‘I just thought that it was such an impossible thing’: A qualitative study of barriers and facilitators to discontinuing long‐term use of benzodiazepine receptor agonists using the Theoretical Domains Framework

Abstract Introduction Existing interventions to reduce long‐term benzodiazepine receptor agonist (BZRA) use lack theoretical underpinning and detailed descriptions. This creates difficulties in understanding how interventions work and how to replicate them in practice. The Theoretical Domains Framework (TDF) can be used to identify behaviour change determinants to target during intervention development. Objective To explore barriers and facilitators to discontinuing BZRA use from the perspective of both current and previous long‐term BZRA users. Design/Setting and Participants Semistructured TDF‐based interviews were conducted with community‐based individuals with current or previous experience of long‐term BZRA use. Data were recorded, transcribed and analysed using the framework method. Results Twenty‐eight individuals were interviewed. Despite commonalities in perceived barriers/facilitators to discontinuing BZRA use within individual TDF domains, individual participants had different experiences of identified determinants of BZRA discontinuation. For example, both similarities and differences existed within and between each participant group in terms of knowledge of the appropriate duration of BZRA use (‘Knowledge’ domain) and experience of withdrawal symptoms (‘Reinforcement’ domain). Compared to previous users, current users typically anticipated more barriers to discontinuing BZRA use and fewer positive consequences of discontinuation. Conclusion This study reports on barriers and facilitators to discontinuing BZRA use from the perspectives of current and previous long‐term users. The findings highlight the challenging nature of BZRA discontinuation and a multitude of barriers that impact participants’ behaviour regarding BZRA use. Future work will involve developing a theory‐based intervention to support BZRA discontinuation in primary care. Patient Contribution The study included patients as participants.


| INTRODUCTION
Benzodiazepines and Z-drugs (e.g., zopiclone) are chemically distinct medication classes with similar mechanisms of action involving gamma-aminobutyric acid, the principal inhibitory neurotransmitter. 1,2 Benzodiazepines are licensed for several indications, including anxiety and insomnia, whereas Z-drugs are only licensed for insomnia. Collectively, they are described as benzodiazepine receptor agonists (BZRAs). Guidelines recommend limiting BZRA prescribing to short-term use (≤4 weeks). 3,4 However long-term BZRA use (>3 months) persists worldwide. 5,6 This is problematic as patients can become physically dependent on BZRAs and experience withdrawal symptoms, even after relatively short-term use (3-4 weeks). 7,8 In some cases, patients may continue BZRA use for longer than originally intended to prevent or suppress withdrawal symptoms. 9 Longterm BZRA use has several negative consequences, including cognitive impairment and increased risk of falls. 10 Potential benefits of discontinuing long-term BZRA use include less daytime sedation, improved cognition, mood and sleep quality and fewer falls. 3 Interventions targeting long-term BZRA use have shown varying effects. [11][12][13][14] For example, cognitive behavioural therapy (CBT) combined with gradual dosage reduction has proven effective in the short term (up to 3 months postintervention) in reducing BZRA use. 12 However, the precision with which CBT-based interventions have been described varies and effects have not been sustained over longer periods. Evidence is lacking to support pharmacological interventions (e.g., anticonvulsants, antidepressants) in facilitating discontinuation of long-term BZRA use. 13 Brief interventions (e.g., short consultations with healthcare professionals recommending discontinuation) have demonstrated effectiveness in reducing and discontinuing long-term BZRA use at 6-12 months postintervention. 14 However, previous interventions have lacked theoretical underpinning and were often poorly described. This makes it difficult to understand how the interventions worked and limits the potential for replication.
In developing interventions, it is increasingly recommended that researchers adopt a systematic approach and explicitly report on the intervention development process. 15,16 For example, the UK Medical Research Council's framework for developing and evaluating complex interventions advocates using the best available evidence and an appropriate theory-base to inform intervention development. 17 Previous qualitative studies have examined patients' experiences and perceptions of BZRA use. 18 While these studies have identified various barriers, including dependence, withdrawal symptoms and absence of supports, they have not comprehensively examined both barriers and facilitators or compared and contrasted the experiences of both current and previous long-term BZRA users. 18 Furthermore, the studies have not employed a theoretical framework of behaviour change during data collection and many have not specifically included Z-drug patients.
This study forms part of a larger multiphase project, which aims to develop a theory-based intervention to support the discontinuation of long-term BZRA use in primary care. The project aligns with the UK Medical Research Council's complex intervention framework. 17 The first phase involved systematically reviewing the evidence-base for brief interventions targeting long-term BZRA use in primary care. 14 The second phase seeks to incorporate a theorybase into the intervention development process by analysing behavioural determinants of discontinuing BZRA use using an integrated framework of behaviour change theories, the Theoretical Domains Framework (TDF). 19,20 The TDF was developed from 33 psychological theories relevant to behaviour change. 19 These theories were condensed into domains that are considered mediators (i.e., barriers, facilitators) of behaviour change. The original TDF consists of 12 domains 19 and a second version (TDFv2), which was produced after a validation exercise, consists of 14 domains. 20 Both are widely used and researchers are free to choose between them based on familiarity and preference. 21 The previous systematic review retrospectively coded included interventions using the TDF to identify domains that may have been implicitly targeted. 14 To ensure that future interventions target relevant behavioural determinants and are theory-based, discontinuation of long-term BZRA use needs to be examined in greater detail from patients' perspectives. This study aimed to explore patients' views and experiences of long-term BZRA use. The objectives were to: 1. Identify mediators (i.e., barriers and facilitators) of discontinuing long-term BZRA use from the perspective of both current and previous users.

