This study investigated if training and practice-based coaching in an evidence-based program was associated with higher observed treatment integrity (adherence and competence) and if these treatment integrity components were associated with teacher report of child behavioral outcomes in the BEST in CLASS efficacy trial. Participants were 462 children (M = 4.32 years, SD = 0.53; 65% male; 17.0% Caucasian, 66.0% African-American, 5.0% Hispanic, and 12.0% other) identified as having problem behavior and their 185 teachers (M = 12.09 years teaching experience; 99% female; 47.0% Caucasian, 48.0% African-American, 1.0% Asian/Pacific Islander, 3.0% Hispanic and 1.0% other). Teachers and focal children were randomly assigned to the intervention (teacher n = 92, children n = 230) or control condition (teacher n = 93, child n = 232). Results of a multilevel mediation analysis indicated that the BEST in CLASS intervention had a positive effect on teacher report of child problem behavior (SSIS-RS) and externalizing problems (C-TRF), as well as having a positive effect on teachers' adherence and competence of delivery of the intervention. There was an indirect effect through competence of delivery for externalizing problems, but not problem behavior. No indirect effects for adherence were found. Implications of these findings and directions for future research are discussed. (author abstract)