The updated Consolidated Framework for Implementation Research based on user feedback

We completed a literature review to identify articles with feedback on the CFIR. The most efficient search criteria for this purpose was the inclusion of the CFIR in the title and/or abstract (see Additional file 1). We searched SCOPUS and Web of Science from 2009 (the year the CFIR was published) to July 6, 2020. This search yielded 376 articles, including (1) original research; (2) systematic reviews; and (3) evaluation of the CFIR as a framework. Two reviewers (MOW, CMR) read the full text of approximately 10% (n = 40/376) of the included articles to independently abstract feedback on the CFIR; discrepancies with abstraction were discussed until consensus was reached. One reviewer (MOW) then read the remaining articles and abstracted relevant passages. Only 59 of the articles contained feedback on the CFIR.

We surveyed unique corresponding authors of the articles included in the literature review (n = 334) in August 2020 to elicit in-depth feedback about their experience using the CFIR. First, the survey elicited information about the author’s use of the CFIR (e.g., the total number of projects completed using the CFIR) (see Table 1). Second, respondents were asked to rate the CFIR based on Flottorp et al.’s “sensibility” criteria for determinant frameworks (e.g., Applicability, Simplicity) [18] (see Table 2). Third, respondents were asked for open-ended feedback about the framework overall as well as existing domains and constructs. Finally, respondents were asked for recommendations to add or remove domains and constructs (see Additional file 2 for the full survey). Survey invitations were sent via email with an embedded link to the survey.

Survey respondents questioned whether the CFIR was intended to evaluate the innovation and/or the strategy being used to implement the innovation, and they found it difficult to differentiate between them. The literature has recognized that the lack of a clean boundary between the innovation and implementation strategies is a contributor to implementation complexity [22]; however, distinguishing between the innovation and implementation strategy is necessary for accurate attribution to implementation outcomes [28] and to identify appropriate areas for future intervention. As a result, the updated CFIR guides users to define the innovation (aka “the thing” [20, 25] being implemented), including the boundary between the innovation and implementation strategies. We encourage use of a reporting guideline to define the innovation (see Table 3).

The word Innovation was added to the name of each construct in the Innovation Domain to orient users to the focus of this domain: the Innovation itself, independent of the implementation strategy. Major revisions were made to the definition of Innovation Complexity: the text “difficulty of implementation” was replaced with “the innovation is complicated” to focus attention on the innovation, not implementation.

While some users recommended dividing the Outer Setting into multiple levels, others wanted to combine the Outer and Inner Settings, describing difficulty understanding boundaries between the two settings. In the original CFIR article's Additional file 1, the boundary between the Inner and Outer Settings was visually depicted using “overlapping, irregular, and thick grayed lines” to highlight that the line between them is not always clear [17]. Lengnick-Hall et al. expand on this reality and call for researchers to take an “open-systems” perspective “to highlight interdependence between outer and inner contexts and [to] view organizations as part of a broader interdependent system that may range from simple to complex, rigid to flexible, and loosely to tightly coupled” [37]. Although embracing an open-systems perspective may be challenging, conceptually differentiating internal and external influences on the performance of organizations has been a central tenet of organization science [38] and highlights the level at which to focus interventions. As a result, the updated CFIR retains the two domains and guides users to objectively define their Outer vs. Inner Settings, including defining multiple levels of the Outer Setting if appropriate.

A few constructs were renamed because users felt the labels were unintuitive (e.g., Cosmopolitanism) or confusing (e.g., Peer Pressure). Patient Needs and Resources was separated into three constructs and relocated to the Inner Setting and Individuals Domains in response to comments that it captured two distinct themes: awareness of patient needs versus prioritization of patient needs [39].

