FRAME-IS provides a template for systematically tracking modifications to an implementation strategy according to discrete modules (4). The modules report the: (1) “Description” of the EBP, implementation strategy, and brief modification description. (2) “What” (type) of strategy modification made according to four categories: content of the strategy itself or its delivery, evaluation for the way the strategy is evaluated, training in terms of the manner of implementer training, and context of how the strategy is delivered, specifically changes to format of delivery, setting of delivery, personnel delivering the strategy, or population targeted for delivery. (3) “Nature” of modification, which refers to the extent and intensity of the modification with pre-specified codes such as tailoring; adding elements; removing elements; lengthening, substituting, spreading, or integrating another strategy into the implementation strategy in primary use; loosening structure; and others. (4) “Rationale” behind the modification. This defines both the goal and level of modification. (4a) Goal codes are sorted by implementation outcomes such as intending to improve reach, effectiveness, adoption, acceptability, appropriateness, feasibility, fidelity, cost, sustainability, or equity. (4b) Level codes dictate identifying the socioecological levels influencing the decision to modify for sociopolitical, organizational, implementer, provider, or patient reasons. (5) “When” (timing) of tailoring the strategy components (pre-implementation, implementation, scale-up, or maintenance) and whether the modification was planned/proactive, planned/reactive, or unplanned/reactive. (6) “Who” defined or is involved in the decision to modify the strategy: political leaders, program leaders, funder, implementer, researcher, providers, community members, or patients.

Consistent with other implementation scientists (14, 15, 19), we use “tailoring” to reference altering strategies to fit context and “modification” as a generic term for change. There is debate but little consensus regarding the scope and definitions of the terms “adaptation,” “tailoring,” and “modification.” Notably, some researchers may classify AD tailoring as an “adaptive implementation strategy” (2, 20).

To understand and guide modifications, we agreed on a set of components common to AD. Due to the vast range of AD conceptualizations in the literature (5–8, 10–12, 21, 22), the decision was made to identify common components from publicly available implementation and training materials (23–25). Priority was given to AD components named by the National Resource Center for Academic Detailing (NaRCAD) and the VHA Pharmacy Benefits Management's Academic Detailing Services (ADS), as both are national leaders in AD development and practice. One project investigator (AMD) cross-walked these training materials to identify 18 components that tend to exist in AD programs both within and beyond VHA. Components were grouped by the FRAME-IS module 2 (“What”) categories (content, evaluation, training, and context) to create an initial matrix of the AD common components. The 18 components were further typed according to the Expert Recommendations for Implementing Change (ERIC) taxonomy of implementation strategies (26) using a coding specification tool (27). ERIC is a sorted list of implementation strategies derived from expert consensus and concept mapping. The phrasing used in ERIC has become common nomenclature in implementation science. ERIC terms are widely used to ensure implementation tools are generalizable across contexts and innovations, and to assist future meta-analyses of implementation strategy effectiveness. Coding strategy components to ERIC is not prescribed by FRAME-IS developers but was conducted here to standardize strategy terms. By standardizing AD component definitions to a general taxonomy, this will enable future AD analyses and ensure that AD modifications are consistently defined.

An overarching AD program was developed to support AD projects for three hybrid III cluster randomized implementation trials (28) and a non-randomized intervention project as part of the Maintaining Implementation through Dynamic Adaptations (MIDAS) Quality Enhancement Research Initiative (QUERI) center. Within each of these projects, AD was used as a strategy to support a specific quality improvement program (29) within participating healthcare sites across VHA. The goals for the four projects were: to reduce inappropriate polypharmacy among older adults, improve safe use of Direct Oral Anticoagulants (DOACs), promote Cognitive Behavioral Therapy for Insomnia (CBT-I) as a first-line treatment over medications, and increase referrals to a telehealth suicide prevention program. The MIDAS AD program deployed in 23 sites across VHA from 2021–2024 to support the four projects.

