Improving medication adherence in stroke survivors: the intervention development process

This stage comprises four steps (detailed in Fig. 1) that facilitate defining the public health problem in context. This was approached by reviewing evidence to define the behavioural problem (Step 1), breaking this down into specific behaviours and eventually selecting one target behaviour to focus upon (Step 2 and 3). A systematic review of psychological determinants of medication adherence in stroke survivors [26] supported understanding of influences upon the behaviour (Step 4). Mapping the identified determinants into the TDF and identifying key domains that were more influential to behaviour change enabled a better understanding of what needed to change.

This stage of intervention development was supported by the Behaviour Change Wheel guidance [23, 25], encompassing two steps. This drew on previous research, whereby a systematic review identified 19 frameworks of behaviour change that encompassed the nine intervention functions and seven policy categories used in the BCW [23]. Moreover, a consensus exercise undertaken by experts in the field, linked theoretical domains to intervention functions and subsequent policy categories well suited to facilitate the behaviour change [23, 25]. Intervention functions are defined as “broad categories of means by which an intervention can change behaviour” [25] and policy categories can be understood as “types of decisions made by authorities that help to support and enact the intervention” [25]. Use of this literature facilitated identification of relevant intervention functions (Step 1) and policy categories (Step 2). Extensive work has been carried out to identify the intervention functions and policy categories most likely to bring about change, when linked to underlying theoretical domains form the TDF [25]. A consideration of the likely services (e.g. health, voluntary, social etc.) that the intervention could be implemented into also facilitated choice of intervention functions and realistic policy categories. Application of APEASE [25] provided evaluative criteria and also narrowed down the potential intervention functions that would be carried forward for this intervention development.

Whilst effectiveness of an intervention is certainly important to consider during the design process, there are other factors that are important, such as evaluation of intervention component suitability in the intended setting and social context. APEASE evaluative criteria support this broader consideration by challenging an intervention designer to ask questions such as [25]:

Therefore, the use of APEASE criteria supplements the selection of intervention components through the application of contextual criteria.

An emerging literature base from experts in the field, identifying BCTs better suited to certain intervention functions and underlying theoretical domains, supported evidence-based selection of appropriate BCTs [27, 28]. In addition, extensive literature searching of current BCTs used in interventions deemed to be effective, targeting akin behaviours or patient groups, as well as application of APEASE criteria supported BCT choice. Affordability, practicality and acceptability were deemed to be of particular importance in this evaluative process by the research team, given the NHS context for implementation. To set the scene, the NHS provides universal health coverage and it is a service in high demand, in part because it is free at the point of use. At present, the UK norm for length of a general practitioner (GP) consultation is a 10-min appointment, although some practices are able to offer longer appointments. Nurse appointments in primary care are often longer. Community pharmacists do not usually operate an appointment system but offer consultations based on clinical requirements. HCPs such as community pharmacists are increasingly enlisted to deliver interventions to enhance healthcare. For example, pharmacists offer medication use reviews (MURs) annually to patients and this may offer an alternative to GP care. Specific to stroke services, patients discharged from a secondary care setting will often only have one or two follow up appointments with the secondary care team before being discharged to primary care services, where the secondary prevention of stroke through medication use, for example, will then be managed. BCTs found not to meet these three APEASE criteria were not carried forward to the next stage of intervention design.

The target behaviour for this intervention is medication adherence, defined as “the extent to which the patient’s action matches the agreed recommendations” [6]. As discussed earlier, up to a third of stroke survivors are non-adherent to their medications, indicating the behavioural problem for this intervention (Step 1).

The target behaviour for this intervention was further defined as: stroke survivors taking medication at the prescribed times, doses and frequencies (Step 2 and 3). A decision was made, following consultation with a patient and public involvement group, to target all medication as opposed to narrowing the focus down to one medication, such as an antihypertensive. A systematic review identifying the psychological factors that influence stroke survivors’ medication adherence was already available [26] (Step 4). Determinants including concerns about medications, beliefs about medication necessity, knowledge about medications and negative emotions had been identified as influential [26]. Once mapped into the TDF, three key domains were identified as influential in behaviour change: ‘Knowledge’ (knowledge about medications), ‘Beliefs about consequences’ (concerns about medications, beliefs about medication necessity) and ‘Emotions’ (negative emotions) domains.

