designs are beginning to be used to study the process from effectiveness to large-scale implementation. For example, the community epidemiology model of Kellam and colleagues can examine questions related to effectiveness, sustainability, and moving to scale, along with randomization (Kellam, Koretz, and Moscicki, 1999; Kellam, 2000). This approach examines variation in the community through the use of random assignment of the intervention conditions to different contexts and across time. To test effectiveness of a classroom-based intervention, for example, classes in a school can be randomized to intervention, and the impact for intervention and control classrooms can be compared in the first cohort (Brown, Wang, et al., 2008). This can be followed by examination of intervention sustainability by measuring the level of program fidelity that intervention teachers deliver in the second year with new students. Teachers remain in the same intervention condition, but the support structure in the schools for monitoring, supervision, and resource allocation changes from the first year to reflect the way such a program is likely to be delivered over time. Finally, scalability can be examined in a third cohort in which all the teachers implement the full intervention. Again, with resources allocated as one would anticipate in a scaled-up program, this third cohort can be used to compare the level of program fidelity as well as child outcomes with those of previous cohorts subjected to different levels of infrastructure support. This model, however, allows limited testing of the components of a higher level implementation strategy involving the full school district, since there is only one such district studied at a time (Kellam, 2000).
A valuable approach that would increase knowledge of successful implementation strategies is to test alternative strategies using a randomized trial design (implementation trials). This would necessarily require multiple location and multilevel analyses to fully examine impact. One such implementation trial is comparing the CTC model (see Box 11-1) to an implementation plan with passive assistance (Hawkins, 2006).
In Project Adapt, Smith, Swisher, and colleagues (2004) tested two types of implementation of Life Skills Training (LST) using group-based random assignment. They compared a standard implementation model in which the LST curriculum stood apart from the day-to-day teaching activities with an infusion model that integrated the curriculum into traditional courses. Three rural school districts were randomly assigned to the traditional LST condition, the infused LST condition, and a control condition. There was some suggestion of beneficial impact of both intervention conditions against control for girls in the first year of the study, although these findings generally disappeared by the second year and did not show at all for boys. There were