Unpacking High Quality Implementation of SEL Programs: A Latent Class Approach

Introduction: Given the increased national attention to SEL, and adoption of SEL programming, there is corresponding attention on the role of implementation in obtaining desired outcomes.  However, it in order to advance science, we must move beyond a question of ‘whether’ implementation matters, to “what” components matter the most. Thus, the primary aim of the current project was to utilize latent class analysis to identify underlying patterns or classes of implementation.  Person-centered approaches are ideal for advancing implementation science- wherein the aim is to bridge research and practice (Fixen & Blasé, 2006). To do so, we utilized data from a large scale randomized controlled trial of the elementary version of Second Step^® .

Methods: Data were drawn from a larger RCT of Second Step in 61 Elementary schools in Arizona and Washington State. In all, 321 teachers provided implementation data (on dosage, adherence, engagement and generalization), and provided survey data on 7,300 students in grade K-2, pertaining to social-emotional skills.  In addition, observational data was collected on disruptive behavior. Students were diverse with regard to ethnicity (with approximately 31% Latino, 14% African American) and SES. 

Results: A three class solution fit the data best: 54% of teachers fell into a High Quality Implementation Class, 24% fell into a Low Engagement Class, and 21% fell into a Low Adherence Class. Membership was then used to determine impact on student gains from Second Step. The probability of membership in the Low Engagement IC predicted gains in emotional symptoms and conduct problems and reductions in social-emotional composite, skills learning, empathy, emotion management, and problem solving when compared to the High IC.  Probability of membership in the Low Adherence class did not predict changes in the gains in student social-emotional measures over those of predicted by the probability of membership in the Low Engagement IC (when compared to the High IC).

Conclusion: The data suggest that low engagement likely has a negative influence on implementation, whereas the impact of low adherence, is contextualized by engagement. Maintenance of engagement, in light of adaptations or adjustments, may maintain key intervention components.  Although generlization was not as distinguishing as engagement, further reserach is warranted, given its correspondence with engagement. Findings have important implications for theory building, as well as program development and practice.