Background

Implementation strategies ‘have unparalleled importance in implementation science’ [1] and generating evidence for their effectiveness is a priority for the field [2, 3]. Implementation strategies are defined as ‘methods or techniques used to enhance the adoption, implementation, and sustainability of a clinical program or practice” [1] (Table 1). The opportunity to replicate strategies, tailored to local contexts where needed, and build cumulative knowledge, is partly limited by inconsistent labelling and insufficient description [4]. To address the problem, several taxonomies have been developed to (a) provide a standardised language for describing and reporting interventions and (b) provide a guide for those studying and those seeking to solve implementation problems [5,6,7,8,9]. These taxonomies can be used to describe the components of multifaceted and multilevel implementation interventions with varying degrees of detail [9]. However, there has been little exploration of the potential overlap, duplication, or complementarity between these taxonomies. Failure to consider this could exacerbate the problem of inconsistent labelling that these taxonomies were designed to address by contributing new terminology instead of synthesising and exploring connections between existing labels. It could also limit opportunities to synthesise results across studies that use different taxonomies.

Table 1 Definitions of key terms
Full size table

The Expert Recommendations of Implementation Change (ERIC) compilation was developed to provide a system to classify and organise the myriad of implementation strategies being used in research and practice [7, 15]. Building on a review of the health and mental health literatures, the initial ERIC compilation was developed using a modified Delphi process with a panel of experts from implementation science and clinical practice in North America, many of whom were affiliated with the Veterans Health Administration [15]. The project established consensus among researchers and practitioners for a common nomenclature for 73 implementation strategies [7]. In a subsequent study, concept mapping was used to organise the 73 implementation strategies into nine groups [16]. Since then, the ERIC compilation has been widely used in health implementation research and practice and has been adapted for other settings (e.g., schools [17, 18]), applied to particular types of interventions (e.g., digital mental health interventions [19]), and served as a foundation for expanding the range of specific implementation strategies (e.g., financing strategies [20]). However, several primary studies and reviews continue to note insufficient description of the content and format of popular strategies such as the use of local opinion leaders [21], continuing professional development [22], and development of education materials [23].

Most reporting guidelines recommend clear reporting of intervention content. The AIMD (Aims, Ingredients, Mechanism, Delivery) meta-framework validation project found that 95% of major reporting guidelines recommend the description of intervention ingredients defined as ‘the observable, replicable, and irreducible aspects of an intervention’ [24]. The Template for Intervention Description and Replication (TIDieR) checklist promotes the reporting of procedures and activities used in an intervention, including any activities to enable or support an intervention [25]. Similarly, the Standards for Reporting Implementation Studies (StaRI) Statement recommends dual reporting of the clinical, healthcare, or public health intervention being implemented and the implementation strategy [26]. Recommendations for specifying implementation strategies advise description of ‘the actions, steps, or processes and sequences of behaviour’ needed to enact a strategy using ‘dynamic verb statements’ that ideally should be behaviourally defined a priori [1].

One potentially complementary taxonomy, designed in part to enhance the intervention description, is the behaviour change technique (BCT) taxonomy (v1) [6]. The BCT taxonomy contains 93 discrete techniques. A BCT is defined as an ‘observable, replicable, and irreducible component of an intervention’ that has the potential to change behaviour [6], echoing the TIDieR definition of an intervention ingredient [27]. The BCT taxonomy was first developed in 2008, as a cross-behaviour classification system, using consensus methods and iterative reliability testing with international behaviour change experts [6]. It is now one of the most common classification systems used for describing activities in behaviour change interventions including interventions to support implementation and change professional practice in healthcare [13, 27,28,29,30]. The BCT taxonomy (v1) underpins the behaviour change wheel, a multistage framework for designing behaviour change interventions [31]. Within this approach, it is used to identify intervention content which can best serve intervention functions such as education and enablement.

