Supporting General Practitioners to Deprescribe Benzodiazepines and Z‐Drugs in Primary Care: Findings From a Modified Delphi Study

ABSTRACT

Long‐term use of benzodiazepines and Z‐drugs (collectively known as benzodiazepine receptor agonists; BZRAs) is associated with a range of adverse effects including dependence and withdrawal on stopping or reducing the dose. Deprescribing is recommended to improve patient outcomes and reduce medication‐related harm, but its implementation in primary care is hindered by various barriers. This study aimed to identify the strategies that Australian general practitioners (GPs) agree were most useful in supporting them to deprescribe BZRAs. A modified Delphi was conducted online over three rounds with 38 Australian GPs. Participants were registered GPs with ≥ 10 years' experience in either addiction, mental health, sleep, pain management, or aged care. Participants rated and provided feedback on 61 implementation strategies over three rounds until they achieved either consensus or stability. Included strategies were mapped to the theoretical domains framework to understand their links to behaviour change mechanisms. Twenty‐five strategies reached a consensus for inclusion indicating their importance to supporting GPs to deprescribe BZRAs. The most important strategies included extended consultation times, provision of clear deprescribing protocols and establishing practice‐level agreement on deprescribing policies.

Findings provide a prioritised set of implementation strategies to address the barriers to BZRA deprescribing and offer a strong foundation for future research and policy to improve deprescribing in primary care.

Plain Language Summary

Benzodiazepines and Z‐drugs (or benzodiazepine receptor agonists; BZRAs) are commonly prescribed in primary care for conditions such as anxiety and insomnia, but with long‐term use, the harms can outweigh the benefits. Deprescribing is often required to curb these harms, but a range of barriers make deprescribing difficult for general practitioners (GPs). This study involved a three‐round Delphi survey with 38 Australian GPs to identify the most useful strategies to support them to deprescribe BZRAs. Of the 61 strategies reviewed, 25 were included in the final list and provide practical recommendations to strengthen BZRA deprescribing support in primary care.

1. Background

For over 40 years, there have been calls for increased vigilance in the prescribing of benzodiazepines, and more recently, Z‐drugs—collectively known as benzodiazepine receptor agonists (BZRAs) [1]. BZRAs are commonly prescribed to manage insomnia and anxiety; however, guidelines internationally recommend they should not be prescribed for longer than 4 weeks due to a range of adverse effects, including the risk of dependence and withdrawal on stopping or reducing [2, 3, 4]. Despite these recommendations, long‐term use (defined as regular use for 6 months or longer) [5] of BZRAs is not uncommon in both Australia [6] and internationally [7, 8, 9, 10], with estimates that long‐term use occurs in 30%–55% of all BZRA use [5, 11].

The potential adverse effects of BZRAs are multifaceted, impacting individuals' physical, cognitive and psychological functioning. Cognitive impairments can include learning impairment, anterograde amnesia, drowsiness and delayed reaction times [2, 12], which can increase the risk of road accidents by 60%–80% [13]. In older adults, BZRAs pose a significant risk for falls and fractures [14, 15] and are linked to a possible risk for dementia, although causal inferences remain uncertain [16, 17]. Prolonged use can aggravate depression and anxiety symptoms, even after discontinuation, and withdrawal from even therapeutic doses can be protracted and debilitating [18, 19, 20]. The most significant concern, however, is how frequently BZRAs contribute to overdose deaths. Although BZRAs are rarely fatal alone, the risk increases substantially when used in combination with other central nervous system depressants, such as opioids or alcohol. Over the past 20 years, deaths involving BZRAs increased sixfold, fivefold and fourfold in the United States, Canada and Australia respectively [21, 22, 23], and in the United Kingdom, these statistics almost trebled in that same time [24].