2.
Select key theoretical domains to target to change the target behaviour (i.e., discontinuation of long-term BZRA use).

| METHODS
Semistructured interviews were conducted with two patient cohorts using separate topic guides (Supporting Information Appendices S1 and S2) based on TDFv2. 20 The first cohort consisted of participants that had successfully discontinued long-term BZRA use. The second cohort comprised participants taking BZRAs on a long-term basis at the time of interview. The methods were adapted from a previous TDF-based intervention development study 22

| Sampling and recruitment
A multistrand convenience sampling method (outlined below) was used to recruit eligible participants. The following inclusion criteria applied: ≥18 years old, prescribed BZRAs for a period equating to ≥3 months' supply in the previous year, community-dwelling in the Republic of Ireland. Patients receiving BZRA prescriptions for any condition (including anxiety and insomnia) were excluded from participation in the following circumstances: cognitive impairment, epilepsy, serious mental illness (e.g., prescribed antipsychotics or lithium) or receiving opioid substitution treatment. This was because the care of these groups and the role of BZRAs within their treatment plans may have been different from individuals who had been prescribed a BZRA for conditions, such as anxiety or insomnia, and remained on the medication for an extended period. Recruitment of both cohorts was conducted concurrently from September 2019 to May 2020.

| Community pharmacy strategy
The researcher (T. L.) identified and telephoned community pharmacies in the greater Dublin area (sequenced according to postcode) using a publicly available register, and provided a brief summary of the study to the pharmacist. Pharmacists that expressed interest in facilitating patient recruitment were posted an information pack containing copies of the participant cover letter and information sheet. Pharmacists who agreed to facilitate patient recruitment provided written informed consent.
Recruited pharmacists used an eligibility screening algorithm to assess known patients currently or previously receiving BZRA prescriptions against the above inclusion criteria and provided patients with relevant study documentation (i.e., information sheet, consent form) when they presented in the pharmacy to pick up their regular prescription. Patients interested in participating were asked to provide a contact telephone number to the pharmacist for them to share with the researcher who would then follow-up with patients directly and complete the recruitment process.

| General practice strategy
To supplement the pharmacy recruitment strategy, a convenience sample of three general practitioners (GPs) known to the research team were contacted to seek assistance with recruitment. This followed a similar process to the above pharmacy strategy. The key difference was that eligible patients who received study information were instructed to contact the researcher directly if they were interested in participating. GPs had no further involvement in recruitment and, therefore, were not formally recruited.

| Social media strategy
Social media was used to maximize recruitment, particularly in terms of previous long-term BZRA users who may no longer have been regularly attending a general practice or pharmacy. A brief summary of the study was disseminated through the research team members' individual Twitter accounts and interested individuals were instructed to make direct contact by telephone or email. The same brief summary of the study information was advertised on an Irish-based online public message forum (www.boards.ie).
The researcher also emailed podcasts with a specific interest or focus on mental health and/or BZRA use (Supporting Information Appendix S3) seeking assistance in disseminating study information to their listenership together with the researcher's contact details. Interview topic guides (Supporting Information Appendices S1 and S2) were based on TDFv2 20 (Table 1) and piloted on two individuals. Although separate topic guides were designed for each cohort, they followed a similar line of questioning comprising three key areas: (1) information about previous/current BZRA use; (2) perceived barriers and facilitators to discontinuing long-term BZRA use; (3) interventions/strategies to facilitate discontinuation of longterm BZRA use.

| Data collection and analysis
All interviews were audio-recorded, transcribed verbatim and deidentified by assigning a unique code. Participants had the opportunity to review and edit their transcripts. Participants were not asked for feedback on the findings.
Interviews within each cohort were first analysed separately, before findings within each data set were compared. Initially, the researcher engaged in a familiarization process by reading and rereading transcripts and listening to audio recordings. The first stage of analysis involved a framework analysis using a deductive approach. 24  The second stage involved a content analysis of charted data using an inductive approach. 25 This sought to capture any emerging themes from the charted data. Subthemes were identified and recurrent beliefs expressed by participants regarding each domain were reported. The content analysis was led by the researcher and independently reviewed by the research supervisor. Throughout the analysis, coding was compared and any disagreements were resolved through discussion. Data saturation was assessed by reviewing the charted data under each domain to determine whether similar information was elicited across interviews and if additional interviews were likely to elicit new information.