Both the ERIC compilation and BCT taxonomy have been used to describe the ‘how to’ of implementation, albeit their applications differ in the scope of activities included and level of detail used to describe those activities. The ERIC compilation focusses primarily on team and organisational level strategies while the BCT Taxonomy focusses on change in individual or group behaviour, where the individual/groups may be at different and multiple organisational levels. Given that most implementation efforts require multilevel interventions, it is important to explore whether and how we can combine these approaches. We could identify few studies that have integrated or linked the ERIC compilation and BCT taxonomy to describe implementation in action [32, 33]. This is in contrast to the combined use of determinant frameworks to identify barriers and enablers in implementation science [34]. The proliferation of theories, models, and frameworks is a common criticism of implementation science; over time researchers become familiar with a particular approach and stick with it. To advance implementation science, there are calls for greater integration and comparison of approaches to avoid silos within the field [35].

The objective of our analysis was to examine the extent and nature of the overlap between the ERIC compilation [7] and the BCT taxonomy (v1) [6]. We chose to compare these two taxonomies as they are commonly used approaches to describe implementation interventions originating from different expert groups, they are typically applied separately, and they have varying levels of granularity in their descriptions suggesting scope for integration. By examining the potential links between these taxonomies, we can move from general descriptions of implementation strategies to more detailed and consistent descriptions of their content. By examining the overlap, we can move beyond differences in labels and bridge siloed approaches to designing and reporting implementation interventions.

Methods

Design

We conducted a qualitative content analysis of ERIC implementation strategy definitions to identify the inclusion of, or overlap with, BCTs. The study is reported in line with the COREQ reporting guidelines for qualitative studies (Additional file 1).

Step 1: Content analysis of ERIC strategy descriptions

We used directed content analysis [36]. The BCT taxonomy (v1) was used as the pre-determined codebook used to code ERIC implementation strategy definitions [6]. It contains 93 techniques organised into 16 hierarchical categories. One author (SMcH) coded the definitions of the 73 ERIC strategies and the descriptions in the published in the ERIC paper and accompanying additional files [7]. The coder had completed online training in BCT coding and was experienced in using the BCT taxonomy to specify implementation interventions [29].

A sequential coding process was used, following the steps for coding intervention content outlined in the behaviour change wheel guide to designing interventions [37]. First, each ERIC strategy description was coded to identify one or more BCTs. Segments of text were checked against the BCT definition to assess whether that BCT was present or absent. The coder focussed on action words and verbs in the strategy definition to select the appropriate BCT. Where appropriate, multiple BCTs were assigned to the strategy description. On completion, all 93 BCT labels and definitions were re-considered for each of the strategy definitions. The coder noted her coding rationale throughout the process. Coding was managed using NVivo software (V12).

Step 2: Developing a classification system

Using the results of the coding in step 1, a typology with five a priori categories was developed to categorise the nature of the overlap between ERIC strategies and BCTs. A typology is a formal system for classifying multifaceted complex phenomena according to a set of common conceptual dimensions in order to increase the clarity in defining and comparing complex phenomena [38]. The terms ‘typology’ and ‘taxonomy’ are often used interchangeable in the literature to describe this type of analytic output. We use the term ‘typology’ to avoid confusion with the BCT taxonomy used in the analysis. Also, it has been suggested that typologies are conceptually developed as is the case here, while taxonomies are empirically derived configurations [39].

The five ‘types’ or categories in the typology were developed initially by examining the patterns of overlap or links between ERIC strategies and BCTs (by SMcH). Each category was given a label and coding definition. Categories were reviewed by all authors and further refinements were made to the category labels and definitions. The typology consisted of five categories of overlap (Table 2). First, there were instances of direct 1-1 overlap between an ERIC strategy and a BCT, allowing for some differences in terminology. Second, there were instances where at least one clear BCT was indicated in the ERIC strategy description which could be used to guide initial operationalisation. Third, there were instances where at least one probable BCT(s) was indicated in the ERIC strategy description, that is to say the BCT was logically indicated but was not clearly or explicitly stated. Fourth, there were instances where no BCTs were clearly indicated in the ERIC strategy definition or description. Fifth, some ERIC strategies did not appear to target behaviour change to support implementation; thus, an underlying behavioural target was not clear.

