摘要：Translational research applies basic science discoveries in clinical and community settings. Implementation research is often limited by tremendous variability among settings; therefore, generalization of findings may be limited. Adoption of a novel procedure in a community practice is usually a local decision guided by setting-specific knowledge. The conventional statistical framework that aims to produce generalizable knowledge is inappropriate for local quality improvement investigations. We propose an analytic framework based on cost-effectiveness of the implementation study design, taking into account prior knowledge from local experts. When prior knowledge does not indicate a clear preference between the new and standard procedures, local investigation should guide the choice. The proposed approach requires substantially smaller sample sizes than the conventional approach. Sample size formulae and general guidance are provided. It is common in implementation studies to adapt existing procedures, or to improvise new procedures, to accommodate local conditions in a specific community care setting. From an evidence-based principle, these decisions can often be informed by conducting a local investigation or quality improvement study to evaluate the pros and cons for viable options under consideration. For example, within the primary care setting and emerging medical home practices, primary care clinics may implement focused procedures to engage high-risk, high-complexity patients with emotional and medical conditions such as depression and diabetes. Such procedures might include focused physical and emotional health screening and multidisciplinary health care delivery procedures. Before replacing the existing intake and care delivery procedures, the clinic may benefit from pilot testing the new procedure to examine its potential utility and impact on patient care. The results of this pilot investigation can then be used to inform the decision (which intake procedure to use) for future patients including a broader “roll-out” of the successful procedures. The primary purpose of these local investigations is to produce local knowledge (such as which intake procedure is more effective for the specific clinic) to inform local implementation decisions. The pilot testing of these procedures is not intended to produce generalizable knowledge that can be applied universally to other care settings. This underlines the difference between implementation science that aims to produce generalizable knowledge 1 and quality improvement projects that aim to produce local knowledge with a focus on an organization’s own delivery system and process. 2 As such, quality improvement projects for local knowledge do not meet the criterion for human participants research as defined by the Office for Human Research Protections, the federal agency with oversight over human participants protection and institutional review boards. Research means a systematic investigation, including research development, testing, and evaluation, designed to develop or contribute to generalizable knowledge. 3 To avoid the confusion with human participants research that aims to produce generalizable knowledge, we use the term local investigation to highlight the distinction, and also to suggest that those investigations might qualify for exemptions from certain human participants regulations. Local investigations often deal with components (“nuts and bolts,” such as a specific intake procedure) of an overall implementation program. Making appropriate nuts-and-bolts decisions is important for the successful assemblage of the overall implementation “engine.” A specific implementation engine might consist of numerous nuts and bolts. Therefore, it is conceivable that multiple local investigations might be conducted in an implementation program for a specific care setting, each addressing the needs for a specific component of the engine. A variety of designs can be used for local investigations, including randomized designs and nonrandomized quasiexperimental designs. 4 To illustrate the difference between local investigations that aim primarily to produce local knowledge, and the usual research studies that aim primarily to produce generalizable knowledge, we focused on randomized designs for this type of quality improvement project.
