Background
Healthcare research costs over 100 billion dollars annually in North America alone[
1,
2]. This considerable investment yields important new knowledge that can significantly improve the health of patients and populations, provided that the knowledge is implemented appropriately. Over the last 20 years or so, it has become increasingly apparent that ‘appropriate implementation’ is an extraordinarily complex and multifactorial problem[
3]. The techniques people have used to implement new knowledge have most often lacked any substantive justification, and have instead been based on past practice and logistical constraints rather than any in-depth understanding of what is likely to work. Perhaps not surprisingly, systematic reviews of many commonly used implementation techniques[
4‐
8] have shown their effectiveness to be highly variable. While these realizations have coalesced into the science of implementation, also referred to as knowledge translation (KT), much remains to be understood about why strategies aimed at improving the use of new research knowledge to improve healthcare have proven so inconsistent in effect.
An important debate in this developing discipline pertains to the use of theory to understand the techniques and processes underlying KT. Some have argued that KT interventions are too heterogeneous for anything to be gained by trying to develop a generalizable theory[
9,
10]. More recently, however, the literature has largely subscribed to the arguments put forward by Eccles and others[
3,
11] that the benefits to be derived from theory in terms of generalizability of findings and standardization of methodology outweigh any risks.
Early efforts to understand KT techniques and processes have focused on identifying which existing social and health psychological theories can most successfully be applied in new KT contexts[
12]. This approach correctly assumes that the psychological principles underlying one area of human endeavor often transfer to others[
13] and, as such, may provide a head start towards identifying causal mechanisms that determine the effectiveness of the KT intervention in question[
3]. Using existing theory (
i.e., ‘a coherent and non-contradictory set of statements, concepts, or ideas that organises, predicts, and explains phenomena, events, behavior, etc.’[
3,
14]) has the further benefits of a foundation of both empirical work performed in other contexts and methodological innovations establishing strategies for measuring relevant constructs (
i.e., the abstracted concepts or explanatory variables on which theory is built)[
3].
The literature has now seen the application of a variety of theories to various KT endeavors (see Godin[
15] for a review). Based on this work, a number of lessons can be identified. First, the number of theories potentially available for study is enormous[
16,
17]. Second, theory may guide implementation in many ways[
18], ranging from closely directing the development and implementation of the intervention, to crudely being used as a crutch to justify an intervention
post hoc[
19]. Third, considering theory alone, without empirical observation to identify the influence of a specific context, can miss important information relevant to improving KT[
9,
12]. Fourth, theories can differ in the roles they are intended to serve, with some clearly designed to provide explanation and prediction and others being intended primarily to serve communicative roles, a fact that can complicate application to new contexts[
20]. In part because of these challenges, the majority of theory application work has focused on a very small number of theories[
15] and yielded only modest progress on issues about how to improve KT interventions[
21]. The best example of this is the Theory of Planned Behavior (TPB)[
21,
22], which has been attractive to KT researchers because of its broad scope (
i.e., it is explicitly intended to be applicable to all voluntary human behaviors) and its demonstrated applicability to a wide range of behaviors. Unfortunately, the TPB has proven less useful for generating clear recommendations for improving KT.
In response to these issues, some have called for an increased focus on theory construction in KT research, rather than the wholesale application of existing theories to contexts that may extend beyond their original mandate. For example, Rycroft-Malone[
12] argues that development of context-specific ‘micro-theories’ would result in a better understanding of the KT issues that are specific to different stakeholders, disciplines, and settings. This might develop important localized information, but the inevitable risks of coarse application of such an approach would include lack of generalizability of findings and the necessity for considerable methodological groundwork for each new context.
We propose that there is a middle ground that might be usefully explored. Construction (as opposed to application) of a theory need not involve de novo, start-from-scratch theory building. Instead, when seeking causal mechanisms that contribute to successful KT, one could combine theory-building activities with individual constructs from any number of relevant theories. We will refer to this as the ‘menu of constructs’ approach, and argue that researchers looking to explain an area of KT should consider the entire menu, not just constructs associated with a particular theory.
We see the menu of constructs approach as attractive for a number of reasons. It allows for inclusion of only those constructs that are relevant to the new context, rather than requiring transport of the entire theory. Very often, it is the individual constructs, rather than the theory as a whole, that recommends the theory in the first place. For example, the notion that channel factors (
i.e., features of the environment that lead individuals to act in particular ways) will influence a clinician’s behavior may be more readily applied to KT interventions than the entire social psychological theory of situationism (
i.e., the notion that the situation in which one finds themselves is the dominant factor in determining an individual’s behavior[
23]) in which channel factors are embedded. Theory newly constructed from a menu of constructs can: leverage measurement/methodological advances from the domain in which the various constructs were generated; incorporate both theory-based constructs and components of the specific context and behavior[
24]; and propose KT interventions that are rooted in, but not restricted by, the larger body of theoretical literature.
