A refined compilation of implementation strategies: results from the Expert Recommendations for Implementing Change (ERIC) project

Background Identifying, developing, and testing implementation strategies are important goals of implementation science. However, these efforts have been complicated by the use of inconsistent language and inadequate descriptions of implementation strategies in the literature. The Expert Recommendations for Implementing Change (ERIC) study aimed to refine a published compilation of implementation strategy terms and definitions by systematically gathering input from a wide range of stakeholders with expertise in implementation science and clinical practice. Methods Purposive sampling was used to recruit a panel of experts in implementation and clinical practice who engaged in three rounds of a modified Delphi process to generate consensus on implementation strategies and definitions. The first and second rounds involved Web-based surveys soliciting comments on implementation strategy terms and definitions. After each round, iterative refinements were made based upon participant feedback. The third round involved a live polling and consensus process via a Web-based platform and conference call. Results Participants identified substantial concerns with 31% of the terms and/or definitions and suggested five additional strategies. Seventy-five percent of definitions from the originally published compilation of strategies were retained after voting. Ultimately, the expert panel reached consensus on a final compilation of 73 implementation strategies. Conclusions This research advances the field by improving the conceptual clarity, relevance, and comprehensiveness of implementation strategies that can be used in isolation or combination in implementation research and practice. Future phases of ERIC will focus on developing conceptually distinct categories of strategies as well as ratings for each strategy’s importance and feasibility. Next, the expert panel will recommend multifaceted strategies for hypothetical yet real-world scenarios that vary by sites’ endorsement of evidence-based programs and practices and the strength of contextual supports that surround the effort. Electronic supplementary material The online version of this article (doi:10.1186/s13012-015-0209-1) contains supplementary material, which is available to authorized users.


Background
Research focusing on implementation strategies, defined as "methods or techniques used to enhance the adoption, implementation, and sustainability of a clinical program or practice" [1], has been prioritized in order to bridge the quality chasm in health and mental health services [2][3][4][5]. a However, efforts to identify, develop, and test implementation strategies have been complicated by a lack of conceptual clarity [1,[6][7][8][9]. This lack of conceptual clarity manifests in two primary ways. First, terms and definitions for implementation strategies are inconsistent [7,10]. Idiosyncratic use of implementation strategy terms involve homonymy (i.e., same term has multiple meanings), synonymy (i.e., different terms have the same meanings), and instability (i.e., terms shift unpredictably over time) [11]. Implementation scientists have responded by calling for efforts to clarify terminology and use it consistently [1,[5][6][7]12]. Second, published descriptions of implementation strategies too often do not include sufficient detail to enable either scientific or real-world replication [1,6], leading some to suggest guidelines for specifying and reporting implementation strategies [1,6,13,14]. Taken together, these two deficiencies complicate the acquisition and interpretation of knowledge, preclude research syntheses such as systematic reviews and meta-analyses, limit replication in both research and practice, and ultimately stymie the translation and application of empirical studies that could inform implementation processes [1,6,9].
A number of taxonomies of implementation strategies have been developed, in part, to address these shortcomings pertaining to the published literature, e.g., [10,[15][16][17][18]. Powell et al. [10] reviewed 41 compilations and reviews of implementation strategies and summarized them according to their foci and disciplines/clinical specialties that they represented (this can be found in Table One of that publication). While they acknowledge that many of those compilations represent seminal contributions to the field, they also argue that most of the compilations were not necessarily intended to be consolidated "menus" of potential implementation strategies for a broad range of stakeholders in health and mental health. Powell et al. [10] note that many compilations and reviews: are purposely narrow in scope, focusing on strategies with known evidence on effectiveness, e.g. [19][20][21][22]; specific medical conditions, fields of practice, or disciplines, e.g. [23][24][25]; strategies that were used in a specific setting or study, e.g. [26,27]; "exemplar" programs or strategies, e.g. [28,29]; one level of target such as consumers or practitioners, e.g. [30]; or one type of strategy such as educational or organizational strategies, e.g. [24,31]. The characteristics of some of these reviews and compilations may lead health care stakeholders to believe that there are relatively few strategies from which to choose. Additionally, many of these compilations do not provide definitions or provide definitions that do not adequately describe the specific actions that need to be taken by stakeholders.
In response to those limitations, Powell et al. [10] proposed a consolidated compilation of 68 discrete (as opposed to multifaceted) implementation strategies and definitions based upon a review of the health and mental health literatures. While the review was conducted by an interdisciplinary team of health services researchers, the development of the compilation was not informed by a wide-range of implementation and clinical experts, and the authors did not seek to generate consensus on the strategy terms and definitions beyond the study team [10]. This raises the question of whether the strategy terms and definitions identified would resonate with a broader array of researchers and implementers in real-world settings. The Expert Recommendations for Implementing Change (ERIC) study [9] builds upon the Powell et al. [10] review by generating expert consensus "on a common nomenclature for implementation strategy terms, definitions, and categories that can be used to guide implementation research and practice in mental health service settings" [9]. We pursued this aim by recruiting a panel of stakeholders with expertise in implementation science and clinical practice and engaging them in a three-round modified-Delphi process to refine Powell et al.'s [10] compilation of implementation strategies. While many other efforts to generate consensus have relied solely upon qualitative approaches, e.g., [8,10,32], this study's mixed methods approach provides more structure for the expert recommendation process and derives consensus quantitatively. We describe these processes below, and more details about our methodological approach have been published elsewhere [9].

