and their relation to outcomes, several attempts have been made to characterize the nature of adaptations. This paper examines how three published adaptation models retrospectively classify adaptations. The Moore, Bumbarger, & Cooper (MBC; 2013) model was derived empirically. It uses a three-category taxonomy (fit, timing, and valence). The Castro, Barrera, & Martinez (CBM; 2004) conceptual model proposes a two-category taxonomy that classifies adaptation according to content and form of delivery. The Ecological Validity Model (EVM; Bernal, Bonilla, & Bellido, 1995) is another conceptual adaptation taxonomy that differentiates among eight different types of intervention modifications.

Methods: A social emotional learning program (TOOLBOX) was implemented by teachers during the 2014-2015 academic year. A teacher survey (N=137; 73% response rate, 84% female, 59% white) revealed that over 80% of teachers reported adaptations.  Teachers provided 90 distinct responses to three open-ended questions about the nature of program adaptation. Of these, 12 responses were not regarding adaptation. The remaining 78 responses were classified according the aforementioned adaptation taxonomies.

Results: The MBC model classified 85% of responses (with 58% of these responses related to philosophical fit, 9% related to reactive timing, and 54% to neutral valance). The CBM model classified 68% of responses (with 58% of these responses related to intervention content and 51% related to intervention delivery). The EVM model classified 89% of responses (with a 34% plurality of these responses related to changes in response to the intervention context).  One adaptation was unclassified by any model.

Conclusions and Implications: Both the MBC and EVM models suggest that teachers are most likely to make adaptations for reasons of fit (e.g., adjusting program language to grade level).  The CBM model suggests that adaptations are made to the delivery of the intervention nearly as often as to the content itself.  The EVM Model classified the largest number of adaptations, but with so many discrete dimensions, classification consensus between various raters was difficult to achieve.  Each model provided a unique contribution toward the goal of comprehensively and reliably classifying adaptations, however no single model achieved this goal fully.  Additional work is needed to develop a taxonomy that enables research on the relationship between adaptations and program outcomes.  Ultimately, the planful, systematic, and effective use of adaptation will help providers better fit intervention to diverse communities, ensuring healthy development for all youth.