The goal of this paper is to spur debate about the range of roles that theory (and theory-relevant constructs) should play in the overall endeavor to improve KT interventions. Our central claim is that theory development may progress more quickly if we allow ourselves to incorporate constructs derived from a range of theories, rather than feeling restricted to align/justify/use any particular theory in its entirety, thus broadening and tailoring the conceptual underpinning of specific KT interventions. In the following sections, we make our arguments in the context of audit and feedback (A&F) as an example KT intervention, but believe that the general logic will apply more broadly. We begin by explaining why we have chosen A&F as our example. We then provide some examples about how the menu of constructs approach is already being explored. We then discuss several examples of constructs that may inform our understanding of KT interventions, but have not been considered in the context of a theory of A&F. We discuss how the menu of constructs approach relates to other emerging paradigms for theory use in KT, particularly the theoretical ‘domains’ approach suggested by the work of Michie
et al.[
17]. And finally, we address some limitations and areas for future work suggested by this approach to theory building.
Summary
Our examples of theoretical constructs derived from cognitive and educational psychology all have one thing in common. They all stem from theories that, because of the context and/or level of abstraction at which they were originally developed, could not hope to provide anything like a full picture of an A&F intervention. As a result, these theories would generally not be considered relevant to KT interventions, because the context for which they were developed is so far removed from complex A&F interventions. Nevertheless, they do provide specific, testable hypotheses about ways in which A&F interventions might be improved. Thus the need for adopting a menu of constructs approach; each theoretical construct mentioned (and the many more that were not mentioned) promise productive lines of inquiry that can yield greater guidance regarding how to adapt A&F strategies to particular settings and how to productively and efficiently test the effectiveness of such strategies.
Menu of constructs and the theoretical domains framework
This menu of constructs approach is not inconsistent with an important new approach to the use of theory in KT research started by Michie
et al.[
17]. Noting the number of health and social psychological theories potentially relevant to such research, as well as the considerable overlap in constructs among them, this group engaged in a consensus process that distilled from 33 different theories a set of 12 behavior change ‘domains’ agreed to be relevant to implementation research. These domains are intended to be distillations of different, but related, constructs from different theories, ones that nevertheless have common implications for behavior change and implementation research. This offers an example of how the menu of constructs approach might be implemented. By beginning from these 12 domains (now updated to 13)[
70], this Theoretical Domains Framework (TDF) offers a systematic means to consider a wide range of theoretical approaches, and to narrow one’s search for theories relevant to a specific KT context.
The TDF has since been extended in at least two different ways. First, it has been directly employed to recommend specific implementation intervention techniques, based on expert agreement on what intervention techniques are implied by each domain[
71,
72]. Such an approach can provide concrete guidance for interventionists, but may do so without explicit reference to specific theories or constructs.
A second approach, referred to as Theoretical Domains Interviewing (TDI)[
73,
74], involves developing an intervention by interviewing or surveying the target audience and using the domains to prompt participants to identify barriers and facilitators to the target behavior. Responses are then categorized into the most relevant theory-specific constructs, and an intervention is developed based on the recommendations of the theory with the most constructs identified as being relevant to the target behavior. TDI therefore provides a systematic approach to identifying which theory may be most appropriate to the new context.
This second approach to applying the TDF specifically seeks to select ‘the most appropriate theories to develop interventions for changing specific behaviors’[
73]. Because the goal is to identify what is relevant at the level of the theory rather than the level of the construct, however, the approach might fail to target constructs such as those discussed in this paper. Because theories from cognitive psychology often seek to explain mechanisms rather than behaviors (
i.e., they explain at a lower level of abstraction), they may not, in isolation, lend themselves to explaining complex behaviors and, as such, exclusive use of TDI may lead us back to the issues that began this paper (
i.e., those that arise when individual theories are adopted in an all-or-none manner). This is likely the reason why the original TDF[
17] combined much of cognitive psychology into a single construct (memory, attention, and decision processes); because of the level of abstraction problem, it is not clear how the many theories within cognitive psychology might be relevant to implementation. It is only at the level of the construct that it becomes clear how such processes can inform KT interventions.
