Scoping implementation science for the beginner: locating yourself on the “subway line” of translational research

There are growing efforts to build capacity for a cadre of implementation science researchers from both federal funders such as the National Institutes of Health [8] and academic universities [9]. We have been part of efforts to train researchers at our respective institutions and in regional, national, and international training efforts. These programs include conventional graduate-level courses, intensive 3-day immersion experiences, and informal and formal mentorship across a range of training stages including undergraduate, graduate, and postgraduate trainees. We engage trainees through a variety of mechanisms including both experiential and didactic approaches. Trainees across the training continuum (e.g., undergraduates to seasoned researchers making a lateral move into implementation science) often ask basic questions like the following: What kinds of research questions fall under the umbrella of implementation science? How do I know if my research focus is ready to be examined with an implementation science lens? How do I know when my intervention is ready for implementation? To answer these questions, we draw on a subway metaphor (Fig. 1), iteratively created with input from our trainees, that we have found to be a useful teaching aid. The goal of this brief commentary is to share our thinking and subway metaphor with the broader research community in the hopes that it can be useful.

We first encourage our trainees to identify the practice that they want to implement, a “practice of interest,” or POI. We use this terminology because there are myriad interventions, procedures, guidelines, tools, and practices that individuals or organizations might seek to implement [10]. Although our trainees tend to be interested in improving health care delivery and outcomes, we note that a “practice of interest” could refer to practices deployed in educational or community settings.

The next step is to evaluate the evidence in support of the POI. Although this step is second nature to implementation scientists, we find that other types of investigators may make their way to this discipline having identified an implementation strategy (e.g., audit and feedback, education) without having scrutinized the degree to which the practice of interest has a strong basis in evidence. Occasionally, students present an “evidence-based” concept or bundle - an example is the use of advanced care planning in patients with advanced illness [11] - whose particular manifestations or components have varying levels of evidence.

Therefore, the first branch point after identifying the practice of interest is to consider whether the POI has been shown to be efficacious. That is, does the POI improve clinical outcomes of interest when deployed under tightly controlled, ideally randomized, conditions? In the development of new drugs, this stage of research often occurs in Phase 2 or Phase 3 testing [12]. For non-drug interventions meant to improve health, efficacy testing may take many forms, including high-fidelity simulation or randomized controlled trials [13]. If the practice of interest has not yet shown efficacy, efficacy studies are needed. In this case, we encourage students to consider future implementation in the development and refinement of the intervention; we call this “designing for dissemination and implementation” [12].

If the POI has shown efficacy, the next question to consider is whether it has also shown effectiveness. We define effectiveness as “real world efficacy,” or evidence of benefit outside the realm of randomized controlled trials with strict inclusion and exclusion criteria [13]. Effectiveness studies should reflect real world practice with heterogeneous settings, clinicians, service providers, and patients or clients. If effectiveness studies showing benefit in the target population are lacking, we submit that two courses of action are reasonable. If no real world studies of the evidence-based practice have been conducted at all, these studies are needed to address the question of whether the POI works in the real world. Oftentimes however, effectiveness studies exist, but the setting or population of interest to the researcher is not adequately represented in the scientific literature. In this case, it may be prudent to test both effectiveness and implementation research questions using an effectiveness-implementation hybrid design [14]. For action-oriented researchers, this family of study designs is highly appealing because it allows simultaneous study of effectiveness and implementation outcomes, thereby expediting the translation of research findings into practice. However, these hybrid studies are often resource intensive and may be unfamiliar to readers and reviewers unfamiliar with implementation science. For these reasons, hybrid trials should only be undertaken by research teams with the expertise to design, execute, analyze, and disseminate them.

In the event that the POI has shown both efficacy and effectiveness, it is appropriate to proceed with studies focused on implementation [15]. Depending on what is already known about the implementation context and potential strategies, these studies might focus on contextual inquiry of implementation determinants (i.e., barriers and facilitators), the development and/or selection of implementation strategies [16, 17], and/or comparative effectiveness studies of different implementation strategies. In our experience, clinical audiences and some funders may demand effectiveness outcomes even in the context of strong effectiveness evidence. Later stage hybrid designs (Type 2 or Type 3) [14] represent one approach to focus on implementation while satisfying a desire to measure effectiveness outcomes. Irrespective of the implementation focus, mixed methods designs are commonly used in implementation research [18].

In our color-coded schematic, only those arms in green (studies of implementation or hybrid studies) fall under the umbrella of “implementation science.” The following hypothetical research questions illustrate the utility of this schematic.