Patterns of Teacher Implementation in the Head Start Cares Social-Emotional Intervention Demonstration

Recent research highlights the potential for pre-K social-emotional interventions to support children's learning and social adjustment across the early school years and beyond, particularly among low-income children at risk for poor outcomes (Bierman & Motamedi, 2015; Domitrovich et al., 2012). Much of this evidence comes from smaller-scale trials conducted under the close supervision of intervention developers, and less is known about implementation and impacts under "real-world" conditions at scale. A notable exception to this has been MDRC's Head Start CARES demonstration, a large-scale evaluation of three social-emotional interventions implemented in over 300 classrooms in 104 Head Start centers across the U.S., which showed intervention impacts on teacher practice, classroom interactions, and child outcomes in pre-K and kindergarten (Morris et al., 2014). Analyses conducted by MDRC indicated that classroom enhancements were, on average, implemented with satisfactory fidelity (i.e., implemented as intended), and that program implementation improved from fall to spring during the year of implementation (Mattera et al., 2013). The current paper will extend this work by examining change and variation in implementation over time, using weekly and monthly data collected from coaches and trainers who supported teachers implementing the CARES enhancements, now available for secondary analysis through the Interuniversity Consortium for Political and Social Research (ICPSR).

Multi-level models, including best-fitting growth curve models and heterogeneous variance models (Hedeker & Mermelstein, 2007), will explore whether there are distinct patterns of implementation over these levels of micro time, and whether patterns are predicted by baseline characteristics of coaches and teachers. Further, individual parameters (e.g., slopes and residuals from growth curves) will be saved as predictors to explore whether patterns of implementation relate to variation in impacts. For example, we hypothesize that greater consistency in implementation will lead to greater improvements in teacher practice and child outcomes relative to more variable implementation across the school year, and this may vary by program component.

Results will identify patterns of the classroom social-emotional enhancements' implementation over time that are beneficial to teacher practice and child outcomes as well as components for which these patterns are more critical to outcomes. These results will serve to inform models of professional development that support optimal patterns of implementation over time and facilitate coaches' future efforts as they prioritize aspects of implementation for teacher professional development.