Methods for designing interventions to change healthcare professionals’ behaviour: a systematic review

Our project sought to advance the methods for translating research knowledge into practice. Knowledge translation (KT) is ‘a dynamic and iterative process that includes the synthesis, dissemination, exchange and ethically sound application of knowledge to improve health, provide more effective health services and products and strengthen the healthcare system’ [1]. One of the critical aspects of KT is that it requires healthcare professionals (HCPs) to change practice [2].

HCPs’ practice can be influenced by a wide range of factors; for example, a recent review identified 57 clusters of factors [3]. Specific interventions range from interventions targeted at HCPs (e.g. educational materials, audit and feedback) to interventions targeted towards consumers and policy-makers. The evidence base for many of these interventions remains incomplete [4], and there is an on-going need to design more effective interventions.

Systematic reviews of KT interventions to change HCPs’ practice consistently indicate that interventions are haphazardly designed and poorly specified, limiting our ability for replication, understanding, and generalizability [5, 6]. Limitations in intervention design impede evaluations of interventions [7, 8]. One issue contributing to the shortcomings in intervention design is a lack of agreed, practical ‘how to’ guidance for designing KT interventions.

Recommendations have been made to ensure that intervention design includes an assessment and prioritization of barriers, identification of potential adopters and practice environments, and consideration of both the potential effectiveness and feasibility of the chosen strategies [2], but these recommendations do not necessarily provide an approach to the design or development of the intervention [9]. Various potential tools (e.g. Behaviour Change Technique (BCT) taxonomy [10]) and sources describing a range of methods for mapping barriers and facilitators to KT interventions exist [11], but sources describing a range of complete methods for intervention design are few.

The aim of the present study was to contribute to the design of such a resource by synthesising literature about methods for designing KT interventions. Our specific objective was to systematically identify published methods for designing interventions to change HCPs’ behaviour.

A systematic review was undertaken. We did not publish a protocol. The initial literature search included MEDLINE, Embase, and PsycINFO from1996 to April 2013. An identical search was conducted on April 20, 2015 to identify papers published since the initial search. A sensitive search strategy was designed in consultation with an information science specialist (CF) using both subject headings and text terms and comprised a combination of three facets: professional behaviour change; theory, framework or technique; and interventions. The search strategies used are detailed in Additional file 1. Reference lists of included papers were screened for additional papers as were articles known to the review team. Our search started in 1996 as this was consistent with the introduction of the evidence-based medicine movement [12], and an associated increase in evidence-to-practice related publications [13].

Papers were included if two criteria were met: (1) the paper described a method (process, tasks, approach) for designing interventions to change HCPs’ behaviour or practice, and (2) the primary focus of the paper was on the intervention design process (as opposed to, e.g. on intervention evaluation). We defined interventions as: ‘a method or technique designed to enhance adoption, implementation and sustainability of a clinical/therapeutic program or practice, a specific clinical/therapeutic practice or delivery system/organizational arrangement being tested or implemented to improve healthcare outcomes’ [14]. A HCP was defined as any member of the healthcare team providing care, and their behaviour was defined as objectively observable actions (as opposed to, e.g. their knowledge or reasoning).

Protocol papers were included if the primary aim of the protocol was to describe intervention design methods or process. Papers were excluded if they pertained to HCPs’ behaviours not related to their clinical practice (e.g. HCPs’ eating healthily, exercising). While papers that report the implementation and evaluation of interventions may include descriptions of how the intervention was designed, this is rarely in a detailed and replicable manner [15]. As our interest was in providing a resource to guide researchers in the process of intervention design, we excluded papers that lacked enough detail for replication. These decisions were made based on the consensus of the three reviewers (HLC, JES, NK). Due to resource limitations, we also excluded articles that were not in English.

A screen of titles and abstracts was conducted by one rater independently (shared by HLC, JES, NK) and was followed by a review of full papers by three raters independently (HLC, JES, NK). An interrater reliability analysis using the Kappa statistic was performed to determine consistency among raters for the full-text review.

For all included full texts, general descriptive information (authors, year, journal, name of method if so named) were extracted and tabulated. To extract and analyse data about the methods, a two-stage process was carried out. Stage 1 involved generating a framework for data extraction and analysis. Three reviewers (HLC, JES, NK) progressed in iterative cycles of reading and discussing the included papers to identify similarities and differences between them and used these discussions to develop a list of items to be extracted. The iterative cycles were continued until an agreement between the three authors was reached. In part, this process was necessary to improve our understanding of the tasks that constitute intervention design and allow us to extract data outside of simply a brief description of the method. The resulting descriptive variables to be extracted were believed to be the most critical, and it included a brief description of the design tasks. Data extraction was conducted by two individuals (HLC, JES) independently first followed by consensus discussions for discrepancies (Stage 2).

We conducted a systematic review of methods for designing interventions to change HCPs’ behaviour. We found 15 papers that outlined 15 methods. All of the papers reported methods that were developed within a specific context. Thirteen papers targeted only individual HCPs with one paper targeting change across individual, organization, and system levels. The methods consisted of at least two tasks and, at most, seven tasks. Thirteen papers included some form of barrier identification and 13 provided direction for linking barriers to intervention components; however, these were not the same 13 papers. Thirteen papers included the use of theory, and another 13 included gathering input on the design of the intervention from the targets of the intervention.