| Identification of key theoretical domains
It was intended at the outset that the selection of key theoretical domains would involve a consensus-based approach within the research team. The number of times domains were indexed in the framework analysis was to be used as a basic indicator of a domain's particular relevance. The results from the content analysis, such as subthemes and recurrent beliefs, expressed by participants in relation to domains were also to be considered during the process of selecting key theoretical domains to target as part of a future intervention. As noted in Section 3 below, this process did not proceed as planned.

| RESULTS
Twenty-eight participants were recruited comprising 13 previous and 15 current BZRA users. Fifty-seven percent (n = 16) of participants were female and the median age was 46 years (range 24-65). The median duration of BZRA use was 5 years (range 3 months to 21 years). The most common clinical indications for BZRA use were insomnia and anxiety. Each cohort's demographic characteristics were broadly comparable ( Table 2).
Most participants were recruited through social media (10 participants through Twitter, 4 participants through www.boardsi.e, 9 participants through the Instagram account of a prominent GP who shared the study tweet as an Instagram story). Two community pharmacies were formally recruited (out of 232 pharmacies contacted) and only two patient participants were subsequently recruited. Three additional participants were recruited after hearing about the study by word of mouth. No participants were recruited through general practices (n = 3) or podcasts (n = 3). In the latter case, this was because those who contacted the researcher were not residents in Ireland and, therefore, not eligible to participate.

| DISCUSSION
This study provides a detailed exploration of barriers and facilitators to discontinuing long-term BZRA use. The study forms part of a systematic approach to developing a theory-based intervention to support BZRA discontinuation. Previous research in this area has highlighted a lack of transparency in relation to the intervention development process and a lack of theoretical underpinning. 14,26 Moreover, previous qualitative studies have not employed a theoretical framework of behaviour change during data collection and many have not specifically included Z-drug patients. 18 38 and evidence to support its use is lacking. 39 However, in the current study, the rationale for a reserve BZRA supply was not to act as a form of maintenance therapy but for use in exceptional circumstances. This option should be explored further as part of strategies to limit and reduce BZRA prescribing and use.
In addition to gradual dosage reduction, some previous users reported substituting BZRAs for other medications (e.g., antidepressants). However, high-quality evidence to support using pharmacological interventions in discontinuing BZRA use is lacking. 13 Most previous users reported engaging with various other resources and supports in discontinuing BZRA use. Several of them had availed of counselling, typically involving CBT, which they often found beneficial. CBT in conjunction with gradual dosage reduction has been found to be effective at reducing BZRA in the short-term (threemonth follow-up); however, existing evidence indicates that this effect is not sustained over longer periods. 12 In contrast, fewer current users reported availing of such supports. One participant who had been referred to a counselling service was still waiting to be contacted by the service. Issues relating to the availability and accessibility of counselling services in Ireland are well recognized. [40][41][42] This clearly needs to be addressed as part of a holistic approach to reducing long-term BZRA use and improving mental healthcare. In the absence of adequate resources, previous users may be able to offer important peer support to current users attempting discontinuation. 43 This needs to be explored further in future research.
This study has also provided useful information regarding recruitment strategies for BZRA-related research. In anticipation of recruitment challenges, a multistrand sampling approach was used.
Participant recruitment via general practices and community pharmacies, which commenced before the COVID-19 outbreak in Ireland, was largely unsuccessful. It was beyond the study's scope to formally establish reasons for this. However, an unexpected study finding was the vital role of social media in recruiting participants. This may relate to the use of online resources and support groups among individuals experiencing medication-related withdrawal symptoms. 44,45 There is some evidence to suggest that social media may be particularly effective for recruiting so-called 'hard to reach' participants through more conventional methods. 46

| Strengths and limitations
The study's strengths include the contextually rich and descriptive data obtained. All interviews were coded independently by two researchers, which enhanced the robustness of the analysis. The analysis of two participant cohorts provides comprehensive insights into behaviour change mediators that can be targeted by a future intervention using behaviour change techniques. This will be outlined in a future paper. In terms of limitations, it must be noted that the results cannot be generalized to the wider population of long-term BZRA users. Another limitation is that older adults who account for the largest proportion of long-term BZRA use 5,6 were not recruited, which reduces the transferability of findings to this cohort. This may have been a consequence of the reliance on social media. While >70% of Irish adults aged ≥50 years have internet access in their own homes, internet use declines with increased age, with social media use experiencing the largest age-associated decline. 47 As data collection overlapped with the first national COVID-19 lockdown in Ireland, it was not feasible to remedy this. Future research should look to develop strategies to optimize recruitment for BZRArelated research and explore the transferability of findings to older adults.

| CONCLUSION
The study reports on barriers and facilitators to discontinuing BZRAs from the perspectives of current and previous long-term users. The findings highlight the challenging nature of BZRA discontinuation and a multitude of barriers that exist. As an individual's experiences of determinants of BZRA discontinuation vary, future interventions may need to be tailored based on an individualized assessment of behaviour change mediators. Future work will involve developing a theory-based intervention to support BZRA discontinuation in primary care.