Table 2 Data-driven a priori typology developed to classify the nature and extent of overlap between ERIC strategies and BCTs
Full size table

It is important to note that the typology was applied to describe the pattern of overlap between a BCT and aspects of an ERIC strategy description. During analysis, it became apparent that more than one type of overlap could apply within a single strategy description, depending on the BCT being considered. In these cases, a strategy was categorised as having multiple types of overlap indicated. However, as this was not one of the five a priori categories in the typology, we present this in the results section.

Step 3: Independent rating

The first round of coding (by SMcH) was tabulated in Excel. Three implementation scientists (JP, BP, CL) independently rated their level of agreement with the BCT(s) coded to an ERIC strategy description and the type of overlap assigned. They rated their agreement on a scale of 0 (complete disagreement) to 10 (complete agreement). They also provided suggestions for BCTs to be removed or added, and changes to categorisations. All coders were had experience coding qualitative data and expertise in both frameworks. Average agreement scores were calculated for each BCT and feedback was collated (by SMH). Coded BCTs with agreement scores of ≥7 were deemed to have reached consensus. A revised list of ERIC strategies was compiled, comprising those that had agreement score of <7 or where suggestions were made to reclassify a strategy or add/remove additional BCTs.

Step 4: Review

The remaining undecided ERIC strategies were reviewed by team. Two virtual meetings were held to review discrepancies in agreement and suggested changes to classification and BCT coding. In cases where team members disagreed, we revisited the full ERIC description and examples given in the ERIC compilation and considered its grouping and function as a change strategy (e.g., is the function to educate, to enable, or incentivise practitioners). During team discussions, groups of ERIC strategies were reviewed together to ensure consistency of coding. When reviewing groups of strategies, we identified additional patterns of complementarity between ERIC strategies and BCTs. These patterns were labelled thematically. The results were drafted and circulated to all co-authors and final revisions were made to the classification and BCTs coded.

Data analysis

Across ERIC strategies, we quantified the frequency of BCT occurring (e.g., X BCT occurred 10 times across strategies). We estimated the number of BCT groupings represented in ERIC strategies (n=16 possible hierarchical groupings in the BCT taxonomy). We quantified the total number of ERIC strategies mapped to each of the five types of overlap in the typology (Table 2). The results were compared across the nine groups of strategies in the ERIC compilation: (1) use evaluative and iterative strategies, (2) provide interactive assistance, (3) adapt and tailor to context, (4) develop stakeholder relationships, (5) train and educate stakeholders, (6) support clinicians, (7) engage consumers, (8) utilise financial strategies, and (9) change infrastructure [16].

Results

The results are organized into three sections. First, we summarize the number of BCTs coded to ERIC strategies. Second, we summarize the number of ERIC strategies assigned to the types of overlap. Finally, we describe additional patterns of complementarity between ERIC strategies and BCTs developed during the analysis and provide examples to illustrate those themes. ERIC implementation strategies titles are written in italics (e.g., remind clinicians), and BCTs are reported using quotation marks (e.g., ‘prompts and cues’).

Characteristics of coded BCTs

Of the 73 ERIC strategies analysed, BCTs were coded 150 times. Overall, 41 out of the 93 BCTs (44%) were identified. At least one BCT was coded from 13 of the 16 possible groupings from the BCT Taxonomy; no BCTs were coded from regulation, self-belief, or covert learning groups. The most frequently coded BCT was ‘restructuring the social environment’ (n=18 strategies, coded in 25% of strategies) (Fig. 1).