Given these harms, in many cases, deprescribing BZRAs is vital for enhancing patient outcomes. Defined as the systematic process of reducing or stopping medications where the harms outweigh the benefits, deprescribing was initially proposed to reduce polypharmacy in older adults [25] but is increasingly recommended to address any potential inappropriate medication use [26]. Numerous studies show that the most effective way to deprescribe BZRAs is to gradually taper the medication at a rate adjusted to the individual, which can be supplemented with additional support such as psychological treatment [27, 28]. With good evidence supporting deprescribing, primary care physicians (PCPs) (e.g., general practitioners (GPs) or family physicians) still require clear guidance on how to effectively implement the deprescribing process [29, 30]. Although several deprescribing guidelines exist, few provide actionable recommendations on how to deprescribe BZRAs [26]. This was until the release of the Maudsley Deprescribing Guidelines (MDG) [31] and the Joint Clinical Practice Guideline on Benzodiazepine Tapering [32], which provide the most detailed recommendations to‐date on how to safely taper BZRAs. Translation of these guidelines into routine practice should thus be a priority for clinicians, researchers and policymakers, with emphasis on identifying strategies that address the myriad of barriers to deprescribing BZRAs.

It is evident that there are many challenges preventing PCPs from deprescribing BZRAs in primary care [33, 34]. These include difficulties in knowing how to initiate a deprescribing conversation [35, 36], inconsistent communication about prescribing information between physicians and patients [37] or between physicians and other healthcare providers [38], hesitance to stop a medication started by a specialist [39, 40], negative perceptions of deprescribing [41], persistent prescribing cultures [30, 42, 43], perceived clinical issues with cessation of the medication [36, 39], anticipated patient resistance or fear of damaging the therapeutic relationship [43, 44, 45, 46] and a lack of physician confidence, skills or knowledge to deprescribe [35, 36, 47].

System‐level barriers in primary care can further hinder the implementation of BZRA deprescribing, such as inadequate time for deprescribing discussions [41, 44, 47], insufficient reimbursement/funding [38], barriers to accessing other evidence‐based treatment (i.e., psychological therapy) [36, 43, 47], lack of care coordination and access to specialists [46] and unclear roles for staff and PCPs regarding deprescribing responsibilities [37]. Likewise, despite the availability of deprescribing resources, negative perceptions of deprescribing and PCP apprehension persist in primary care practice [34]. As a result, many have called for greater effort to support deprescribing in primary care [33, 34, 48].

Implementation science plays a pivotal role in translating guidelines into practice and seeks to understand and promote the adoption, use and sustainability of evidence‐based practices within real‐world healthcare settings [49]. The discipline encompasses various methodologies and frameworks that aim to identify barriers and facilitators to implementation, tailor strategies to support uptake and sustainability of evidence‐based practices and evaluate the impact [50]. Among these, the theoretical domains framework (TDF) [51, 52] is an integrative framework commonly used to assess the determinants (i.e., barriers and facilitators) of behaviour to then inform the design of strategies that support implementation. The TDF has been widely applied across a variety of healthcare settings [53, 54], in particular in primary care [55], and has previously been used to evaluate barriers and facilitators to deprescribing BZRAs in hospital settings among a range of healthcare practitioners [36, 56]. The TDF not only provides a strong theoretical basis for implementation research but is also a helpful framework through which to synthesise evidence [57]. Importantly, determinants identified via the TDF can be operationalised by mapping them to their corresponding behaviour change technique (BCT). BCTs are replicable, observable and evidence‐based techniques designed to facilitate behaviour change and are key to understanding how and why implementation strategies work [58]. Mapping strategies to the TDF offers a structured approach to selecting implementation strategies that directly address relevant determinants. Linking these strategies to BCTs helps to clarify their mechanisms of action and ensures strategies are theory informed [59].

Whilst many studies have utilised implementation strategies to address barriers and facilitators to BZRA deprescribing in primary care, few have focused on understanding if, how and why those strategies were effective [60]. As a result, there is limited insight about which implementation strategies are the most effective in supporting PCPs to deprescribe BZRAs. To advance knowledge in this field, the first step is to reduce the vast list of implementation strategies described in the empirical literature. To achieve this, the current study aimed to gain a consensus on the strategies that Australian GPs perceive as most useful in supporting BZRA deprescribing in primary care. A three‐round Delphi was conducted using the TDF as the organising framework.