Expert panel participants
We employed a purposive sampling procedure [33] that began with an initial list of implementation science experts generated by members of the study team. The team targeted a number of groups based upon their substantial expertise in implementation research, including members of the editorial board for the journal Implementation Science, implementation research coordinators for the VA Quality Enhancement Research Initiatives (QUERIs) [34], and faculty and fellows from the National Institute of Mental Health funded Implementation Research Institute [35]. Nominees were encouraged to identify peers with expertise in implementation science and clinical management related to implementing evidence-based programs and practices. Efforts were made to ensure a diverse sample by including VA and non-VA implementation experts and by attempting to obtain a balance between implementation and clinical expertise. Recruitment was limited to individuals residing in the four primary time zones of North America (i.e., Eastern through Pacific) in order to minimize scheduling conflicts for the live Webinar (described below). Ultimately, we recruited a panel of 71 experts (see "Contributors" section for a full list of participants), each of whom participated in at least one of the three Delphi rounds (see Table 1). Ninety-seven percent of the experts were affiliated with academic or health-care institutions in the USA, and 3% were affiliated with Canadian universities. Ninety percent of participants had expertise in implementation science and practice, and 45% were also experts in clinical practice. Nearly two-thirds of participants had some affiliation with the VA, though most of those individuals also had academic appointments in social science or health-related schools or departments.

Modified Delphi process
The modified Delphi process [36] had three rounds. The first two rounds provided the opportunity for panel members to offer feedback on a list of strategies and definitions via two Web-based surveys. After each of the first two rounds, iterative refinements were made to the compilation based upon participant feedback. The third round involved a live, Web-based polling process to obtain consensus on the final compilation of strategies.

Round 1
Fifty-seven experts completed the Round 1 Web-based survey. Section one of the Round 1 survey listed terms and definitions from Powell et al.'s [10] published taxonomy of 68 strategies. Each "item" included a strategy term, its definition, a text box for participants to write in possible synonyms, and a text box for further comments, proposed definitions, or concerns regarding the strategy term or definition. Section 2 of the Round 1 survey asked panelists to propose strategy terms and definitions not included in Powell et al.'s [10] compilation. The full survey can be viewed in Additional file 1.

Round 2
Forty-three experts completed the Round 2 Web-based survey, which included the implementation strategy terms and definitions from Round 1 along with a summary of the panelists' comments and suggestions regarding additional strategies. This included both a qualitative summary and, where possible, a quantitative characterization of participants' Round 1 responses (e.g., 72% of panelists made no comment). The core definitions from the original compilation [10] were separated from their accompanying "ancillary material" (additional details that may be helpful in understanding the nuances of the strategy). This allowed us to summarize and group the feedback from Round 1 according to whether the concerns panel members expressed pertained to the core definition, alternate definitions (proposed by participants in Round 1), or concerns or addendum to the ancillary material. The full Round 2 survey can be viewed in Additional file 2. Once again, participants could suggest additional strategies and make additional comments in response to the strategies, definitions, or feedback from Round 1. Panelists' feedback from Round 2 was used to construct a final list of strategies and definitions for the consensus meeting in Round 3. Terms and definitions were considered "acceptable" to the expert panel and were not included in the Round 3 voting if no panelist suggested alternatives or expressed concerns about the core definition.