Fig. 1
figure 1

Frequency of BCTs coded in ERIC implementation strategy descriptions

Full size image

Types of overlap between strategies and BCTs

Table 3 outlines the number of strategies categorised to each type of overlap in the typology. One strategy was categorized as directly overlapping with a BCT; the strategy to change physical structure and equipment equates to ‘restructuring the physical environment’. Most ERIC strategy descriptions (n=64) contained BCTs that were clearly indicated or BCTs that were probable or indicated multiple types of overlap. This pattern was reflected across the nine ERIC strategy groups. Four strategies were categorised as having no BCTs indicated in the strategy definition or detailed description. Four strategies were categorised as not focusing on behaviour change to support implementation.

Table 3 Number of ERIC strategies categorised to each type of overlap, per group, and overall (n=73)
Full size table

Direct 1-1 overlap between ERIC  strategy and BCT

There was one instance where the ERIC implementation strategy directly overlapped with a BCT. The strategy to change physical structure and equipment directly overlapped with ‘restructuring the physical environment’. The strategy and BCT referred to the same change at the same level, albeit using different terminology.

Clear and probable BCTs indicated in ERIC strategy descriptions

Most ERIC strategy descriptions (n=64) contained BCTs that were clearly indicated, BCTs that were probable, or indicated multiple types of overlap (Table 4).

Table 4 ERIC strategies with BCTs coded (where appropriate) and type of overlap indicated
Full size table

For some strategies (n=18), one or more BCTs were clearly indicated in the description. For example, three BCTs were clearly indicated in the description for the ERIC strategy to develop a formal implementation blueprint. An implementation blueprint should include the aim/purpose of the implementation and the scope of the change (e.g., what organizational units are affected), (3) timeframe and milestones, and (4) appropriate performance/progress measures [7]. The BCT ‘action planning’ was clearly indicated as it refers to prompting detailed planning or performance of behaviour and must include at least one of context, frequency, duration, and intensity. ‘Goal-setting (behaviour)’, which refers to setting or agreeing a goal defined in terms of the behaviour to be achieved, was coded given references in the full description to the purpose of implementation, the scope of change, coordinating the blueprint with a fidelity monitoring tool and the types of intervention required at different organisational levels. ‘Goal setting (outcome)’, which refers to setting or agreeing a goal defined in terms of a positive outcome of wanted behaviour, was coded on account of reference to appropriate performance/progress measures.

More often, strategy descriptions indicated BCTs that were probable (n=31), that is logically indicated given the title, definition, and/or description of the strategy but not clearly or explicitly described. The designation of ‘probable’ reflected a lack of specification in the strategy description and the scope for strategies to operationalised with different degrees of change. The BCT ‘restructuring the physical environment’ was most often coded as probable as it was not clear to what extent strategies involved full-scale physical change. This applied to eight strategies: develop and organise quality monitoring systems, develop and implement tools for quality monitoring, centralize technical assistance, use data warehousing techniques, facilitate relay of clinical data to providers, make billing easier, change record systems, create or change credentialing, and/or licensure standards.

The descriptions of some strategies (n=15) indicated multiple types of overlap depending on the BCT being considered and the explicitness of the strategy description. This mix of clear and probable BCTs was evident across the nine ERIC strategy groups.

ERIC strategies with no BCTs indicated or not targeting behaviour change

Four ERIC strategy descriptions had no BCTs indicated explicitly or logically: make training dynamic, assess for readiness and identify barriers and facilitators, promote adaptability, and develop educational materials. Four ERIC strategies were categorised as not focusing on behaviour change for implementation: conduct local needs assessment, develop an implementation glossary, work with educational institutions, and start a dissemination organisation.

Patterns of complementarity

Several themes were identified during the analysis that reflected different patterns of complementarity between ERIC strategies and BCTs.