2. Material and Methods

2.1. The Delphi Methodology

The Delphi method is widely used for determining health research priorities [61]. It involves an iterative process, whereby experts are repeatedly surveyed on a subject until consensus is reached and is particularly useful for capturing practice‐based evidence in areas with little or no empirical evidence [62] In this study, a modified Delphi was used where the list of implementation strategies was initially generated from the literature and then supplemented by suggestions made from participants in the first round. Reporting criteria adhered to the guidance for conducting and reporting of Delphi studies (CREDES) [63], and the study was conducted in accordance with the Basic and Clinical Pharmacology and Toxicology policy for experimental and clinical studies [64]. The Delphi process was guided by the six‐step approach outlined by Belton et al. [65], which provides direction on questionnaire development, method of delivery, attrition management and data analysis. The research team included six GP advisors who provided specialist knowledge and oversight, as recommended when identifying priorities in health research [66]. It was decided a priori to conclude the Delphi after three rounds to provide GPs with clarity on their commitment and reduce the burden that might result from a protracted process.

2.2. Survey Development

A scoping review [60] was conducted to identify implementation strategies used internationally in primary care to support BZRA deprescribing. The list of strategies was then cross‐referenced with the barriers and facilitators (see Table S1) to deprescribing documented in four relevant systematic reviews [33, 34, 67, 68]. This cross‐referencing resulted in a comprehensive list of potential strategies to support BZRA deprescribing. To ensure each strategy was actionable and linked to an evidence‐based BCT, they were operationalised using the Theory and Techniques Tool [69]. This is an online interactive tool that links BCTs (e.g., goal setting; problem solving) to their mechanisms of action (i.e., the TDF). Minor refinements and adjustments to the wording of some implementation strategies were made to ensure they aligned with the tool, and some BCTs were deemed to be incompatible with deprescribing (i.e., biofeedback). This process generated an initial pool of 49 strategies that were mapped against the relevant TDF (see Table S2). This formed the baseline survey that was piloted with the research team, GP advisors and six laypeople resulting in changes to wording, amount of information provided and survey layout.

2.3. Participants

Australian GPs were recruited through purposive sampling and were defined as ‘experts’ based on the following three criteria: (1) medical practitioners trained in general practice with general registration with the Australian Health Practitioner Regulation Agency (AHPRA); (2) a minimum of 10 years' experience/special interest in at least one of the five relevant areas to BZRAs, including addiction, mental health, aged care, sleep and/or pain; and (3) English fluency. We aimed to recruit at least 35 participants, based on expected attrition rates of up to 30% [70, 71], to ensure our final sample was above what has been deemed sufficient in similar research (n = 23 [72]). Potential participants were identified through online searches and research team networks; these participants were categorised (i.e., area of experience, location and organisations), ranked and prioritised [73] and then systematically contacted to achieve a diverse group of experts until the target sample size was achieved [62, 74].

Participants were recruited via email inviting them to be part of an important project shaping BZRA deprescribing recommendations. This sought to encourage interest and ownership in the study, as responses from each round relied on their contribution and commitment [75]. Additionally, participants had the opportunity to win a gift voucher (or donate to their chosen charity) and be acknowledged in any associated publications and presentations [76, 77].

2.4. Consensus

Agreement and stability were both measured to determine consensus [78, 79, 80]. A 5‐point Likert scale was used, ranging from 1 (not at all useful) to 5 (essential), with a non‐neutral mid‐point 3 (moderately useful) and an option for ‘do not know/unsure’. Agreement was defined as any implementation strategy being endorsed by 75% or more of experts between two consecutive scale points (i.e., between very useful and essential) [78, 79, 80]. Stability of responses was also assessed so that disagreement could exist. Stability was assessed at both the group and individual level. Group stability was defined as less than 15% movement between rounds, suggesting that opinions were not likely to be affected further [81, 82], with individual stability assessed using a Wilcoxon matched‐pairs signed‐rank test [83, 84]. Where these two measures of stability conflicted, the research team resolved the conflict collaboratively, and within‐person change was given greater weight than group‐level change.

