Background
Innovations in health research have the potential to improve health but harnessing this potential requires that effective innovations be translated into routine healthcare. Unfortunately, evidence-practice gaps continue to be documented: medicines are inappropriately prescribed [
1], patient safety practices are not enacted [
2] and harmful practices persist [
3]. Sub-optimal clinical practices (over, under and misuse of tests/treatments) result in avoidable morbidity and mortality [
4]. Such gaps in care are consistent across countries and clinical areas, leading some to suggest that health research is
‘all breakthrough, no follow through’ [
5]. Implementation science emerged in response to this, focusing on the rigorous scientific study and development of a cumulative evidence-base for how best to address evidence-practice gaps. A foundational requirement of implementation is the need for
someone (usually more than one person or group),
somewhere (from organisational leadership through to those providing direct patient care) doing
something (usually more than one thing)
differently. In short, taking up new evidence requires healthcare providers and other health system stakeholders to change their behaviour.
Given the centrality of behaviour in implementation science, there is a need for describing behaviour as clearly as possible. Doing so may help to (a) clarify evidence-practice gaps, (b) clarify the various people and groups at different levels that need to do something differently, (c) identify modifiable barriers and enablers and design implementation interventions to address them, (d) provide an indicator of what to measure to evaluate an intervention’s effect and (e) ultimately facilitate evidence synthesis. A generalisable framework may help to ensure consistency in the description and specification of behaviour in implementation research.
In the mid-twentieth century, after social psychologists identified that scores on attitude measures were not associated with actual behaviour [
6], Fishbein [
7] proposed that the low predictive validity of attitude measures could be addressed by assessing attitude to an
action (e.g.
voting for a specific political party) rather than assessing attitude to the
target of that action (e.g. the political party). While now taken for granted, this focus on the
action led to a paradigm shift in attitude-behaviour research and was a key principle underlying the Theory of Reasoned Action. Extending the approach, Ajzen [
8] proposed the Theory of Planned Behaviour for predicting and explaining human behaviour in a specific context at a specific time. Together, these ideas gave rise to the specification of behaviour according to what became known as TACT: Target, Action, Context, Time [
9]. Similarly, Michie and Johnston [
10] proposed that when behaviours are described in terms of what, who, when, where and how, they are more actionable and hence more likely to be performed.
Clear specification of the behaviour is a key though often overlooked first step in conducting implementation research for a range of study objectives, such as identifying influences (barriers and enablers, and determinants) on behaviour or designing implementation interventions to support behaviour change among stakeholders in the health system. Despite half a century of guidance on behaviour specification, research is frequently published in which the behaviour is poorly specified. A systematic review of 67 reports of behaviour change interventions found that the Action domain was clearly specified in 69% of reports, and that all components of the TACT framework were described in only 5 (7.5%) reports [
11]. Poor specification makes it difficult to measure behaviour and behaviour change. Clear specification facilitates strong compatibility between the behaviour under investigation and the theoretical constructs that predict that behaviour, which enhances prediction (cf. the principle of compatibility) [
12,
13].
Consider the following description of a potential guideline-recommended clinical behaviour for primary care practitioners: ‘For people with diabetes, record a blood pressure reading in the patient’s medical records’ (Example 1). Using the TACT framework to unpack this recommendation, the two specified components are the Target (people with diabetes) and the Action (record a blood pressure reading in the patient’s medical record). While a seemingly straightforward description, it is not clear who should do the behaviour and when and where it should take place. Leaving this implicit introduces ambiguity that may undermine change efforts as well as measurement of whether or not the behaviour has been performed.
Drawing upon Fishbein and Ajzen’s [
9] early advice, a clearer, more actionable specification could be: ‘For people with diabetes (
Target), record a blood pressure reading in the patient’s medical records (
Action) in the primary care clinic (
Context) when they attend for their annual diabetes review (
Time)’ (Example 2). Example 2 specifies further components of the behaviour: Time (when patients attend) and Context (primary care clinic). This enhanced specification corresponds to the TACT framework, but still lacks a fundamental component:
who (i.e. which person or people on the primary care team) is responsible for performing the Action. Furthermore, Example 2 arguably involves a sequence of discrete Actions (taking the blood pressure reading, accessing the patient’s medical records, entering the blood pressure reading into the record). It may also include ancillary behaviours such as inviting the patient to attend the clinic for the annual review. These actions may be performed by different primary care staff (e.g. physician, nurse, administrator) to support the focal behaviour of interest. This dimension of specification is not included in the TACT framework, which assumes that the individual is performing the behaviour for themselves. In implementation research, individuals often perform a behaviour for someone else’s benefit (i.e. the Target, such as a patient). We propose to expand the TACT framework to guide specification of behaviour in terms of not only Target, Action, Context and Time but also Actor—the person(s) who will perform the Action(s). By clarifying the Actor, the Action then becomes clearer and more specific, allowing for clarification of complex behaviours (or sequences of behaviour) in terms of different Actions performed by different Actors in the health care setting at different times (i.e. preparatory and sequential Actions).