Users remarked that the Outer Setting domain was underdeveloped [40, 41]. The updated CFIR adds constructs to capture the potential influence of Local Attitudes, i.e., sociocultural values and beliefs, and Local Conditions, i.e., economic, environmental, political, and/or technological conditions, on the willingness and ability of entities within the Outer Setting to support implementation and delivery of the innovation [42‐47], which may influence equity in implementation. These constructs are especially important for innovations that require support from community entities, such as Housing First models of care [48], and for capturing common resource constraints in LMICs [42].

The original CFIR’s broad construct, External Policies and Incentives, was separated into several new constructs, including, for example, the key role of Financing [46, 49‐51]. The updated CFIR also better captures diverse sources of External Pressures [46], including Societal Pressure (e.g., pressure from social movements and protests) [45], Market Pressure (e.g., pressure to compete with and/or imitate peer entities), and Performance Measurement Pressure (e.g., pressure to meet publicly reported goals).

Some users recommended dividing the Inner Setting into multiple levels [52] to account for teams or units [53, 54]. We added guidance for users to objectively define their Inner Setting and to add additional levels as needed. For example, Safaeinili et al. adapted the CFIR to accommodate three embedded levels: (1) pilot clinics, (2) peer clinics, and (3) the larger health system [54].

New constructs and subconstructs were added to the Inner Setting to address several critiques. For example, Culture was felt to be too broad, with one survey respondent stating, it “ends up becoming my ‘I don’t know where else this fits’ bucket” (survey response). Additionally, users noted the absence of equity considerations [40, 42], including “more specifically racism, patriarchy and misogyny, that [are] so much a part of the care that we provide” (survey response). As a result, four subconstructs were added to Culture, including Human Equality-Centeredness, Recipient-Centeredness, Deliverer-Centeredness, and Learning-Centeredness, which serve to orient users to determinants that may influence equity in implementation.

In addition, as described in our companion article, The CFIR Outcomes Addendum [19], Implementation Climate and Readiness for Implementation were removed from the updated CFIR. Though few users commented on these constructs, some questioned their meaning and “nesting” of subconstructs within each in the framework (e.g., Leadership Engagement, Available Resources, and Access to Knowledge and Information were all nested within Readiness for Implementation). Though there is broad recognition that implementation climate and readiness are a function of multiple implementation determinants, there is no consensus on precisely which determinants. Therefore, we have reclassified these constructs to more appropriately position them as antecedent assessments [55], on the pathway between implementation determinants and outcomes in the CFIR Outcomes Addendum [19].

Many users felt the CFIR did not provide “sufficient individual-level constructs” [45] and were unclear about which individuals were included [45‐47, 56‐59]. Furthermore, they felt that constructs in this domain overlapped with constructs in other domains and failed to capture more important individual-level characteristics. One user summarized this feedback well: “[The CFIR needs to focus] more on who the individuals are and their underlying characteristics” (survey response). As a result, the Individuals Domain was significantly reorganized and now includes two subdomains: Roles and Characteristics.

In the original CFIR, roles were spread across three different domains: Patient Needs and Resources was listed in the Outer Setting, Leadership Engagement was listed in the Inner Setting, and multiple implementation-specific roles were listed in the Process Domain (e.g., Formally Appointed Internal Implementation Leaders). All roles have been relocated to this new subdomain, and additional roles were added, including Implementation Team Members [60]. In addition, the Formally Appointed Internal Implementation Leader and Champion constructs were combined into the Implementation Leads role because of the inability of users to distinguish between the two roles [61], and as affirmed in a review of champions [62].

Users felt that the existing Characteristics constructs overlapped with constructs in other domains, e.g., Knowledge and Beliefs overlapped with all constructs in the Innovation Domain. In addition, they thought the domain failed to capture more relevant characteristics related to professional roles and identities, skills and capabilities, autonomy, and level of involvement [46, 47, 59]. Some CFIR users combine this domain with the Theoretical Domains Framework (TDF), which was developed with the intent “to simplify and integrate a plethora of behavior change theories and make theory more accessible to, and usable by, other disciplines” [63]. The COM-B system was developed as an even more simplified system by which to acknowledge key domains related to behavior change based on US consensus of behavioral theorists and a principle of criminal law defining specific prerequisites for volitional behavior [29]. As a result, the original CFIR Characteristics constructs were replaced with constructs based on Michie et al.’s COM-B system [29]. The COM-B posits that broad categories of Capability (e.g., skills), Opportunity (e.g., autonomy), and Motivation (e.g., commitment) shape behavior.