Both agreement and stability were reviewed between each round to determine if strategies would be included or excluded in the subsequent round. Strategies were excluded from future rounds if consensus was reached between any of the lowest three scale points (‘not at all useful’ to ‘moderately useful’; i.e., agreed to have low usefulness to GPs) and included if consensus was achieved across any of the top three scale points (‘moderately useful’ to ‘essential’; i.e., agreed to be useful to GPs). Likewise, remaining strategies that demonstrated within‐group and/or within‐person stability across rounds (i.e., GPs were not changing their opinion based on the feedback from their peers) were excluded from subsequent rounds to reduce participant burden.

2.5. Procedure

Participants completed three rounds of online surveys between February and July 2024. Participation was conditional on participants reading a Plain Language Statement and providing their informed consent. In the first round, participants were asked to rate the usefulness of 49 implementation strategies drawn from the literature. To reduce participant fatigue, strategies were presented within seven contextual and descriptive categories (see Table 1). For a full list of how each strategy was categorised, see Table S2.

TABLE 1.

Descriptive categories used to cluster implementation strategies (n = 49).

Category	Description	N (%)
Education and training	Strategies aimed at expanding GPs' knowledge, practical skills and competency	5 (10.2)
Resources and support	Strategies related to physical or digital tools/resources to be provided to GPs or implemented in consultations	11 (22.5)
Multidisciplinary collaboration	Strategies that facilitate or encourage communication and collaboration between various care providers	4 (8.2)
Remuneration	Strategies that remunerate or incentivise GPs	3 (6.1)
Administrative support	Strategies that are implemented by practice staff to support GPs with deprescribing processes	10 (20.4)
Practice policies and procedures	Broader strategies that focus on adjusting the culture of a practice toward deprescribing	10 (20.4)
Patient resources and support	Physical or digital resources provided to patients and external strategies that facilitate behaviour change	6 (12.2)

Note: N = number of implementation strategies in each category.

Abbreviation: GP, general practitioner.

During the first round, participants were invited to propose additional strategies that were not already included in the survey. After each round, strategies that did not achieve consensus (i.e., ≥ 75% consensus between two scale points) progressed to the subsequent round for re‐rating alongside additional feedback. This included the range and frequency of previous group responses, each participants' previous response and a synopsis of participant comments. Providing these comments afforded insight and rationale, as opinion change is unlikely to occur without strong causal reasoning [85]. Likewise, comments were presented in ‘for’ and ‘against’ columns rather than alongside individual ratings to reduce normative influence, and anonymity removed any undue influence or collusion (see Figure S1 for an example). All communication with participants was conducted via email, and up to five follow‐up attempts were made for nonresponders. The study was approved by the relevant Human Research and Ethics Committee (HEAG‐H 202_2023).

3. Data Analysis

3.1. Quantitative Analyses

Quantitative data were analysed using SPSS v29.0.2.0 [86]. To assess consensus, frequencies were checked, and percentages were calculated by totalling experts' Likert scores. Group stability was assessed by calculating the percentage change, where the difference between endorsements for each round was aggregated to form total units of change and divided by 2 to account for net person changes [81]. The net changes were then divided by the total number of participants to determine the percentage of change (i.e., < 15%). Individual level change was determined by a Wilcoxon matched‐pairs signed rank test p value of ≤ 0.05 [83, 84].

3.2. Qualitative Analyses

In the present study, qualitative feedback was analysed independently by two researchers (EO, MP). Participants' comments were categorised as either supporting or opposing each strategy, and then summarised and aggregated, with original wording retained where possible [87]. Responses challenging conventional thinking were included [85], whilst repetitive comments were eliminated to minimise same thinking. Feedback that focused on ‘how’ a strategy might be modified or translated were removed to ensure participants focused on the usefulness of the implementation strategies as opposed to their feasibility. The additional strategies suggested by the participants were evaluated by the research team to confirm that they were (a) not duplicated, (b) aligned with the study scope (i.e., could be delivered within primary care setting) and (c) clearly articulated [61]. Where they were not, the researchers reviewed and adjusted the description accordingly. These additional strategies were then added to subsequent survey rounds.