Within a single strategy, there are different types of overlap with BCTs

Within a single strategy, there were multiple different types of overlap with different BCTs; as mentioned, 15 strategies were a mix of BCTs that were clearly indicated and others that were probable. Strategy descriptions contained explicit text which clearly indicated a BCT and non-specific text which suggested a BCT was probable and logical given the description. For example, the description of the strategy to inform local opinion leaders clearly indicated the BCT ‘identification of self as role model’ while ‘credible source’ was probable depending on the colleagues identified.

‘Broad strategies’ indicate relatively few and broad BCTs but the list of probable BCTs is extensive

In the ERIC compilation, there were what we referred to as ‘broad strategies’ that is descriptions were short and/or not overly specific about what the strategy involved. Most strategies that were broadly defined indicated a single BCT in their description. The BCTs indicated were similarly broad in scope; typically ‘restructuring the physical environment’ or ‘restructuring the social environment’. Restructuring the physical environment was the only BCT indicated in four ERIC strategy descriptions (change service sites, use data warehousing techniques, make billing easier, change record systems). Restructuring the social environment was the only BCT indicated in six ERIC strategy descriptions (build a coalition, create new clinical teams, revise professional roles, develop academic partnerships, recruit, designate and train for leadership, involve patients/consumers to enhance uptake and adherence).

While broad strategies indicated few BCTs in their description, the list of probable BCTs was considered extensive and with limited information in the full description, and it was not possible to code further BCTs as clear or probable. For example, the strategy to involve patients/consumers and family members is described as engaging or including patients/consumers and families in the implementation effort, and the BCT restructuring the social environment was coded as probable. Other BCTs are probable depending on how this strategy is operationalised, but there was no information to inform further coding.

ERIC strategies serve as examples of BCTs operationalised for implementation

ERIC strategies, including some of those mentioned above, provide contextual examples of how BCTs could be operationalised for implementation. This pattern was evident for BCTs that we considered to be broadly defined, and these BCTs were among the most frequently coded; ‘restructuring the social environment’ (18 strategies: 7 clear indications and 10 probable indications), ‘credible source’ (12 strategies: 1 clear indication and 11 probable), restructuring the physical environment (9 strategies: 1 direct overlap, 1 clear, 8 probable), and ‘social support (practical)’ (8 strategies: 3 clear and 5 probable). The ERIC strategies provided examples as to how the same BCT could be operationalised in different ways to support implementation.

Presence of some BCTs is dependent on form of delivery

The presence/absence of some BCTs was dependent on the form of delivery indicated in the strategy description. For example, the BCT most frequently coded as probable was ‘credible source’ (for 12 strategies) as the presence/absence of this BCT is dependent on who delivers the information and whether they are credible to the target population. This BCT was coded as probable for all strategies except providing ongoing consultation which refers explicitly to the use of experts in the strategies to support implementation of the innovation.

Other ERIC strategy descriptions make suggestions about or imply the form of delivery. Strategies involving the BCT restructuring the social environment suggest operationalisation using either group delivery methods (e.g., creating a learning collaborative) or individual level delivery (e.g., providing local technical assistance). Other strategies suggest the setting where the strategy would be delivered (e.g., visit other sites) or delivery features relating to the provider (e.g., identify and prepare local champions, provide clinical supervision). The strategy make training dynamic refers to several elements of the form of delivery including delivery format and intensity which could implicate BCTs in the operationalisation of this strategy.

Some ERIC strategies are steps in the implementation process and target collective cognition at the system level

Eight ERIC strategies, categorised as having no BCTs indicated or not targeting behaviour change, demonstrated that some strategies are part of early phases of planning for implementation. For example, the strategies to assess for readiness and identify barriers and enablers and promote adaptability are overlapping/interrelated processes that could be used during pre-implementation to inform strategy or BCT selection. Their purpose is not to directly influence behaviour for implementation at the service-delivery level but rather to inform decision-makers and change minds at a system level.

Other strategies, which we categorised as not targeting behaviour change, are system-level processes focussed on information gathering or sharing (strategies: work with educational institutions, start a dissemination organisation, develop an implementation glossary, conduct local needs assessment). These strategies could also be part of the exploration or preparation stages of implementation but are not informing the selection of strategies for implementation or execution of the innovation itself.