Round 3
Forty experts participated in Round 3 of the modified Delphi, which involved a live polling and consensus process conducted via a Web-based interactive discussion platform. Prior to the meeting, panelists were e-mailed a voting guide describing the voting process along with a ballot, allowing them to prepare responses in advance (the voting guide and ballot can be viewed in Additional files 3 and 4, respectively). During the consensus meeting, each implementation strategy term and core definition for which concerns were raised during Round 1 or 2 was presented along with the alternative definitions proposed from the earlier rounds. Terms with only one alternative definition were presented first, followed by those with multiple alternatives. This strategy was used so panelists could "warm up" by voting under the least complicated circumstances, with voting continuing with increasingly difficult scenarios and ending with voting on new terms proposed by panelists. The first stage of voting involved Total represents the total number of unique experts participating in at least one round of the modified Delphi process.
"approval voting", in which panelists were given the option to vote for as many definitions (original and alternative) they thought acceptable. Approval voting is particularly useful for efficiently identifying the most acceptable choice [37], as it has been deemed the most "sincere and strategy proof" form of voting [38]. It promotes collaborative versus adversarial forms of decision making. Furthermore, it allowed us to determine whether the definitions from the original compilation [10] were acceptable even when alternative definitions may have been preferred. Approval ratings for existing definitions, when low, pointed to the need for improving definitional clarity. While no research literature could be found to support a supermajority cutoff, we drew upon supermajority benchmarks from the US Senate [39]. Three fifths (60%) is required to end debate for most issues, while two thirds (66%) is required for other actions. We opted for the convention used to end debate (60%). This ended up being fortuitous for timely completion of the Webinar, as there would have been six additional debates and runoff votes had we opted for a higher supermajority rate. We acknowledge that we may have received different results if we had used 66%. In the first stage of voting, a definition that received a supermajority of votes (≥60%) and also received more votes than any other definition was declared the "winner", and the poll was advanced to the next term. When there was no clear supermajority winner, panelists discussed the definitions. Discussions were highly structured to maximize productivity during the 60-min Webinar. Panelists indicated if they wanted to make a comment by clicking a virtual hand raise button in the Webinar platform and had up to 1 min to make comments. Subsequent discussion was then limited to 5 min per strategy. Following open discussion, the second stage of voting involved "runoff voting", in which participants selected only their top choice. If only two alternatives were presented, the definition receiving the most votes was declared the winner. If three or more alternatives were presented and a majority (i.e., more than 50%) was not obtained in the first runoff vote, then the top two alternatives from the first runoff round would advance to a final runoff round to determine the winner. If a tie between the original and alternative definition occurred in the runoff round, the definition already published in the literature was retained. These same voting procedures were applied to the additional strategies proposed by the expert panel in Rounds 1 and 2 of the Delphi process; however, the approval poll also included an option for the proposed strategy to be rejected if a supermajority (≥60%) of panelists deemed the strategy unworthy of inclusion. Figure 1 provides an overview of the voting process [9].
Four of the forty panelists were unable to successfully utilize the Webinar program but did participate in polling by e-mail while following the Webinar proceedings using their voting guide (Additional file 3) and participating in the discussion using the teleconference line. The multiple sources of votes (through Webinar polling and e-mails) were aggregated in real time.
The Institutional Review Board at Central Arkansas Veterans Healthcare System has approved all study procedures.

Rounds 1 and 2
Expert panelists suggested a number of changes to Powell et al.'s [10] terms and definitions and proposed additional strategies. For example, suggested changes to strategy terms included changing "tailor strategies to overcome barriers and honor preferences" to simply "tailor strategies", and "penalize" to "develop disincentives". The alternate definition for the term "develop an implementation glossary" is illustrative of the participants' efforts to ensure strategy definitions were clear. The original definition was "develop a glossary to promote common understanding about implementation among the different stakeholders". A new definition was proposed, "Develop and distribute a list of terms describing the innovation, implementation, and the stakeholders in the organizational change." Finally, five new terms and definitions were suggested in Round 1, including "promote adaptability", "external facilitation", "identify early adopters", "promote network weaving", and "provide local technical assistance". Table 2 provides a summary of the types of changes to original strategy terms and definitions that were suggested in Rounds 1 and 2, as well as the new strategy terms that were proposed. The majority of the expert feedback received in Rounds 1 and 2 did not focus on strategy terms and core definitions, but rather involved concerns, additions, or clarifications pertaining to the ancillary material. For example, for the strategy "provide ongoing consultation", participants noted that consultation can be conducted by individuals outside of the organization and that it can focus on system and culture change in addition to clinical concerns. Feedback on ancillary materials did not impact the core definition of the strategy and was thus integrated into the ancillary material at the discretion of the study team. A more comprehensive description of the types of feedback received in Rounds 1 and 2 can be viewed in Additional file 2.