The MIDAS team developed an AD program to support these projects. The AD program needed to meet four key criteria and work within immutable constraints. First, the program must be able to support the broad range of project topics. Second, we were not able to support separate detailers with content expertise for each of the four project topics. Third, national policies prevented us from employing VHA providers or funding VHA ADS to staff the projects (6, 9, 10, 30). Fourth, rigorous evaluation required consistency across projects, clear definition of AD strategies used, clear descriptions of adaptations, and clearly defined outcomes. MIDAS AD program development was informed by a comprehensive review of existing AD program materials, particularly from the VHA ADS (23), NaRCAD (24), and existing research publications (5–7, 11, 12, 21) and from our experience with quality improvement and implementation science interventions. Based on these sources, we created a master protocol specifying project content, evaluation design, and detailer training tailored to MIDAS's needs (Supplementary Material A). Among these materials are specifications of evaluation tools including a fidelity assessment using self-ratings and peer review (Supplementary Material B), a detailer script (Supplementary Material C), and a participant satisfaction survey (Supplementary Material D).

To classify MIDAS modifications, we followed five steps for rigorous document analysis (31–34). First, two trained detailers (GH, MF) independently defined the MIDAS AD program components by conducting a MIDAS document review using modified directed content analysis (31, 32). We reviewed materials from across the development and implementation of the four projects including protocol drafts, training materials, fidelity assessment entries (such as post-detailing session notes and ratings), and peer reviewer notes. Second, we classified components according to the FRAME-IS modules. One investigator (AMD) applied predetermined FRAME-IS module codes (4) to component descriptions for MIDAS AD, then re-coded components based on ERIC labels and definitions. Labels were applied based on identifying the best fit between detailed MIDAS AD component descriptions and ERIC labels provided by the developers (26, 27), with an emphasis on the function of the component. As needed, the investigator consulted with an ERIC developer (LD) for accuracy. Third, two collaborators (LE, MS) iteratively reviewed the MIDAS AD component descriptions, categorizations, and codes as a form of validity check via peer debriefing (33). Fourth, for reliability checks (34), content and codes were reviewed and revised in five reflexivity discussions held over the course of three months with the investigative team (GH, MF, AMD, LE, MS). Finally, the final reporting structure was reviewed and approved by two project leaders (JBS, LD) as representative of the MIDAS AD program.

The AD documentation system used FRAME-IS codes for each module (Table 1). These definitions were extracted from the method developers (4) and worded to be AD-specific. The AD tailoring definitions are the codes within each module we suggest for use when documenting AD program customization to context, as in MIDAS (below). Supplementary Material E provides a blank AD FRAME-IS template for other programs to describe their AD components and modifications.

We identified common core components and subcomponents of AD programs structured by the FRAME-IS “what” module (Table 2). Content comprises three core components: campaign development, campaign initiation and recruitment, and in-session delivery. Each component has subcomponents; for example, developing key messages, meeting with clinical staff, and delivering a presentation about the topic. The only common component for AD evaluation is collecting outreach process data to document which providers were educated, the duration and timing of those educational outreach sessions, and the content of the session such as barriers brought up by the provider (for example, a necessary medication is not in their order set) and whether the detailer secured the provider's verbal commitment to change their prescribing behavior to be more consistent with the innovation. The three common components for training dictate having detailers attend a basic skills workshop, meeting with detailer colleagues or consultants for consultation, and practicing via role play. Finally, four common components of context show that AD outreach visits are usually delivered one-on-one, in-person, to front-line staff, and by someone with a background in a health-related field.

Mapping each component to ERIC reveals that four AD common components pertain to conducting educational outreach visits, three to making training dynamic, and one each to seven other strategies: develop a formal implementation blueprint, tailor strategies, distribute educational materials, develop and implement tools for quality monitoring, train facilitator, provide ongoing consultation, and remind clinicians (Table 2). Taken together, Table 2 shows that, as commonly used, AD programs are a multi-faceted implementation strategy that bundle together a range of educational and interactive components.

Key to beginning the program was identifying potential modifications of AD components, questions to consider for our program, and the MIDAS-specific context from which all decisions were made (Table 3). We strove to apply a consistent MIDAS AD strategy to all four EBP interventions. While based on our experiences developing the AD program, we structured this paper around providing guiding questions and points of consideration for future AD program developers when making tailoring decisions.