Finally, there is a sequence inherent in some strategies without (and sometimes with) BCTs which would likely be coupled in practice. For example, developing educational materials alone does not indicate any BCT in its description but would most likely be coupled with distributing educational materials, a strategy with probable BCTs to shape knowledge and provide information about consequences (‘instruction on how to perform behaviour’ and ‘information about health consequences’).

Discussion

The aim of this study was to examine the extent and nature of the overlap between the ERIC compilation of implementation strategies and the BCT Taxonomy. Overall, we identified complementarity rather than redundancy when integrating these two taxonomies. There was only one instance where the ERIC implementation strategy directly overlapped with a BCT; the ERIC strategy to change physical structure and equipment overlapped directly with the BCT ‘restructuring the physical environment’. Most strategy descriptions had BCTs that were clearly indicated, BCTs that were probable but not explicitly described, or indicated both types of overlap within a single strategy. This study can be considered a foundational step to move from general descriptions of implementation strategies to full and consistent description of actions.

Enhancing levels of specificity

Some ERIC strategy descriptions contained BCTs that are clearly indicated but more that are probable, depending on how the strategy is operationalised. With limited information in the strategy description, it was not feasible to code an exhaustive list of all BCTs. A number of studies have highlighted that ERIC strategies vary in their level of specificity [40, 41]. It has been suggested that taxonomies such as the ERIC compilation do not possess the granularity and specificity the BCT Taxonomy contains [27]. Our results challenge this assumption as we identified mixed levels of granularity in both taxonomies. Similar to other studies [13, 27], in our analysis broadly defined BCTs such as ‘restructuring the social environment’, ‘restructuring the physical environment’, and ‘social support (practical)’ were among the most frequently coded in ERIC strategy descriptions. Some ERIC strategies were more granular, describing what was being restructured in the environment or, the nature of social support provided to implementers. As such, they provided contextual examples of how broad BCTs could be operationalised for implementation. The BCT taxonomy is intended to apply to any behaviour so these instances from the ERIC compilation may provide examples for inclusion/cross-referencing with future versions of the BCT taxonomy. The ERIC compilation could also serve as a basis for identifying additional BCTs that are not reflected in version one of the BCT taxonomy.

Overall, 44% of all possible BCTs were identified in ERIC strategy descriptions. For the remaining 56% of BCTs, it may not be logical to ever consider them in any implementation strategy (e.g., biofeedback) or they may apply to broad strategies that are currently under-specified. Considering how those underutilised BCTs could be incorporated may be an opportunity to enhance the description and novelty of some ERIC strategies.

Moving towards mechanisms of action

Specifying the BCTs indicated in ERIC strategy descriptions provides a path by which we could begin to understand mechanisms of action. In implementation science, a mechanism of action is defined as a ‘process or event through which an implementation strategy operates to affect desired implementation outcomes’ [42] Using the Theory and Technique Tool developed to link BCTs to mechanisms of action [43,44,45], we can suggest mechanisms of action linked to BCTs that are indicated in ERIC strategies. For example, four BCTs were clearly indicated in the description of the ERIC strategy to provide ongoing consultation. There is evidence of a link between each of the four BCTs and at least one mechanism of action (MoA): ‘social support (unspecified)’/social influences (MoA), ‘credible source’/attitude toward behaviour and general attitudes and beliefs (MoAs), ‘feedback on behaviour’/motivation and feedback processes (MoAs), and ‘instruction on how to perform the behaviour’/knowledge, and skills and beliefs about capabilities (MoAs). Strategy-BCT-MoA linkages could provide the building blocks for testable causal models that can be refined over time. The MoA linkages mentioned here are based on those described in behaviour change interventions and coded using the Theoretical Domains Framework and additional MoA constructs from theories of behaviour change [43]. Mechanisms can operate at different levels including the organisational, community or macro policy level [42] and mechanisms at those levels may not be captured sufficiently by current tools which concentrate on the individual level. Efforts are underway to develop a research agenda to advance understanding of the mechanisms of implementation strategies [46].