Round 3
The majority of the terms and definitions (69%) from the Powell et al. [10] compilation were considered "no contest" and were not subjected to voting in Round 3 as participants did not raise substantial concerns or suggest alternative definitions for them. Twenty-one strategies and five new strategies were subjected to voting in Round 3. The complete results from the Round 3 voting can be viewed in Additional file 5. For each vote, there was a small number of abstainers; the percentage of participants casting votes ranged from 83 to 94%. In the majority of cases, the initial vote (i.e., the approval voting stage) yielded a clear winner; however, in two cases, no strategy received over 60% of the vote in the approval voting stage and in another case there was a tie between two strategies, each receiving 66% of the votes. In these situations, the participants discussed their thoughts and concerns, after which the runoff vote successfully identified a winning definition.
For the 21 alternative definitions suggested, an alternative definition was selected 81% of the time and the original definition was maintained 19% of the time. One of the advantages of approval voting was determining the acceptability of the original definitions even when alternatives were thought to be superior. In each of the 17 times in which an alternative was ultimately selected, the original definitions failed to reach the supermajority approval level of 60% (average 30%, range 3 to 51%).
Each of the five new strategies that the panel proposed was maintained in some form. Panelists had the opportunity to reject the proposed additions, but on average, across the five strategies, 84% of panelists voted to retain the new strategy (range 100 to 71%). Each of the new strategies had an initial proposed definition in Round 1. Panelists had the opportunity to suggest alternative definitions in Round 2. In two cases ("promote network weaving" and "provide local technical assistance"), no alternative definitions were proposed, and the new definition was retained with approval votes of 71 and 73%, respectively. In one case ("identify early adopters") the alternative definition won in the approval vote. Finally, in two cases ("facilitation" and "promote adaptability"), the original new definition was selected over the alternatives in the runoff vote. Figure 1 Overview of the voting process in the final round of the modified Delphi task. In the third and final round of the modified Delphi task, expert panelists will vote on all strategies where concerns were raised regarding the core definition in the first two online survey rounds. For each strategy, the original and proposed alternate definitions will be presented for an approval poll where participants can vote to approve all definition alternatives they find acceptable. In the first round of voting, if one definition receives a supermajority of votes (≥60%) and receives more votes than all others, that definition will be declared the winner and the poll will move to the next term. If there is no consensus, a 5-min discussion period is opened. When the discussion concludes, a runoff poll is conducted to determine the most acceptable definition alternative [13].

Final compilation
The final compilation included 73 discrete strategies (Table 3). Consistent with the Powell et al. [10] compilation, active verbs were used to describe the implementation strategy terms. We attempted to strike a balance between economy of expression and comprehensiveness. Thus, in some cases, we used verbs like "develop" or "create" instead of "develop and implement" or "create and implement", though the implementation or use of the strategies developed or created should be thought of as part of the same process. In many cases, this is clarified in the definition. For example, the strategy "develop a formal implementation blueprint" specifies in the definition that the blueprint should be used and updated. Each of the strategies, including those in which the verb "use" is included in the strategy term, should be thought of as discretionary for researchers and implementers.
Our intent was to highlight the range of discrete strategies that could potentially be used to implement new programs and practices, not to present a checklist of strategies that must be used in all efforts. Additional file 6 contains the full compilation with ancillary material that contains additional references and details that may be useful to implementation stakeholders, such as advice about how a particular strategy might be used.