The COM-B constructs are each mapped to 14 domains in the TDF, which provides CFIR users a wide portal into a repository of 84 behavior-change theoretical constructs. In addition, we encourage users to add additional constructs and map them to the COM-B as appropriate. For example, theories, models, and frameworks related to:

We also added the Need construct, based on feedback about its importance for all constituents [56], and to capture facets of the original CFIR Patient Needs and Resources construct.

We added guidance to encourage users to describe their overall approach or implementation process framework to guide implementation, e.g., the Interactive Systems Framework [71]. Doing so helps distinguish the Innovation from the Implementation Process and accompanying implementation strategies.

Some users questioned the inclusion of the Implementation Process Domain in the CFIR because it appears to include strategies, not contextual factors. We clarify that the goal of this domain is to capture “the degree to which” each of these processes occur during implementation and influence implementation outcomes. Additional constructs were added in the updated CFIR to acknowledge scientific advancement since 2009 that are common across many process frameworks [8] and collective-level change theories [72]. Depending on the process framework used for a particular project and the implementation strategies used [26, 27], there may be other components of the implementation process that users should add.

The updated CFIR has expanded the number of constructs within the Implementation Process Domain in response to critiques that key processes and strategies were missing. Though it is outside the scope of the CFIR to include all 73 implementation strategies from the Expert Recommendations for Implementing Change (ERIC) [26, 27], a few best practices have been added: Teaming [42, 46, 73], Assessing Needs [46, 47], Assessing Context, Tailoring Strategies [14], and Adapting [45, 74, 75]. Published guidance highlights the importance of Adapting the innovation [76], and the updated CFIR notes the importance of adapting the setting as well [77]. The addition of Assessing Needs: Innovation Recipients and Engaging: Innovation Recipients serves to better center Innovation Recipients in the updated CFIR and orient users to these as important determinants to equity in implementation.

The updated CFIR represents an incremental change from the original based on feedback from CFIR users, approximately two-thirds of whom have applied the CFIR in more than one study. The CFIR (and technical assistance website [110]) is a public resource and common good—free and open to all—and it must continue to evolve. We call on implementation scientists to collaborate with researchers in other disciplines (e.g., equity and justice, business, organizational science) to continue developing the CFIR, building on feedback from an ever-larger community of users. Our call for critique and reflection is echoed by others [16, 111].

Our team can help support these advances, but we do not own the framework, and we represent a narrow slice of the world. We extend an open invitation for others within alternative spheres to move the CFIR into the next generation. Further development is needed to: operationalize the framework to address equity; adapt the framework for a series of specific scenarios such as LMIC settings; map the framework to other determinant frameworks to identify gaps and resolve synonymy and polysemy issues (i.e., construct fallacies) [112]; develop qualitative, mixed, and quantitative methods for application; continue development of validated measures; establish foundations for iterative evolution and strengthen the theories encapsulated in the updated CFIR, including understanding relationships between constructs and with outcomes; and further exploring and establishing semantic identity for each construct [112]. Systematic reviews of empirical findings are needed to further inform or refine theoretical concepts encapsulated within and across constructs and middle-range theories need to be developed to understand interrelationships between constructs [16]. Our team and others have explored the use of causation [113] and relationship coding to highlight how determinants may interact within an implementation project [114]. For example, Kerins et al. developed an adapted version of the CFIR that included construct relationships based on a systematic review of menu labeling implementation projects [45]. Coincidence analysis and other novel analysis methods can be applied to explore clusters of constructs that lead to desired outcomes [115].