Form of delivery

While we did not formally code all elements of form of delivery, some ERIC strategy descriptions suggest or imply forms of delivery and as a result additional BCTs were coded. Delineating and describing the strategy and form of delivery is an important step when designing and reporting implementation interventions. Even with this distinction, the taxonomies examined in this study are not intended to deliver intervention content ready to use ‘off the shelf’. Other dimensions that also need to be clarified to fully operationalise an implementation strategy include the actors delivering the strategy, targets of the action, temporality, and dose (frequency and intensity) [1, 47].

The presence or absence of BCTs in a strategy description is not an indicator of strategy effectiveness. Furthermore, for some techniques, there is evidence of when and how they should be applied to ensure theoretical coherence and effectiveness, these characteristics should be persevered during operationalisation [47, 48]. In this analysis ‘credible source’ was one of the most common BCTs indicated in ERIC strategy descriptions. It was primarily coded as a probable BCT as it depends on the source of information and whether they are identified as credible by recipients of the information. According to Peters and colleagues there are a number of parameters of effectiveness for modelling, an implicit part of this BCT [48]: the recipient must attend to the communication, must remember it, and must have a sufficient skill to perform the desired behaviour; then, the recipients must identify with the model; the model must be positively reinforced for the desirable behaviour; and the model should be a coping model as opposed to a mastery model. When operationalising an ERIC strategy, it is not enough to consider whether the BCT is present or absent, we also need to consider how it is applied in practice to ensure it is functioning as intended. This is essential for understanding the mechanisms of action and inaction when assessing the effectiveness of implementation strategies.

Deconstructing high-level and preparatory strategies

Some may consider that the BCT taxonomy and ERIC compilation focus, to varying degrees, on different levels of individual and organisational change, and reflect tensions between the traditions of behaviour-change-oriented health psychology and system-oriented organizational psychology/change management. We believe the separate application of the BCT taxonomy and ERIC compilation may constrain our thinking about how to best draw from both to describe implementation interventions. In this analysis, we demonstrate if and how the BCT taxonomy applies across all levels of ERIC strategies. Strategies can be deployed down the implementation chain to influence actions among frontline implementers and upwards to influence collective decision-making or action at an organisational and system level. Our analysis deconstructs some of the broad organisational strategies into behaviour change techniques to support implementation at these higher levels. It is argued better reporting of who needs to do what differently at a higher level would more comprehensively capture the multilevel changes involved in implementation [13]. In our analysis, nine ERIC strategies were categorised as having no BCTs indicated in their description or were not targeting behaviour change for the implementation of an innovation. The results align with recent efforts to organise implementation strategies into categories related to their timing [40] and target [49]. In terms of timing, strategies such as assessing readiness and tailoring could be used to prepare for the execution of the innovation or to inform the selection of other strategies. Vax et al. assigned these strategies, among others, to readiness stages of pre-contemplation, contemplation, and preparation [40]. In terms of target, we classified four strategies which did not target behaviour change but instead focussed on information gathering or information sharing at an organisational level (work with educational institutions, start a dissemination organisation, develop an implementation glossary, conduct local needs assessment). This grouping reflects aspects of the classification system proposed by Leeman and colleagues in which strategies are organised according to the strategy actors and action targets [49]. Within this system, dissemination strategies are one class of strategy which targets awareness, knowledge, attitudes, and intention to adopt an innovation. Thus, failure to identify BCTs in every strategy does not necessarily reflect a weakness of a strategy but may reflect a difference in the timing or target of a strategy.

Some of the ERIC strategies in this group may more closely reflect policy categories proposed in the behaviour change wheel approach to enable interventions [31]. Although not a formal part of this analysis, ERIC strategies such as working with educational institutions and starting a dissemination organisation appear to overlap with the policy category ‘environmental/social planning’, defined as designing and/or controlling the physical or social environment to enable interventions. Other policy categories in the BCW approach map to ERIC strategies in which specific BCTs were probable, for example, the ERIC strategy to use mass media and the policy category ‘communication/marketing’. Different aspects of the BCW approach (BCTs, intervention functions, policy categories) could be linked to ERIC strategies depending on the level of granularity required. This reflects the range of strategies at different levels of change outlined in the ERIC compilation.

Strengths and limitations

We systematically coded ERIC strategy descriptions using the BCT Taxonomy and classified the level of overlap using a de novo typology. Other classification systems are used to design and specify implementation interventions which may also overlap with these taxonomies [8]. While guidance on coding intervention content using BCTs warns against making inferences or assumptions that a BCT is present unless there is evidence that it has been delivered [37], the purpose of this analysis was to examine the extent and nature of overlap between BCTs and strategies. Therefore, we considered it necessary to code probable BCTs to highlight the lack of specification in certain strategy descriptions. Additional BCTs beyond those coded in this analysis are possible as part of any strategy, depending on how it is operationalised in a given context.

One researcher coded all the ERIC strategy descriptions using the BCT taxonomy. To minimise the influence of researcher assumptions or familiarity with certain strategies, three analysts reviewed and rated their agreement with the coding. They also provided suggestions of BCTs that should be removed or added.

We recognize that depending on how any given ERIC strategy is operationalized, there could be many other potential BCTs possible. Instead of making assumptions, we opted for a conservative approach in our analysis and coded the strategy definition and detailed description in the ERIC compilation only. It is likely these descriptions are a starting point for how a strategy could be operationalised, rather than an exhaustive or prescriptive description. Primary studies of implementation interventions are another source of information on how strategies are operationalised by researchers and implementers. BCT-coding intervention descriptions in future systematic reviews could be another step in the process of specifying ERIC strategies. Researchers have begun to synthesize the content of commonly used strategies [22, 27]. BCT coding is dependent on the richness of the strategy description [13] and the specificity of the BCT definitions. As mentioned previously, broadly defined BCTs were the most frequently coded in ERIC strategy descriptions which could reflect the ease with which their definitions are identifiable in the text.

The results of this study have a number of practical implications. First, the study distinguishes between ERIC strategies based on the extent to which they are similar to or indicate BCTs and thus can be readily operationalised. It places strategies on a continuum from clearly specified to those requiring more work before application. This could be potentially useful to researchers and practitioners trying to design, replicate, scale, and spread implementation interventions. These frameworks could be integrated iteratively in several ways. For example, designers could begin with the ERIC compilation to name strategies in a language that is accessible to stakeholders and specify the activities within those strategies using the BCT taxonomy, using the coding in this analysis as a starting point. Alternatively, designers using the behaviour change wheel approach to intervention design could refer to the ERIC compilation to see how certain intervention functions or BCTs could be operationalised in an implementation context. Second, the results could also inform fidelity assessment by suggesting observable BCTs that could be monitored as strategies are deployed [50, 51]. Finally, each taxonomy can make a unique contribution that can be leveraged to enhance the reporting, replication, and synthesis of strategies. The ERIC compilation provides an accessible language for community partners and practitioners designing implementation strategies. Integrating the BCT taxonomy encourages consideration of the actions and targets of these strategies.

Conclusion

The myriad of theories, models, and frameworks is an accepted part of implementation science. However, there is increasing recognition of the opportunity and needs to combine these tools [52]. This study highlights the complementarity rather than redundancy that can come from combining the ERIC compilation and BCT taxonomy. Each taxonomy can make a unique contribution to enhance the reporting, replication, and synthesis of strategies. More efforts to integrate these taxonomies will aid strategy developers and build links between existing silos in implementation science.