At the end of data collection, strategies across all rounds that achieved consensus between ‘very useful to essential’ were designated as ‘first‐priority’ implementation strategies. Among strategies that were rated between ‘moderately’ to ‘very useful’, those which had higher weighting on ‘very useful’ were designated as ‘second‐priority’, and those with a higher weighting on ‘moderately useful’ were termed ‘third‐priority’ strategies. When mapping the final list of included strategies to the TDF, we identified relevant domains based upon established ‘green’ links in the Theory and Techniques Tool, indicating there is sufficient evidence regarding the specific BCT‐TDF link. In the absence of established links, ‘orange’ (i.e., inconclusive) links were then reviewed and considered as potential evidence for the specific BCT‐TDF link.

4. Results

Of the 40 people who initially agreed to participate, 38 GPs consented and completed Round 1 of the Delphi, reflecting a heterogenous sample (see Table 1). Retention was high across all three rounds, with 38 GPs completing Round 2 and 37 completing Round 3 (97.4%). Notably, 16 GPs (42.1%) indicated being current or former practice owners, which exceeds the national average in Australia of approximately 25% [88], and the sample had a slightly higher representation of female GPs (55.3%) than the Australian average (48%) [89].

In addition to the original 49 strategies identified in the literature, participants' suggestions led to 12 more strategies included in the second round for rating. A flow chart displaying the movement of implementation strategies across the Delphi process is shown in Figure 1. Consensus was achieved for three strategies as being first priority for supporting BZRA deprescribing (see Table 2), whilst an additional 22 strategies were considered either very or moderately useful and classified as second and third priorities respectively. An overview of the consensus levels and range for each strategy over the rounds is shown in Table 3. Following the third round, the Delphi process was deemed complete; however, seven strategies had not reached a consensus. Of these, five demonstrated stability and thus would have been excluded from a subsequent round, meaning only two strategies, both suggestions made by GPs in Round 1 (see Table S3), would have required an additional round.

FIGURE 1.

Flowchart of the Delphi process to identify the most useful implementation strategies for deprescribing BZRAs in primary care.

TABLE 2.

Characteristics of participants.

Characteristic	(N = 38)
Age in years (M ± SD)	55.61 ± 9.57
Age range in years	37–72
Identified gender, n (%)
Man	17 (44.74)
Woman	21 (55.26)
Years of experience (M ± SD)	26.68 ± 10.26
Specific interests, n (%)
Mental health	24 (34.8%)
Addiction	18 (26.1%)
Geriatrics/aged care	6 (8.7%)
Sleep disorders/respiratory health	4 (5.8%)
Pain management	9 (13.0%)
Other	8 (11.6%)
Place of work
Group practice	21 (38.9%)
Solo practice	5 (9.3%)
Tertiary education institute	5 (9.3%)
Aboriginal medical service/controlled health organisation	2 (3.7%)
Aged care facility	3 (5.5%)
Public hospital	6 (11.1%)
Private hospital	1 (1.8%)
Community health service	7 (12.9%)
Other	4 (7.4%)
Region of work
Capital city	24 (58.5%)
Other metro area (pop > 100 000)	9 (21.9%)
Regional (pop 5001–99 999)	6 (14.6%)
Rural (pop < 5000)	2 (4.9%)

Note: Area of special interest, place of work, and region of work are not mutually exclusive categories so do not total 100%. Other areas of special interest were comprehensive care (n = 1), aboriginal health (n = 1), and women and children's health (n = 1). Other places of work were Drug and Alcohol health service (n = 1), the prison sector (n = 1), and self‐employed mindfulness consultation (n = 1).

TABLE 3.

Implementation strategies that achieved consensus for inclusion and assigned first, second, and third priority (n = 25).