A number of publications related to designing interventions were not included in this review, For example, MRC guidance documents for developing and evaluating complex interventions [9], publications outlining the KT process [34, 35], tools, and frameworks that examine barriers assessments for KT intervention [17], and taxonomies of behaviour change techniques [10]. While these publications are certainly of high relevance to intervention development and evaluation process in general, these were not included here as they were all judged to provide limited detail about the specific, replicable actions to design interventions. For example, the MRC guidance emphasises the importance of designing interventions, but is limited in concrete guidance on how to actually do this in practice. In addition, a number of papers were identified that specifically stated intervention design as an aim [36, 37] but that lacked the detail that would allow replication. It could be that there are additional papers not included in this review that could facilitate intervention design. The majority of the methods found (11/15) incorporated other tools or resources, albeit not in identical ways. While we do not know the rationale for doing so, it could be that existing tools alone are felt to be inadequate for intervention design. Future studies on additional design methods that incorporate other existing tools and resources will likely aid in advancing methodologies for designing interventions.

There were two additional papers not included in our review that received significant dialogue during consensus and therefore warrant some discussion. One of these was Eccles et al. [38] which includes a description of using a theory to design a KT intervention. We felt this paper provided a rationale and description of conceptual issues related to using theory to develop an intervention; however, the degree of detail on how to design an intervention in this paper is limited. The second paper warranting discussion was on Intervention Mapping [29]. This paper outlined a method for intervention design for health behaviours, not for HCPs’ behaviours, and was therefore excluded. However, we did find four methods papers for designing interventions to change HCPs’ behaviour that incorporated Intervention Mapping [20, 23, 26, 27]. It is likely that other methods to design interventions to change health behaviours could similarly be adopted to design interventions to change HCPs’ practice.

Two main gaps seem evident in our review of the intervention design literature. First, limited methods target change in organisations or systems, or at least were developed with a focus on the organization or system. We found only one study [14] that did this explicitly. A second study [22] targeted teams as well as individuals but did not do so at the organizational level. We found limited methods to specifically take the organisation and system level contexts into consideration, as well as methods that consider all levels. While it is true that many of the methods’ approaches to barrier identification could result in a focus on the organisation or system should barriers at these levels be identified (for example, see French et al. [21]), the implicit focus of these papers targeted individual behaviour change. Future studies should consider how and under what circumstances to ensure that organisational and system level change is considered.

The process of undertaking this review highlights a second gap: the need for a better understanding of what activities constitute intervention design. Our iterative process of determining the intervention design variables for extraction and the subsequent extraction of those variables has led to a better understanding of the steps inherent in intervention design, at least according to current methods. There appear to be four steps common to intervention design: barrier identification, linking barriers to intervention component selection, use of theory, and user engagement (i.e. seeking input on feasibility or acceptability of the intervention from the potential targets). While we do not necessarily understand the best order for these tasks nor do we know what additional tasks are required, it does represent a simple structure of potential prototypical steps for the design of a KT intervention. Additional understanding of these tasks, as well as a more in-depth consideration of potentially additional tasks that should be but are not yet routinely adopted, would improve intervention design methods.

Almost all of the methods found (13/15) used theory at some point in the tasks for intervention design, yet evidence indicates theory is rarely used in the design of interventions, or at least rarely reported [39, 40]. Selecting one of these published methods or building on these methods is likely to guide researchers to use theory. Our review did not measure the degree to which these methods are used but this would be a useful future area of research. Additionally, future methodological work could focus on best practices for the use of theory to design an intervention.

Several limitations of this review warrant discussion. We used only one rater for the title and abstract review. Although support exists for the validity of using one rater [41], having two raters would have reduced the possibility of omitting a potentially relevant study. All of the included methods were developed specifically for healthcare environments. This could be in part due to our search strategy. Other methods from disciplines outside of healthcare could yield additional and suitable methods, as could methods developed prior to 1996. Several limitations exist that might have reduced the number of potential methods found. Due to the challenges in adequately searching books, we did not include books or book chapters in our search. Our inclusion criteria meant that we did not include any studies that reported on the testing of an intervention in addition to the development of the intervention. In part, we did this to isolate methods that were described in enough detail to be able to replicate and adequately guide the design of an intervention. Lastly, we did not search grey literature, making the review susceptible to publication bias [42], and we only chose to use three databases. It is feasible that additional methods exist.

This systematic review outlined 15 published and replicable methods for designing interventions to change HCPs’ behaviour. Its use as a resource and as a catalyst for improved quality and quantity of methods is encouraged. Although these methods included varied steps, there was a general agreement that designing an intervention for individual-level change includes identifying barriers, selecting intervention components, using theory, and engaging end-users. Methods for designing organisation and system-level interventions were limited. Further comparative analysis of how the common tasks are completed in the different methods will provide a starting point for developing more detailed guidelines for designing KT interventions. Future research should focus on the degree to which these methods have been used, determining how such methods could be better adopted and further development of both guidance for the existing methods and, potentially, new methods.