Comparing variations in implementation processes and influences across multiple sites: What works, for whom, and how?

Traditional analyses and interpretation of controlled trials rely on measures of central tendency (e.g., mean findings for treatment versus control) to detect treatment effects. These trial designs therefore emphasize homogeneity of results, with variations within the experimental or control groups treated as error to be controlled for or ignored. For implementation trials, however, heterogeneity of results is an expected result to be explored rather than an imperfection to be minimized. Thus, many implementation trials seek to understand not only "Does it work?" but also "What works, for whom, and how?" Hence, mixed quantitative-qualitative methods that can capitalize on heterogeneity are needed to (i) comprehensively identify factors that influence the implementation process and (ii) understand their impact on implementation outcomes. This paper outlines the matrixed multiple case study approach, which allows for understanding how these processes and influences similarly or differently interact with outcomes across multiple implementation sites. We provide an example of this approach using data from a multi-site trial that tested the implementation of the evidence-based Collaborative Chronic Care Model at nine US Department of Veterans Affairs medical centers.