Implementation strategy	Agreement level			TDF domain(s)
	Moderately useful	Very useful	Essential
First‐priority strategies
Extended consultation times for patients during the deprescribing process (e.g., deprescribing consults/complex medicine consults)	13.2%	26.3%	55.3%	SPRI, ECR
Provision of a tapering protocol (e.g., with dose equivalencies, recommended reduction rates, goals of reduction)	15.8%	34.2%	47.4%	Knowledge, Skills, BaCa, MADP, ECR
*GPs review BZRA guidelines and agree on a broad practice policy from a suite of options (e.g., see same GP for BZRA script; no sudden cessation; tailored approaches to tapering)	8.1%	32.4%	48.6%	Knowledge, SPRI, BaCo, ISoC, Goals, MADP, ECR, SocIn, Emotion
Second‐priority strategies
GP payment for completing medication reviews	24.3%	70.3%	0%	Reinforcement, ISoC
*Online modules (incl. videos/webinars, case studies) with CPD points	43.2%	45.9%	0%	Knowledge, Skills, SPRI, BaCa, BaCo, Reinforcement
Offering patients access to free counselling or deprescribing support services following appointments where deprescribing is discussed	18.4%	68.4%	13.2%	ECR
Practice providing independent clinical consultation for deprescribing decision‐making (i.e., where patient can meet with MDT or a deprescribing expert 1:1)	34.2%	50.0%	5.3%	ECR
Academic detailing (i.e., a one‐off educational visit from a specialist in BZRA deprescribing)	36.8%	47.4%	2.6%	Knowledge, Skills, SPRI, BaCa, BaCo, ECR, SocIn, Emotion
*Increased funding for longer consultations (e.g., deprescribing consults, complex medicine consult levels C and D)	16.2%	67.6%	13.5%	Reinforcement
Information on how to conduct deprescribing consultations (e.g., a step‐by‐step guide breaking down deprescribing into stages)	24.3%	59.5%	8.1%	Knowledge, Skills, BaCa, BaCo, ISoC, MADP, ECR,
Decision‐making tool/algorithm (i.e., to determine if patient appropriate for deprescribing)	36.8%	44.7%	5.3%	Knowledge, Skills, MADP, ECR
Practice adopting a deprescribing protocol (i.e., for all GPs to agree to and follow)	28.9%	50.0%	0%	Knowledge, SPRI, BaCa, ISoC, Goals, MADP, ECR, SocIn, Emotion, BR
Printed resources for patients to use in consultations with patients (e.g., to discuss risks/benefits of long‐term use, other evidence‐based treatments)	36.8%	42.1%	2.6%	MADP, ECR
*Patient resources that focus on underlying conditions (i.e., insomnia/anxiety)	21.6%	56.8%	5.4%	ECR
*Mentoring or peer‐to‐peer led education (i.e., to build GP capacity by more experienced GPs)	27.0%	51.4%	8.1%	Knowledge, Skills, SPRI, BaCa, Optimism, BaCo, ECR, SocIn, Emotion, BR
*Checklist to ensure prescribing occurs after making recommendation regarding lifestyle interventions	29.7%	46.8%	8.1%	ECR
Multidisciplinary team–based approach to deprescribing (e.g., pharmacist generates tapering plan, psychiatrist consults on complex cases)	34.2%	42.1%	7.9%	SPRI, ISoC, ECR, SocIn
Digital resources for patients (e.g., online hub with educational videos, fact sheets, relaxation techniques)	31.6%	44.7%	13.2%	ECR
*Online BZRA calculator and tapering regime calculator triggered by software	13.5%	62.2%	5.4%	Knowledge, Skills, MADP, ECR
An educational dossier (including fact sheets, tapering protocols, evidence on deprescribing, decision‐making algorithm)	27.0%	48.6%	2.7%	Knowledge, Skills, BaCa, BaCo, MADP, ECR
Third‐priority strategies
Case conferencing (i.e., online meetings with MDT to review complex cases)	57.9%	31.6%	5.3%	Knowledge, Skills, SPRI, BaCa, Optimism, BaCo, ISoC, ECR, SocIn, BR
An online self‐help program to support patients reducing their BZRA that can be shared with the GP	50.0%	31.6%	0%	ECR
Ongoing outreach support from specialist in BZRA deprescribing (e.g., to provide PCPs with consultations, decision‐making support)	39.5%	36.8%	21.1%	Knowledge, Skills, SPRI, BaCa, Optimism, BaCo, ISoC, Goals, MADP, ECR, SocIn, BR
Digital resources via an online hub to support GPs with deprescribing and managing withdrawal	44.7%	31.6%	10.5%	Knowledge, Skills, BaCa, BaCo, ECR
*Mandatory requirement to review RTPM (e.g., SafeScript) prior to writing scripts for BZRAs	48.6%	27.0%	13.5%	Reinforcement, MADP, ECR

Note: *strategies recommended by participants in Round 1; percentages in bold indicate the combined ratings that reached consensus.

