This study serves to add to existing stroke and medication adherence literature by presenting an intervention developed using the BCW, underpinned by the TDF, and designed in consideration of implementation into the NHS. Eleven BCTs were identified as potential components for his intervention. BCTs identified included habit formation (8.4), action planning (1.4) and information about health consequences (5.1).
Consideration of the intervention context (e.g. time and financial pressures within the NHS), facilitated by using APEASE evaluative criteria , has enhanced this process and enabled the development of a focused intervention. This process was important as there is currently a lack of experimental evidence in the literature looking at the effectiveness of particular BCTs targeting specific psychological determinants of medication adherence of stroke survivors. Moreover, inconsistencies in descriptions of published interventions means that it is difficult to establish which BCTs are more effective at targeting medication adherence in stroke survivors. For example, it is often difficult to identify the type of information provision given to participants in an intervention, as varying terms have been used for this intervention component e.g. “an educational booklet”  or “reinforcing relevant knowledge on the chronic diseases they are suffering from” . With the development of checklists to support reporting of interventions, such as the Template for Intervention Description and Replication (TIDieR) checklist , clear and transparent recording of what is included within interventions may be enhanced.
An understanding of underlying modifiable factors is required in order to change behaviour (e.g. ). To achieve this, a systematic review was used in this study that identified potential determinants of medication adherence in stroke survivors to target with intervention . The shortlisted selection of BCTs was generated through systematic linking of determinants to intervention functions, policy categories and associated BCTs, ensuring that the intervention remains evidence-based, theory driven and targets the known modifiable psychological influences on medication adherence in stroke survivors. This will facilitate evaluation of the interventions’ effectiveness when feasibility and pilot testing is carried out, as identification of core intervention components will be possible, something highlighted as important in behaviour change intervention development and reporting .
This approach to intervention development (use of TDF, BCW and APEASE) has focused the identification of intervention functions through which BCTs will be delivered (Education, Persuasion, Training, Environmental Restructuring and Enablement). It has also identified the most likely effective BCTs that could be delivered in intervention (for example, information about health consequences, use of a credible source, self-monitoring of the behaviour, social support (emotional), identifying the pros and cons to taking medications and using habit formation). The effectiveness of these BCTs has been supported in previous research discussed below. A Cochrane review, focused on effectiveness of medication adherence interventions, reported that information, reminders and self-monitoring were included in almost all interventions that showed good effect for improvement in adherence . In 2016, Conn and colleagues  conducted a meta-analysis, assessing blood pressure outcomes for medication adherence interventions among adults with hypertension. It was found that BCTs focused on habit formation were effective at improving diastolic blood pressure (habit d = 0.477; no habit d = 0.181; p < .001) . Moreover, O’Carroll and colleagues  piloted a randomised controlled trial (RCT), testing an intervention (incorporating components to support habit formation) targeting adherence to antihypertensive medications in stroke survivors. Significant results were reported, with 10% more doses taken on schedule in the intervention group (intervention, 97%; control, 87%; [95% CI for difference 0.2,16.2]; p = 0.048), encompassing BCTs such as action planning .
This study highlights a key methodological challenge of applying the TDF and BCW in intervention design: selection of suitable BCTs to target underlying theoretical domains was less systematic than identification of intervention functions and policy categories. However, work in 2015 by Cane and colleagues  and previous mapping working in 2008 by Michie and colleagues  assisted in this process, giving some support and guidance as to which BCTs are likely to target underlying theoretical domains. The use of this literature, along with wider reading of BCTs that are better suited to certain intervention functions and consideration of BCTs that have been previously reported to show reasonable effect in interventions targeting similar behaviours, has been useful in guiding BCT selection.
Strengths and limitations
A key strength of this work was the application of the APEASE evaluative criteria to refine selection of intervention functions, policy categories and BCTs. The multiple dimensions covered by the tool lead to a more careful consideration of the realities of implementing this intervention in the NHS context. For example, the APEASE criteria remind intervention designers to consider not only potential acceptability but also potential side effects and the equity implications of a new intervention. As advocated in the literature (e.g. [15, 23]), the present study also demonstrates a transparent and explicit approach to intervention development. In addition, the research team came from multidisciplinary backgrounds (primary care; health psychology) and two of the three authors were also members of an internationally-recognised applied stroke care research group, factors which enabled a more holistic decision making process when attempting to narrow down selection of BCTs. Input from the health psychologists, who have advanced training on the TDF and intervention development applying the BCW facilitated effective use of theory to underpin intervention design.