Discussion
This study aimed to refine and achieve consensus on a compilation of implementation strategy terms and definitions by systematically gathering input from a wide range of stakeholders. A large, accomplished panel of implementation and clinical experts was successfully engaged in a rigorous consensus development process. Participants identified substantial concerns with 31% of the terms and/or definitions from the original Powell et al. [10] compilation and suggested five additional strategies. Seventy-five percent of the definitions from the original compilation were retained after voting. The expert panel achieved consensus on a final compilation of 73 implementation strategies. This study has improved the original published compilation by enhancing the clarity, relevance, and comprehensiveness of included strategies and ensuring that they resonate with a wide range of stakeholders conducting implementation research and practice. There are several immediate uses of this compilation. First, it provides a list of discrete strategies that can serve as "building blocks" for constructing multifaceted, multilevel implementation strategies for implementation efforts or in comparative effectiveness research [4]. Second, the core definitions and ancillary materials (see Additional file 6) can be used in conjunction with available reporting guidelines [1,13,14,40,41] to improve the specification and reporting of implementation strategies in efficacy, effectiveness, and implementation research [42]. Finally, the refined compilation can be used as a tool to assess discrete strategies that have been used in published implementation research. Mazza et al. [18] recently demonstrated how taxonomies can be used for that purpose.
The subsequent stages of the ERIC project [9] will further enhance the utility of this compilation in a number of ways. First, expert panelists will complete concept mapping [43] and rating exercises to derive conceptually distinct categories of strategies, interrelationships between them, and a rating for each discrete strategy's importance and feasibility. This information will help users select strategies for their planned implementation efforts by highlighting the broad categories they might consider and providing feasibility and importance ratings of both individual discrete strategies and clusters of strategies. Second, expert panels will be asked to choose the best implementation strategies to use in real-world scenarios that describe implementations of specific evidence-based practices (e.g., measurement-based care for depression) in hypothetical VA mental health clinic settings that vary on certain contextual characteristics [9]. This stage of ERIC will yield recommendations about which multifaceted, multilevel strategy is best matched to specific scenarios. This information will help provide guidance for similar implementation efforts and insights into how recommendations may change based on clearly described differences in context.
As Powell et al. [10] cautioned, this compilation should not be thought of as a checklist. No implementation effort could feasibly utilize every one of these strategies. The ERIC compilation provides a list by which to select discrete strategies that can be used to build a tailored multicomponent strategy for implementation. Future research is needed to identify the contexts and circumstances under which each discrete strategy is effective to help guide users in their selection.
We note that while our attempt was to identify discrete strategies involving one action or process, the included strategies vary in their level of complexity. In fact, active research agendas have focused on determining the essential components of many of these "discrete" implementation strategies, such as audit and feedback [44], learning collaboratives [45], and supervision [46]. The evidence will continue to accumulate, providing more detailed specifications of components for discrete strategies to help inform future iterations of this and other compilations.
The ERIC compilation consolidated discrete implementation strategies that have been identified through other taxonomies and reviews (see Powell et al. [10] for a list of sources and methodological details). Thus, there are many similarities between the ERIC compilation and other taxonomies. However, the ERIC compilation Create or change credentialing and/or licensure standards Create an organization that certifies clinicians in the innovation or encourage an existing organization to do so. Change governmental professional certification or licensure requirements to include delivering the innovation. Work to alter continuing education requirements to shape professional practice toward the innovation Develop a formal implementation blueprint Develop a formal implementation blueprint that includes all goals and strategies. The blueprint should include the following: 1) aim/purpose of the implementation; 2) scope of the change (e.g., what organizational units are affected); 3) timeframe and milestones; and 4) appropriate performance/progress measures. Use and update this plan to guide the implementation effort over time Table 3 ERIC discrete implementation strategy compilation (n = 73) (Continued) Develop academic partnerships Partner with a university or academic unit for the purposes of shared training and bringing research skills to an implementation project Develop an implementation glossary Develop and distribute a list of terms describing the innovation, implementation, and stakeholders in the organizational change Develop and implement tools for quality monitoring Develop, test, and introduce into quality-monitoring systems the right input-the appropriate language, protocols, algorithms, standards, and measures (of processes, patient/consumer outcomes, and implementation outcomes) that are often specific to the innovation being implemented Develop and organize quality monitoring systems Develop and organize systems and procedures that monitor clinical processes and/or outcomes