Few topics in American medicine have generated as much interest or debate as the quality of the healthcare we receive. The United States continues to rank well below its developed peers in many international comparisons due, in part, to the persistence of medical errors, preventable deaths, and sub-optimal patient outcomes. In an effort to close these and other gaps in quality, physicians, health services researchers, and improvement experts have searched--and continue to search--for tools that can consistently and sustainably improve clinical performance. This dissertation explores one potential improvement tool: the quality improvement collaborative (QIC). QICs, which require healthcare organizations to work together on a single quality problem, were initially designed simply to disseminate evidence-based best practices, but have evolved into a more general approach for helping organizations make targeted changes to their care delivery systems. From a theoretical perspective, the QIC approach offers several advantages over single-institution improvement strategies, including the ability to rely upon an external support system and to leverage the practical knowledge of organizations that have already made similar changes. However, the empirical data on QICs’ effectiveness are mixed: a few studies suggest that hospitals achieve more significant improvements in QICs than on their own, but several randomized trials show no difference. To better understand the reasons for this inconsistency, I examined the mechanisms behind QIC function and developed a framework for understanding variability in QIC success. I then went on to explore two ways in which QICs might influence quality improvement more generally: by collecting new data or developing new risk adjustment models and by rapidly adapting an intervention from another setting to the local environment. As a whole, my research suggests that, while QICs offer a series of unique benefits to both hospitals and to the field, they are, by no means, a panacea. While there is still no magic bullet, I believe that continuing to develop the QIC approach, identifying where and when it is most effective, and integrating it into the larger armamentarium of improvement tools offers us the best chance to improve the quality of American healthcare for good.