Abbreviations: BaCa, beliefs about capabilities; BaCo, beliefs about consequences; BR, behaviour regulation; ECR, environmental context and resources; ISoC, intention and stage of change; MADP, memory, attention and decision processes; MDT, multidisciplinary team; RTPM, real‐time prescription monitoring; SocIn, social influences; SPRI, social/professional role and identity.

5. Discussion

This modified Delphi study aimed to determine which implementation strategies Australian GPs considered the most useful to support BZRA deprescribing in primary care. Sixty‐one strategies were reviewed by participants, and 25 ultimately met inclusion criteria. Of these, three achieved consensus as ‘first‐priority’ strategies. Specifically, extended consultation times, provision of clear deprescribing protocols and establishing practice‐level agreement (determined by GPs) on the deprescribing approach. Included strategies mapped to all TDF domains, illustrating that many factors influence the process of deprescribing BZRAs and that supporting GPs with multifaceted strategies may be necessary to overcome existing barriers. The most common TDF domains were Environmental Context and Resources, followed by Knowledge and Skills.

Participants overwhelmingly agreed that deprescribing requires time and that extended consultations are necessary to support effective deprescribing. In fact, all three first‐priority strategies were system‐level, aligning with the Environmental Context and Resources. The importance of funding extended consultations cannot be overstated, not only because this facilitates deprescribing opportunities but it also acknowledges the complexity and clinical importance of deprescribing BZRAs. Participants’ preference to have a practice‐wide approach to deprescribing emphasised the need for consistency of care and a shared responsibility in their approach to BZRA deprescribing. This more broadly reflects the need for a cultural shift [33, 34] and that to embrace deprescribing GPs need to know they are in alignment with their peers. It may also mitigate patient resistance by framing any changes in care as a ‘practice policy’. Together, these system‐level supports would provide the essential scaffolding to enable GPs to deprescribe BZRAs and address a need observed globally, healthcare systems that prioritise and improve capacity to appropriately deprescribe [36, 90].

The last of the first‐priority strategies ‘provision of deprescribing protocols’ echoed the need for consistency and structure, as tapering protocols provide step‐by‐step guidance on how to safely reduce BZRAs. This finding is not unique to Australia, as PCPs internationally report the need for additional guidance on how to develop a tapering plan for BZRAs [91]. Clinical tools providing information and guidance on tapering BZRAs are now widely available; however, few have included end‐users in their design and development. This means as point‐of‐care tools, they do not maximise their clinical utility to function in multiple ways (e.g., communication tool, decision‐aid, planning and goal setting) that might otherwise provide a framework for PCP to use in guiding the patient through the deprescribing process and support PCPs to overcome the challenge of initiating deprescribing conversations [35, 36, 56]. Importantly, any tapering protocol must balance the need for evidence‐based guidance whilst promoting a flexible person‐centred approach so that PCPs can tailor the deprescribing plan to each person's unique presentation [31, 59].

Consensus was also reached on 22 second‐ and third‐priority implementation strategies, most of which (86.4%) mapped to Environmental Context and Resources. Consistent with findings in the Mather et al. [55] systematic review, this study highlighted the need for detailed, evidence‐based resources for both GPs and their patients. Specifically, there was strong support for practical clinical tools, in addition to protocols, such as decision‐aids, equivalency calculators and fact sheets to support GPs to deprescribe BZRAs. Consumer‐facing resources were also strongly favoured, including educational materials, self‐help resources and programs, as well as access to specialised support and counselling. In fact, access to specialist support was the only strategy that achieved 100% agreement. However, presently, these are limited, leading to repeated calls for additional support services [32, 92, 93]. Australia only has two catchment‐based services providing free support for BZRA withdrawal (Reconnexion and the Medication Support and Recovery Service), and the UK only has one specialist service providing comprehensive support (e.g., PostScript 360) and two smaller catchment‐based services (e.g., Benzodiazepine and Opiate Withdrawal Service (BOWS), North‐East London NHS Foundation Trust (NELFT clinic). In other countries, some fee‐for‐service programs exist (e.g., Outro Health), but otherwise, specialised support is limited to individual clinicians, online forums (e.g., Benzo Buddies), podcasts (e.g., Let us talk withdrawal) and websites providing education and advocacy (e.g., Benzo Info Coalition, and BE‐SAFE). Collectively, these findings emphasise the importance of resourcing GPs and their patients to facilitate informed decision‐making, collaboration and person‐centred care in successful BZRA deprescribing.