AACTT: an expanded framework for specifying behaviour
We propose the AACTT framework (Action, Actor, Context, Target, Time) for specifying behaviour. Re-arranging the order of domains in the framework reflects a more easily defined sequence for specifying behaviour than TACT that naturally begins with the Action and who performs it.
Although it may sometimes seem obvious who is to perform the Action, for behaviours that are performed by healthcare professionals or teams for, with or on behalf of their patients, specification of the Actor is particularly helpful. Indeed, healthcare delivery behaviours have been described as ‘collective behaviours’, suggesting that role confusion may be a barrier to performance [
14] that could be illuminated by careful specification of the behaviour at the outset using AACTT.
While Action and Actor are important, specification of Context and Time allows the responder to keep these elements in mind when answering a questionnaire, responding in an interview or changing their behaviour. Behaviours are inherently tied to the time and place in which they occur, and thus clarification of these elements provides an opportunity to situate analyses of barriers and enablers and intervention design within the contexts that behaviours take place. Recent theoretical advances emphasise the role of associative processes in behaviour [
15], including automatic processes [
16‐
19]. Specification of contextual and temporal cues in questionnaire items, interview topic guides and observational tools may increase the validity of responses, especially if the behaviour has an element of automaticity. Context and time may also be important for identifying when and where it is appropriate to perform an Action, thus informing implementation efforts.
When considering the behaviour of individuals and teams as they deliver health care, a further refinement of the ‘Target’ domain is appropriate. ‘Target’ is often explained as (performing) an Action to someone, i.e. who the behaviour is targeted at. However, current models of healthcare delivery place a focus on patients as active participants in their health care, and thus as a collaborator with the healthcare professional. Hence, rather than performing an Action to a passive recipient, the healthcare professional may act with or for the patient. Thus, it is recommended that researchers frame (i) Action and (ii) Target as (i) doing what? (ii) to, for, with or on behalf of whom?
In doing so, the Actor-Target relationship need not only reflect a healthcare professional-patient relationship. As demonstrated in Fig.
3, for healthcare professionals working in a team, one healthcare professional’s (Actor A) behaviour (Action A) may be for the benefit of another healthcare professional (Target A), enabling the latter’s subsequent behaviour (Action B). Such horizontal sequences within teams also apply to specifying the behaviour of vertical sequences of behaviour within the health system, where a policymaker’s (Actor A) behaviour (Action A) sets the stage for a healthcare administrator (Target A/Actor B) to perform a behaviour within their role (Action B) that benefits the healthcare professional (Target B/Actor C) and enables them to provide care (Behaviour C) to benefit their patients (Target C). Thus, a given Actor’s Target can also be another Actor in the system.
AACTT provides common elements that can be used for consistent description and specification of behaviour. By extension, AACTT can be used to describe the sequence of multiple behaviours of multiple Actors at different levels of the organisation required to enact change. For instance, in the case of promoting hand hygiene in hospital, AACTT provides a means for clarifying the behaviour of those engaging in hand sanitizing behaviour, but also the leadership in the organisation whose policy-enacting behaviour sets the stage for middle management to engage in procurement behaviour to provide hand sanitizing stations and gels, through to maintenance staff engaging in refilling behaviour to ensure sanitizing gel is available for the healthcare providers. Each behaviour by these organisational Actors is required for healthcare providers to engage in hand-sanitizing activities. Rather than making implicit assumptions about such a sequence of behaviours or describing them as separate organisational factors, the AACTT framework helps to unpack the complexity and clarify the responsibility of all behaviours in organisational health settings, providing a clear opportunity for behavioural approaches to inform organisational change. The level of granularity or aggregation in the specification of each AACTT domain should be defined by what is measurable, useful and practical for the given application, to ensure practical utility.
In summary, we propose a new framework: AACTT (Action, Actor, Context, Target, Time—see Table
1 for definitions and examples) to allow for the careful delineation of ‘who does what; to, for or with whom; when; where?’ [
10]. Herein, our aim is to demonstrate how the AACTT framework can be used within the main steps of an implementation research process and to provide a simple tool that implementation researchers and practitioners can use to apply the AACTT framework to specify the behaviour(s) of stakeholders.