for the purpose of quality assurance and improvement Fund and contract for the clinical innovation Governments and other payers of services issue requests for proposals to deliver the innovation, use contracting processes to motivate providers to deliver the clinical innovation, and develop new funding formulas that make it more likely that providers will deliver the innovation Identify and prepare champions Identify and prepare individuals who dedicate themselves to supporting, marketing, and driving through an implementation, overcoming indifference or resistance that the intervention may provoke in an organization Identify early adopters Identify early adopters at the local site to learn from their experiences with the practice innovation Increase demand Attempt to influence the market for the clinical innovation to increase competition intensity and to increase the maturity of the market for the clinical innovation Inform local opinion leaders Inform providers identified by colleagues as opinion leaders or "educationally influential" about the clinical innovation in the hopes that they will influence colleagues to adopt it Intervene with patients/consumers to enhance uptake and adherence  (and perhaps beyond). Third, a major strength of this compilation is that it is based on consensus of a broad range of implementation experts. There are several limitations related to the process of generating this compilation. First, had we used a different taxonomy of implementation strategies as a starting point, the modified Delphi process may have yielded different results. However, the original Powell et al. [10] compilation incorporated strategies from several other existing taxonomies, e.g., [15][16][17], increasing the chances that key implementation strategies were included. The fact that the expert panelists suggested few additional strategies also increases our confidence that the compilation was relatively comprehensive. Second, the composition of our expert panel was limited to participants in North America and was mostly composed of implementation and clinical experts from the USA. This was appropriate given the ERIC project's focus on implementing evidence-based mental health programs and practices within the VA and for pragmatic reasons (e.g., scheduling the consensus meeting), but we acknowledge that broader international participation would have been ideal. This may have implications for the content of the compilation, as we discuss below. Third, it is possible that in-person meetings may have generated more nuanced discussions of strategy terms and definitions; however, the asynchronous, online process had the advantage of allowing a wide range of implementation and clinical experts to participate and also ensured anonymity of responses, which limited the possibility of participants simply yielding to the majority opinion in Rounds 1 and 2. Finally, as noted in the "Results" section, a small number of participants abstained from voting for portions of the Round 3 consensus meeting. While we can speculate as to potential reasons (e.g., technical difficulties, other distractions, not finding any of the strategy terms and definitions appropriate), we cannot be certain as to why participants abstained or about whether or not this could have impacted the final results in cases in which voting results were extremely close.
There are also limitations related to the content of the refined compilation. First, the evidence base for each strategy was not considered because the purpose of this work was to identify the range of potential options available. Second, the strategies were not explicitly tied to relevant theories or conceptual models. The compilation's utility would be enhanced by linking each strategy to the domains of prominent conceptual frameworks (e.g., the Consolidated Framework for Implementation Research [47], Theoretical Domains Framework [48,49], Promoting Action on Research Implementation in Health Services (PARIHS) framework [50]). Furthermore, users might benefit from using a recently developed framework by Colquhoun and colleagues [8] to better plan use of the individual strategies by identifying: 1) active ingredients (i.e., the defining characteristics of the implementation strategies); 2) causal mechanisms (i.e., the processes or mediators by which strategies exert change); 3) mode of delivery or practical application (i.e., the way an active ingredient is applied, such as face-to-face, Web-based, mass media, etc.); and 4) intended target (i.e., the implementation strategy's "intended effects and beneficiaries"). Lastly, while we are not aware of evidence that would suggest that the strategies in this compilation would not be applicable to many different contexts, it is possible that some of the strategies may be more applicable to US or North American settings given the focus of the ERIC project and the composition of the expert panel. Engaging a broader international panel may have revealed additional strategies that are applicable to healthcare systems that are organized differently or to settings (e.g., low-and middle-income countries) that are not similarly resourced. The fact that the original compilation drew from taxonomies developed in contexts other than the US, e.g., [15,17] may help mitigate this potential limitation.

Conclusions
This research advances the field by improving the conceptual clarity, relevance, and comprehensiveness of discrete implementation strategies that can be used in isolation or combination in implementation research and practice. The utility of this compilation will be extended in subsequent stages of the ERIC study. We conclude by echoing Powell et al.'s [10] caution that this compilation, while substantially improved, should not be viewed as the final word. We welcome further comments and critiques that will further refine this compilation and enhance its ability to inform implementation research and practice.

Contributors
We would like to acknowledge the contributions of each member of the expert panel: