Using the EPOC checklist as a starting point to code strategies described in abstracts from the 2008 and 2009 G-I-N Conferences, we drafted an implementation taxonomy, which was subsequently used to classify implementation strategies described in the abstracts from the implementation stream of the 2010 G-I-N Conference. Although all four domains (professional, financial, organisation, and regulatory) were utilised in the assessment of the abstracts, our draft taxonomy proved inadequate when attempting to further classify some of the implementation strategies using the subgroups within these domains.
All of the domains in our draft taxonomy were successfully applied to the abstracts, which supports their usefulness as broad categorisations. However, further classification of the abstracts using the categories within each domain yielded a high number of strategies that were classified as ‘other’. Use of the ‘other’ category is likely due to the level of detail available in abstracts, which could be overcome by reviewing a full-text article. However, it may also point to the need for better granularity within the taxonomy to allow more accurate classification of strategies. Further investigation is required to identify and understand the types of strategies that have not been included in the current version of the EPOC taxonomy.
The language and clarity of our draft taxonomy were assessed using feedback received from attendees at a workshop on the taxonomy, which was held at the G-I-N 2011 meeting. In general, the taxonomy was found to be sufficiently detailed to enable classification. However, in some instances, items were found to be less explicit. For example, the strategy ‘identify barriers’ does not specify the type of barrier or the method used for identification.
A number of design-related issues were also identified by the authors when the draft taxonomy was used to classify the abstracts. The level of detail included in each subgroup of the taxonomy was, in some instances, not sufficiently detailed to distinguish differences between two items on the checklist. For example, problems arose when trying to differentiate reminders and recalls from other forms of electronic implementation and when differentiating audit and feedback from other quality improvement activities. These issues support the need for improving the granularity of the taxonomy.
The focus on interventions for health professionals in this taxonomy also overlooks other groups of individuals, such as patients and key decision makers, who play a critical role in the health system. There is an increasing recognition of the benefits of patient involvement in healthcare, particularly in patient-centred health care and shared decision-making [27–30]. As such, it is important to consider implementation strategies aimed at improving the role of consumers in the implementation of guidelines. For example, the draft taxonomy, at present, provides only seven possibilities for patient-directed strategies, which fit within the ‘Financial’ or ‘Structural’ domains, while a substantial portion of the taxonomy describes strategies for health professionals. The scarcity of consumer-related implementation strategies in the EPOC taxonomy may be supplemented with strategies identified by the EPOC Consumers and Communication Group and published via the Cochrane Library, as well as other current literature.
In addition to health professionals and patients, knowledge translation in health care involves other groups such as policy makers, government officials, membership bodies, and non-government organisations. The present structure of the taxonomy does not permit consideration of the specific roles of these groups within the healthcare system and, as such, would require revision to ensure that implementation strategies aimed at behaviour change in a range of participants within the healthcare system are included in the framework.
Following the process of classifying implementation strategies using our draft taxonomy, we found that the majority of abstracts reported the use of two or more strategies. The use of multiple strategies by some guideline implementers (with some abstracts describing six or more activities) suggests that implementers are cautious about confining themselves to a select number of strategies. Alternatively, it suggests that the use of multiple strategies is assumed to be more effective because it addresses more barriers . Nevertheless, the use of multiple strategies highlights the difficulty in determining causality and, in turn, the effectiveness of individual implementation strategies when more than one is used.
It is also currently unclear how guideline implementers select the number of strategies required to facilitate the implementation of a guideline. While a number of studies have demonstrated improved patient outcomes after multifaceted interventions , more evidence is required to demonstrate the optimal number or combination of strategies for guideline implementation and the circumstances under which the number of strategies is most beneficial.
Professional and organisational strategies
The results of our study showed that the commonly used implementation strategies described within the abstracts were either professional or organisational. These results may reflect the fact that: there is more evidence to support the use of these strategies; these strategies are more accessible to guideline implementers in general; or these strategies are more accessible to the cohort of presenters at the G-I-N Conference who are likely to be researchers . Conversely, the low utilisation of financial and regulatory interventions in the selected abstracts may reflect the fact that research involving these implementation strategies is not being sufficiently reported in scientific fora. Low usage of financial and regulatory interventions may also reflect the capabilities of the authors of these research papers, who may be clinicians or academics working in healthcare settings, to access these strategies.
Limitations of the study
We identified a few limitations in our study. The review of abstracts provided a lower level of detail compared to that of a full-text article. In some instances, this lower level of detail made it difficult to accurately classify the implementation strategies described in the abstract. In these instances, it is difficult to ascertain if the inability to classify a strategy is due to the detail and accuracy of the abstract or the draft taxonomy itself. Nevertheless, abstracts are usually the first ‘screening point’ for published papers and, therefore, it is important that abstracts, as well as full-text articles, include a sufficient level of detail to inform the reader about strategies that were used.
Another limitation of this study is the use of abstracts from the G-I-N Conference. Applying the taxonomy on these abstracts may yield optimistic depictions of guideline implementation, because the participants at this conference are likely to be more conscientious about the implementation of guidelines than people in the broader academic or healthcare community. Nevertheless, these abstracts provide a good basis for testing our draft taxonomy as they are screened and selected by an expert panel and, therefore, the reporting of details in these abstracts is likely to be of a higher standard than those published in the wider literature. Further development of our draft taxonomy will include its application on full-text articles, which would greatly enhance its utility, validity, and relevance.
This first draft of the taxonomy utilised a flat structure to create a blueprint of a potential hierarchical format of the taxonomy. Knowledge translation can, however, occur in three dimensions: linear, cyclic, and multi-dimensional , and guideline implementation requires a complex assessment of considerations of different interventions, levels of intervention, target groups, and contextual differences [8, 11]. Our draft taxonomy could be altered so that strategies can be classified across many dimensions or elements. For example, a patient-based implementation tool that is also information technology-based and gives feedback to a doctor may be viewed as having three dimensions. This multi-dimensional approach to implementation could be supported by expanding the taxonomy into a multi-dimensional framework or an ontology to aid guideline developers and implementers to consider the relationships between components that make up implementation strategies [35, 36]. A taxonomy of interventions could also be used in conjunction with an appropriate reporting structure, as it complements current reporting guidelines by providing definitions for a broad categorisation of intervention strategies.
Behavioural change techniques may also intersect with guideline implementation strategies. It is therefore important to consider the reliability of language and coding for implementation strategies used within the guideline, as these could contribute to guideline implementation but could also be classified as behavioural change . Future iterations of the taxonomy should, therefore, consider whether behavioural change techniques need to be differentiated from guideline implementation strategies and if so, how this can be done.
Further development of our draft implementation taxonomy should include a review of other taxonomies for supplementary items that may be relevant to our current taxonomy and the inclusion of strategies that were classified as ‘other’ in the current draft. This could result in a more useful tool for guideline implementers and researchers. A discussion of the strengths and weaknesses of our taxonomy in comparison to other published taxonomies, including those related to specific healthcare settings , would also increase its validity. Specific revisions to the taxonomy will need to consider the expansion of professional strategies to include the range of professionals, consumers, and other groups involved in implementation and the development of a hierarchical structure or possibly a multi-dimensional ontological framework. Further investigation is also required as to which strategies should be included in the ‘other’ category; and the feasibility study needs to be replicated using published research papers instead of abstracts. Input from other implementation science experts should also be sought to strengthen the validity of our draft taxonomy. Nevertheless, we believe that our draft taxonomy is likely to be applicable to other settings, given that the development process involved the use of abstracts from the G-I-N Conference, which attracts implementations researchers from various settings.
While there continue to be issues relating to the reporting of interventions (e.g., the rigour of design and reporting of research versus evaluation), the availability of a taxonomy moves us one step closer to providing structure for reporting and assessing implementation strategies. Guideline implementers and researchers would also benefit from the availability of a well-indexed database of implementation studies, which would facilitate the search for and identification of suitable implementation strategies. One such example is the Health Evidence Database Classification (http://www.mcmasterhealthforum.org/healthsystemsevidence-en), which is a database that contains research evidence about governance, finance, and delivery arrangements of healthcare systems, as well as implementation strategies. Improving the accuracy and granularity of the MeSH vocabulary  for describing implementation research would also facilitate the indexing of implementation strategies.
However, the use of a taxonomy in research reports does not, by any means, reduce the need to describe the implementation strategy in detail. Sufficient information is still needed to enable implementers to replicate the strategy in other locations and settings, which is often not the case even in treatment trials . Recommendations and checklists such as the Guideline for Reporting EBP Educational interventions and Teaching (GREET) statement, which is being developed for educational interventions, will be helpful .