The list of tailored and additional MIDAS AD components is presented in Table 4 along with the reason for tailoring or adding. Out of eighteen common AD components, MIDAS AD program developers retained 12 components with no modification; these are therefore not included in Table 4 (but can be seen in Table 2). MIDAS tailored (i.e., substituted or refined) six of the common components. The MIDAS AD program developers then added an additional seven components by integrating strategies not usually included in common AD. Added components not otherwise found in common AD are denoted with an obelisk symbol. No components of common AD were removed from MIDAS AD.

In Table 4, the FRAME-IS codes show the category of “what” was modified (Module 2), provide a “description” of the modification (Module 1) with ERIC labels, and the “nature” (Module 3) and “rationale” (Module 4) of each modification. Two modules are not included in Table 4: Modules 6 and 7. The “when” (Module 5) was always during pre-implementation and planning was always planned/proactive, except for the use of group visits when the modification was during implementation and the planning was reactive. The “who” (Module 6) of decision-makers was always the MIDAS team.

Per Module 4 (“rationale”), reasons for modification primarily stemmed from needs at the provider and organizational levels, and in some cases the implementer level. The goals of each modification varied and aimed primarily to improve appropriateness, acceptability, adoption, and reach. Per ERIC type (shown in italics in the Module 1 column), the MIDAS AD program tailored components related primarily to the method of conducting educational outreach visits and developing and implementing tools for quality monitoring of AD.

Remaining results are organized according to the “nature” of modification (Module 3), first by those tailored (Module 1, no symbol in Table 4) and then by the integration of other strategies into AD (Module 1, denoted with obelisk symbol in Table 4). Underlined words indicate the categories of “what” (Module 2) was changed with italics to draw attention to specific codes.

Strategy tailoring is necessary to fit implementation processes to specific needs (3, 15, 37). Yet, changing an established strategy raises key questions: how much can it change? At what point has it changed so much that the critical elements are gone? Given the wide range of conceptualizations regarding AD key components and activities (5–8, 11, 12), our work to describe common AD components and their tailoring is an important step. One previous review that attempted to define AD characteristics found a 36.5%–100% variation between AD programs, particularly regarding communication components (12). A subsequent Delphi study (11) specified six key features of AD, among them an emphasis that AD focus on changing provider behavior. Despite these attempts at consensus, disagreement remains. For example, some argue AD is most effective in one-on-one, face-to-face sessions (6, 7, 9), while others report effective virtual delivery (10), or that—in contrast to the Delphi study findings—AD is designed for making workflow process improvements (21). This variation in what defines AD calls for future studies to document and make available programmatic components, resources used for development, and modifications from those resources. Our work starts to fill this aim.

While we propose that choices made for MIDAS may represent areas that are core (i.e., content) and modifiable (i.e., evaluation, context), we cannot conclusively say this is true across contexts. However, based on our literature reviews and experience across four projects, we can provide a template and guidance for tailoring AD programs. When we developed and delivered AD, there were many potential decision points for how to tailor the program. We describe these guiding questions to provide options for tailoring AD depending on context. Table 3 is a resource for AD developers to consider as they embark on their own implementation projects.

While several resources are freely available to learn about AD (10–12, 23, 24), we can only find one guidebook for creating and implementing a new AD program (25). To build the public library on AD development and study, MIDAS AD protocols and evaluation tools are available as supplementary materials. We strongly suggest that AD programs use the FRAME-IS template in Supplementary Material E to track their modifications towards building our understanding of which AD components are core and which are modifiable, in effort towards the broader study of implementation strategy bundles.

Strengths of this report are the use of extensive document review and coding schemes supported by a team of collaborators with strong implementation science experience, the application of a systematic method for reporting modifications, and the demonstration of a generalizable suite of modifications that applied to four projects with different EBPs. A limitation for non-VHA settings is that there may be more variations on common components than we identified, especially given the large number of small AD programs nationwide. Notably, in non-VHA settings the payment and reimbursement systems provide different prescribing incentives and adherence metrics. Another limitation is our inability to assess which modifications were effective for their programs.