Table 1
AACTT framework definitions and examples
Action | A discrete observable behaviour | Prescribing antihypertensives, providing a referral to a specialist, washing hands, setting a policy |
Actor | The individual or group of individuals who perform (or should/could) the Action | Primary care physician, pharmacist, social worker, resident, administrator, middle manager, head of unit, policymaker |
Context | The physical, emotional or social setting in which the Actor performs (or should/could) the Action | Examination room, doctor’s office, outside a patient room, in a boardroom, stressful vs. calm situation, when patients’ relatives are present or not |
Target | The individual or group of individuals for/with/on behalf of whom the Actor performs the Action | Patient with diabetes and blood pressure above 140/80 mmHg, patient wanting to quit smoking |
Time | The time period and duration that the Actor performs the Action in the Context with/for the Target | At annual review, next time a patient visits, every week, over the next 6 months |
Discussion
Herein, we introduced the AACTT framework that can be used to inform the careful specification of behaviour for implementation research and practice and provided a generalisable worksheet to facilitate use of the framework (worked examples in Figs.
1,
2 and
3, blank worksheets in Additional file
1). AACTT can be applied across key steps of implementation research and practice advocated by process models [
20‐
22] to transparently define and measure behaviour(s) in terms of who performs them, for/with whom, when and where. AACTT formalises a natural progression of the TACT framework developed by Fishbein and popularised in social and health psychology for more direct application to behaviour where the individual is performing a behaviour for someone else’s (i.e. Target) benefit. This has direct application not only in implementation science applied to healthcare settings but also public health–, social welfare– and family-based settings in which someone performs an Action for someone else (e.g. school- and family-focused interventions) [
46,
50].
Using the Theory Comparison and Selection Tool (T-CaST) and checklist [
51], the AACTT framework is designed to be
usable (includes relevant domains, has been developed so that key stakeholders can use it, we provide steps for its application and methods for promoting its application across a range of possible studies and an explanation for how the domains influence each other),
testable/valid (can form the basis for testable hypotheses, includes face-valid explanations and has been used in empirical studies),
applicable (focuses on a key implementation outcome, can be applied across a range of methods and across a range of analytical levels, populations and conditions and is generalisable across disciplines) and is likely to be
acceptable (to key stakeholders, and is the historical evolution of a framework rooted in a particular discipline). Thus, in principle it fulfils all the criteria for use of a framework by implementation science researchers and practitioners, though its actual usability, testability, applicability and acceptability will ultimately be determined through application of the tool across a range of types of implementation research [
52].
Michie and Johnston made a call for making clinical practice guidelines more specific by specifying ‘what’, ‘who’, ‘when’, ‘where’ and ‘how’ [
10]. While sharing some similarities with AACTT, there are three important differences that distinguish AACTT and underscore its potential added utility. First, the ‘who’ in Michie and Johnston’s recommendations refers to the ‘Actor’ but does not make any mention of ‘to, for or with whom’ the action is performed (i.e. the ‘Target’ in AACTT, which may be a patient, a healthcare team member or other organisational actor). Second, the ‘where’ is specific to a physical location, whereas ‘Context’ in AACTT can refer to a broader set of contexts that include the physical location but could also include a context that is internal to the actor (e.g. emotional context of a stressful versus a calm situation) or the social context (e.g. when patients’ relatives are present). Third, in proposing AACTT as an extension to Fishbein and Ajzen’s TACT framework, we are deliberate in ensuring cumulative theory and methods development as it extends to applications within implementation science.
Existing calls for better specification and reporting have focused on the description and labelling of implementation intervention strategies (e.g. with taxonomies of change strategies such as the ERIC [
42] and BCT [
43] taxonomies) and on the wider components of interventions (e.g. with checklists such as TIDIER [
40]). Within implementation science, Proctor, Powell and McMillen [
53] proposed seven domains for specifying implementation strategies per se, including who delivers the strategy (actors), how they deliver the strategy (actions), what and to whom the strategy is directed (action target), the sequence of strategy delivery (temporality), intensity (dose), implementation outcomes affected and justification. While some of the Proctor et al. domains share similar or overlapping nomenclature with AACTT (i.e. Actor, Action, Time), the scope of intended application of each framework and its domains differs. As exemplified in Table
1, in the AACTT framework ‘Actor’ refers to the individuals or groups performing the behaviour that could be the recipient of an implementation strategy, the ‘Action’ refers specifically to the behaviour being performed by the Actor and is the object of change and ‘Time’ refers to when that behaviour is performed. Thus, whereas the seven dimensions proposed by Proctor et al. focus on detailing the delivery of an
implementation strategy within an implementation intervention, AACTT domains focus uniquely on detailing
behaviour(s). Thus, in the context of an implementation intervention that would be specified with Proctor et al.’s dimensions, the behaviour specified using AACTT would be an outcome that the (now well-specified) implementation strategy aims to change. Thus, AACTT is designed for specifying behaviour as the object of change, rather than specifying implementation strategies designed to bring about that change.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.