A mapping process, recommended by BCW guidance [23, 25] was followed. This drew on previous research [23, 25], utilised to identify the potential intervention functions and policy categories that would target the three theoretical domains found to be most influential to medication adherence in stroke survivors in Stage 1. The possible intervention functions (Step 1) and policy categories (Step 2) are displayed in Tables 1 and 2.

The use of APEASE criteria [25], along with consideration of the intervention context, assisted in narrowing down the potentially appropriate intervention functions. Likewise, the same process was used to narrow down the potentially appropriate policy categories. Within Tables 1 and 2, reasons for inclusion/exclusion of the intervention functions and policy categories are presented and examples for exclusion of intervention functions and policy categories are discussed below.

Four intervention functions were not considered to be appropriate for this intervention design (Incentivisation; Coercion; Restriction; and Modelling). For example, Restriction i.e. “use rules to reduce the opportunity to engage in the behaviour” [25], when considering APEASE would likely not be acceptable or practical. Medicine taking is often done alone and so there will be no one there to enforce the rules or witness rule breaking, which in turn will limit the effectiveness of this intervention function. Coercion i.e. “Create an expectation of punishment or cost” [25] seems inappropriate, within the service the intervention would likely be implemented (the NHS), morally and ethically (particularly as this conflicts with HCP ethical frameworks). It is inappropriate to create an expectation of punishment if patients do not take their medicines, given that patients can have valid reasons for not wanting to adhere to regimens such as wanting to stop side effects. Moreover, in terms of APEASE, it is likely not acceptable to coerce medication adherence, and there will likely be unwanted side effects from the use of this intervention function, potentially causing the undesired consequence of even worse adherence from patients. Likewise, Incentivisation i.e. “create an expectation of reward” [25], has similar ethical constraints, in addition to potential financial constraints. Although there is emerging evidence (both from systematic review and primary studies) to suggest that incentivisation could have utility in interventions targeting medication adherence [29‐33], these studies have predominantly been conducted within the US and have shown promising but varying effect sizes that are not consistently statistically significant. Within the UK NHS context, financial and economic constraints are such that interventions based on financial incentives are unlikely to be adopted. Finally Modelling i.e. “provide an example of people to aspire to or emulate” [25] was felt to limit some practicality and possibly some equity. The ability to emulate another’s behaviour could be contingent on multiple other factors, such as access to similar healthcare services and patients possessing similar abilities (cognitive, social, physical) to successfully take their medicines in the same manner.

Two policy categories were not considered further for this intervention design (Fiscal Measures and Legislation). For example, Fiscal Measures i.e. “the use of the tax system to reduce or increase the financial cost” [25] is not a practical policy category to consider. This is partly due to the fact that within the UK, residents over the age of 60 years (a category that a large proportion of stroke survivors come under) do not pay for prescriptions and as such an amendment to taxation systems seems a less practical option. Furthermore, individuals considered to be living on low incomes or receiving certain types of welfare benefits (such as those unable to work due to illness) can apply for exemptions of paying for prescriptions. In the UK, prescription costs are also comparatively low in comparison to other healthcare systems such as those in the US (£8.80 per item, which equates to US $11.85 or €10.20), and for individuals requiring multiple prescriptions regularly, there are schemes in place to reduce the maximum annual cost to £104. Fiscal measures would likely require legislation changes, something that would rely upon elected politicians’ willingness to propose such changes. There would also be questions of affordability dependant on the economic climate at the time of the intervention, and thus the use of this policy category could become less acceptable. Legislation i.e. “making or changing laws” [25] was not practical to focus on within this project as the process involved would be out of scope for a research study.

Table 3 displays the process of systematically using an evidence base to select potential BCTs for this intervention. Careful linking of evidence base and theory, underpinned by the BCW [23, 25] and previous work by experts in the field [27, 28] has resulted in a potential 21 BCTs that might be considered for this intervention. Application of APEASE, as well as identifying existing effectiveness of the BCTs within other, similar interventions has enabled the selection to be narrowed down to 11 BCTs (information about health consequences (5.1); self-monitoring of behaviour (2.3); biofeedback (2.6); information about antecedents (4.3); credible source (9:1); self-monitoring of outcome(s) of behaviour (2.4); pros and cons (9.2); prompts/cues (7:1); action planning (1:4); habit formation (8:3); social support (emotional) (3.3)). Table 4 presents all 21 BCTs, separated into BCTs that will be included or excluded from the next stage of this intervention development. Reasons for inclusion/exclusion of each BCT are summarised in Table 4, assessed against APEASE criteria.