The generalisability of this study warrants discussion. The overarching method for medication adherence intervention development applied here is generalisable across patient populations. The systematic review to identify psychological influences on stroke survivors’ adherence included papers from any country and so provides an internationally applicable view of the factors that medication adherence interventions for stroke survivors should target. The ability of particular BCTs to change psychological influences is considered to be generic across contexts, in the absence of evidence to the contrary. Therefore, the initial selection of intervention components that could alter psychological determinants of medication adherence is generalisable across stroke survivors in different contexts. However, the application of the APEASE criteria to narrow the choice of potential intervention components requires one to consider both a specific patient population and context. By explicitly stating our judgements of each intervention component, such as a BCT, vs. the APEASE criteria for stroke survivors within the NHS, we enable others to judge the extent to which our final BCT selection would generalise to interventions for stroke survivors in other countries and contexts.
One limitation of this research may be that assessment of intervention components, such as BCTs, using APEASE criteria involves a certain amount of subjectivity. However, the assessment of potential intervention components were carried out by a multidisciplinary team who have considerable experience and knowledge of adherence interventions and the current healthcare system as it relates to stroke survivors within the UK.
Stroke survivors are a highly diverse patient group, with varying impairments as a result of their strokes, and often significant comorbidities. Our intervention is intended to support stroke survivors who live in the community with some degree of independence, rather than those in institutional care or who are highly dependent on domiciliary carers. Some stroke survivors may have considerable dysphasia. Our intervention is unlikely to entirely meet their needs, and instead, targeted adherence support from doctors, nurses or pharmacists, tailored to the individual’s particular communication difficulties (e.g. expressive or receptive), may be required.
A further potential limitation, and subsequent avenue for future research is the applied definition of adherence the authors have used. There are varying terms used interchangeably to describe a multitude of medication use behaviours including concordance, persistence and adherence. Research efforts have recently provided a more testable and analysable definition , that operationalises adherence into three quantifiable stages: initiation, implementation and discontinuation . Our research focuses on the implementation stage, as many stroke survivors initiated some of their medications (e.g. antihypertensives) prior to their stroke. However, in the future, interventions could be designed to have components tailored to target stroke survivors in the three different phases of medication adherence. Most recently, a new metric for medication adherence measurement has been proposed that allows these aspects of behaviour to be assessed by one measure . However, this metric was derived from a sample of patients using inhalers and considers nonadherence related issues such as inhaler technique. Technique may be less relevant medications prescribed for stroke risk factor control and as such requires further research in this population. Enhancing assessment of adherence will strengthen evaluations of intervention effectiveness.
Work such as the establishment of the links between theoretical domains of the TDF and BCTs is relatively new, and is based on hypothesised links and expert consensus. Therefore, more work is required to provide empirical, experimental evidence showing that behaviour change is possible through delivery of specific BCTs targeting underlying determinants known to influence the behaviour. However, careful exploration of the literature to identify BCTs, alongside pragmatic decision making from a multidisciplinary team regarding suitable BCTs based on evaluative criteria, should enable a more realistic and theory-driven approach.
The next stage of this project will be to explore the acceptability of the proposed intervention components, as well as potential modes of delivery (face-to-face, phone, text, email, website etc.) with key stakeholders (healthcare professionals (HCPs)and stroke survivors). Exploration of the most acceptable way to operationalise the BCTs and the acceptability of the overall BCT will be undertaken. For example, the BCT Action planning will be considered. This BCT could be operationalised as asking a participant to make a plan about exactly when and where they will take their first daily dose of medication, and interviews will explore the acceptability and utility of this. Consideration of how this BCT should be delivered (e.g. in a face-to-face discussion with a HCP, developed over the phone with a HCP, through development of the plans prompted via email) will also be explored in this study.