We employed implementation mapping guided by the TICD framework. Implementation mapping is a systematic process based on five tasks to develop or select strategies for the implementation of evidence-based practice [1]. The TICD framework was chosen because (1) it is an implementation science framework designed to guide efforts to improve care delivery; (2) it is based on a systematic review of 12 prior frameworks; (3) it has been widely used with more than 700 citations in the literature; and (4) it includes a patient factors domain [9]. The TICD includes 57 practice determinants across 7 domains [9]. In the following sections, we describe the implementation mapping tasks used to select and specify implementation strategies for risk-aligned bladder cancer surveillance (Fig. 1). The final task is an ongoing comprehensive evaluation of implementation outcomes to measure the impact of the strategies being pilot tested in four VA sites.

Implementation Mapping process as applied to the current project. The left column shows the specific Implementation Mapping tasks and the right column shows an example strategy that was selected and specified using Implementation Mapping

The implementation mapping process was based on a needs assessment, for which we identified facilitators and barriers of risk-aligned bladder cancer surveillance. This was done via staff interviews across six Department of Veterans Affairs (VA) sites and has recently been published [10]. In this prior mixed-methods work, we used a quantitative approach to identify the six VA sites. Two sites commonly provided risk-aligned surveillance and four sites were deemed to have room for improvement, defined as sites which performed high intensity surveillance for low-risk and low intensity surveillance for high-risk early stage bladder cancer [10]. We purposively sampled 14 participants (6 providers, 2 nurses, 2 schedulers, 4 leaders) from risk-aligned sites and 26 participants (12 providers, 3 nurses, 3 schedulers, 8 leaders) from sites with room for improvement for semi-structured interviews. In sites with room for improvement, we found that absence of routines to incorporate risk-aligned surveillance into clinical workflow was a salient determinant contributing to less risk-aligned surveillance. Irrespective of site type, we found a lack of knowledge of guideline recommendations by nurses and providers, including attending and resident physicians, and advanced practice providers. We concluded that future implementation strategies will need to address the lack of routines to incorporate risk-aligned surveillance into clinical workflow, potentially via reminders or templates. In addition, implementation strategies addressing knowledge and resources could likely contribute to more risk-aligned surveillance [10].

This task entailed identification of two types of objectives, performance objectives and change objectives. Performance objectives are observable actions that need to be performed to provide risk-aligned bladder cancer surveillance and define “who has to do what” [11]. Change objectives are defined by what needs to be changed related to a specific determinant to accomplish the performance objective [11].

The performance objectives were organized by TICD framework domains and determinants and then by employee type (provider, nurse, scheduler, leader, patient). Performance objectives were formulated based on qualitative data from the prior staff interviews [10] and then reviewed and discussed in group sessions with the research team to assure they align with the qualitative data. These performance objectives were then discussed with one patient advisory group and one physician advisory group to solicit input.

To formulate change objectives, we then created a change matrix. Each row represented a specific performance objective. The columns listed the 57 determinants from the TICD framework [9]. In each cell of the change matrix, we denoted the change objective, i.e., what needs to be changed to accomplish the performance objective. Directionality was taken into account, i.e., the change objective had to logically affect the performance objective. To formulate the change objectives, two authors (AOO or FRS) independently filled in a first objective into applicable cells. Next, they reviewed each others’ work and then met to discuss edits, including addition of change objectives that were not identified on the initial pass, or changing cells to not being related to a performance objective after discussion. The change matrix was then reviewed by the research team and edited until consensus was reached on the content for each cell of the change matrix. From this final change matrix, we then obtained the unique change objectives. The change objectives were then reduced by combining change objectives that had conceptually overlapping topics.

First, we developed an implementation strategy matrix linking unique change objectives (rows) to potential implementation strategies (columns). Implementation strategies were obtained and labeled according to the ERIC [12]. We reviewed all 73 ERIC strategies and excluded those that were not applicable for inclusion in our project (e.g., not feasible within confines or budget of the project, not appropriate for the context of working within VA, already completed as part of the mixed-methods needs assessment or as part of the research project development). Specifically, one author (FRS) performed an initial assessment of which ERIC strategies may not be applicable for inclusion in our project and specified reasons for exclusion. These decisions were then reviewed, discussed, and revised in meetings with two additional authors (AOO, LZ), and then with the entire research team. All decisions were documented along with reasons for exclusion (see methods journal tab in final implementation strategy matrix in Supplementary Material). Next, we wrote strategy-specific statements in each cell of the matrix on how each strategy could potentially affect a change objective. These statements were discussed by the team, and we came to consensus on the content for each cell of the implementation strategy matrix. The potential implementation strategies were then discussed with one patient advisory group and one physician advisory group to solicit input.

To prioritize strategies, we then created a plot from this matrix, showing how many and which change objectives are being addressed by each strategy. We categorized strategies into broad versus narrow scope based on whether or not they addressed eleven or more change objectives. Eleven or more was chosen as a cut-point because the median number of change objectives addressed by the strategies was 10.5. Next, we evaluated 3 factors for each strategy: (1) broad versus narrow scope based on number of change objectives addressed, (2) qualitative assessment of the required time commitment from local staff, and (3) likely impact of the strategy in our clinical setting based on the available evidence from prior studies. When drawing conclusions about likely impact, we specifically considered the clinical setting in which the prior studies were conducted and whether that setting was comparable to the setting of the current study. As a final task, we decided which strategy should be included or excluded, and reasons for inclusion and exclusion were documented along with the theoretical change methods driving each strategy [13].

This task included operationalization and specification of each implementation strategy according to seven dimensions described by Proctor, including actors, actions, targets of actions, temporality, dose, implementation outcomes affected, and theoretical justification [14]. In addition, we produced implementation materials for each strategy (e.g., cheat sheets, posters, templates for the electronic medical record) with corresponding implementation activities. These were documented, including fidelity measures (i.e., non-modifiable components of each strategy) and allowable adaptations (i.e., allowable modifications based on local needs). Given the iterative nature of implementation mapping, we occasionally readdressed a prior task throughout the mapping process.