We propose a simple fix to marry the TDI with the menu of constructs approach that will simultaneously indicate how the menu of constructs approach might be productively implemented. Rather than identifying constructs in the interviews as a means to identifying the most relevant theory, in some cases it might be worthwhile to consider all constructs deemed relevant in the interviews, and use them to construct a new theory specific to that KT context. This approach would allow development of interventions and theory that incorporates constructs from various theories. Such an approach would be consistent with our menu of constructs idea, allow for the incorporation of constructs at different levels of abstraction, and also make use of the important methodological advances from the TDF and TDI.
We believe this approach would help overcome some thinkers’ objections to the utility of using theory in KT research[
9,
10]. Use of theory need not be in opposition to detailed empirical understanding, but instead should serve as an orienting conceptual framework that can be used iteratively to both guide and to be influenced by empirical observation. Much of the negative connotation that can be associated with the word ‘theory’ comes from confounding of the term with high level conceptualizations that may describe a problem well, but offer little in the way of concrete guidance regarding specific mechanisms whereby practice can be changed. Theory that is so broad as to be applicable to any situation may inevitably be so weak as to yield little more than adages that frame an outcome after it has occurred. For example, theories that conceive of creativity as a tendency to ‘think outside the box’ can provide adequate descriptions of activities that define creativity, but ultimately offer little guidance regarding how to do so effectively.
The most effective methods for implementing this menu of constructs approach have yet to be established. We plan to present experts in a wide range of theoretical domains with example A&F interventions, eliciting their opinions about theories and constructs that they feel make testable predictions about how to improve the interventions. Once a laundry list of constructs is assembled, we anticipate conducting pilot evaluations in the form of written vignettes, usability testing sessions, or small scale randomized controlled trials to assess which candidate constructs warrant more formal evaluation. This pilot testing process could also incorporate the results of bottom-up analyses of the target behavior and context, of the sort recommended by others[
75].
Limitations
This menu of constructs approach may be seen to have much in common with Bandura’s pejorative term ‘cafeteria-style research’[
76]. He argues that picking constructs from various theories and recombining them can lead to needless proliferation of essentially identical constructs with different names. When one’s goal is integration of multiple overlapping theories into one all-encompassing theory, such proliferation is clearly a problem. However, we see the process of theory development in KT as being distant from such a grand unifying theory. In the context of A&F interventions (only one of many possible KT interventions), we are only beginning to understand what factors predict an effective intervention. What is becoming clear is that the broad social cognitive constructs such as those offered by the TPB have not offered a sufficiently detailed theoretical description to help us to consistently design effective A&F interventions, and that the door must be opened to theory from a broader range of disciplines to understand these complex interventions[
77]. When higher-order models do not provide sufficient help, ‘drilling down’ to more complex and context-specific aspects of behavior seems only sensible.
Another potential limitation of the menu of constructs approach is that by incorporating individual constructs independently of the models with which they were developed, one may lose some of the power of the original theory, and potentially some of the meaning of the construct itself. Theories are comprised not only of constructs, but of the proposed relationships between constructs as well. Porting constructs into new contexts, separate from these relationships, may have unexpected implications for the utility of the construct. For example, will the ‘perceived behavioral control’ construct from the TPB have the same explanatory value independently of the other TPB constructs? We believe that this problem is one of validation. No one is suggesting that all potentially relevant constructs will prove useful in every context. Rather, adoption of a menu of constructs approach is meant to offer specific prompts that necessitate the validation of each construct within the context of the new theory being built. In the newly resulting theoretical context, any individual construct may or may not add explanatory value, and cannot therefore be included solely based on its utility in its original theory.
Finally, while in general we see the flexibility inherent in the menu of constructs approach to be a way forward in the KT literature, that flexibility may sometimes be more of a hindrance than a help. One of the attractions of the TDF approach is that it boils down many theories into a few key domains, which may be seen as more tractable from the point of view of designing implementation strategies. In contrast, the menu of constructs approach widens the number of constructs to consider even further, by incorporating theories that to date have not been considered in a KT context. We feel that different tools will suit different purposes, and that further work attempting to use the menu of constructs approach will allow us to fruitfully explore these issues in more detail.
As with any component of evidence-based medicine, theory should be applied judiciously rather than adopted lock, stock, and barrel with no consideration of the individual idiosyncrasies created by different contexts. Furthermore, as KT researchers, the goal of theory building should remain firmly on how to create more effective interventions. If the menu of constructs approach allows us to develop a better understanding of the range of theories from multiple disciplines available to us, and to engage in systematic study of the applicability of their constructs, we will be better positioned to make clear recommendations about how to implement newly developed knowledge more effectively.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
Both authors contributed to all aspects of the manuscript. Both authors read and approved the final manuscript.