Educational approaches emerged as a key area, equally for the knowledge and skills as the next most common reported TDF domains. This was unsurprising given a common prescriber‐level barrier to deprescribing is the confidence, skills and knowledge to enact a deprescribing plan [35, 36, 47]. Similar to Sibille et al. [59], participants favoured academic detailing, practical guidance on how to conduct deprescribing consultations and online modules that could be accessed at their convenience. This contrasts with research interventions, which commonly use in‐person workshops [94, 95], yet this format is limited in scalability and is often dependent on the research team for delivery. Qualitative feedback on strategies linked to knowledge reflected the need for practical and ‘bite‐sized’ information that could be used as point of care tools. Strategies aimed at developing GP skills to engage with patients through the deprescribing journey were favoured when the education was interactive and specific, either through one‐on‐one or interdisciplinary and peer methods of learning, again, consistent with existing evidence [96]. Overall, these findings underscore the importance of providing targeted, accessible and interactive educational resources to upskill GPs to confidently deprescribe BZRAs.

The last set of TDF domains commonly linked to the included strategies were those reflecting beliefs, attitude and willingness to deprescribe BZRAs. Whilst many strategies addressing these domains involved education, they were often underpinned by an interaction with a peer or expert. Specifically, strategies linked with shared care and collaboration, such as case conferencing, outreach support and independent clinical consultation, were perceived by GPs as important supports for BZRA deprescribing. Similarly, a recent international survey of PCPs highlighted that the availability of multidisciplinary collaboration and effective communication were key factors underpinning deprescribing decision‐making, as was access to deprescribing guidelines and tools [97]. Such results stress the availability of clinical support such as multidisciplinary collaboration for GPs to enact effective change to BZRA prescribing and add to the growing evidence of the important role of other clinicians (e.g., pharmacists) in deprescribing BZRAs [98, 99].

Despite clear consensus for most implementation strategies (70.5%), the Delphi process revealed variability in participants' perspectives on around a third of strategies presented. This was reflected by the number of items (n = 16) excluded based on stability, indicating that GPs disagreed and did not always change their opinion based on their peers' perspectives. Participants provided detailed feedback highlighting the complex and diverse views of GPs. The absence of consensus, therefore, suggests underlying heterogeneity in GP practice environments, deprescribing attitudes and perceived usefulness of strategies rather than a methodological limitation, and the persistence of disagreement highlights the contextual complexity of implementing BZRA deprescribing in primary care. Context‐specific approaches to implementation are therefore required to respond to the diversity of practice environments and patient populations, to address the specific barriers of each setting [100]. Providing GPs with a suite of strategies that they can select and adapt to their specific needs may be the most practical path forward, and accounts for individual preferences and clinical setting. This approach aligns with core implementation science principles, which emphasise the importance of selecting and tailoring strategies to fit the specific context in which they are applied [101].

Interestingly, implementation strategies targeting patients directly, such as sending letters to attend review appointments or displaying waiting room posters, were not seen as useful by the GPs in this study, primarily because they were not perceived to be person‐centred. This contrasts with existing evidence, which indicates that GP‐signed letters to patients can be effective for prompting deprescribing conversations [102]. However, this meta‐analysis was based on five studies from 1994 to 2004 and may not reflect contemporary, person‐centred and technology‐enabled care. The person‐centredness of our GP participants may also be why several technology‐driven solutions (e.g., software prompts, email reminders) were seen as less valuable than those that were more responsive to GP and patient needs (e.g., online training modules, tapering calculators, digital self‐help programs). This illustrates the value of GP expert feedback in determining whether implementation strategies reported in the literature remain applicable in current clinical practice.

6. Implications

The findings of this study have several key implications for healthcare policy and practice. GPs indicated that current funding models hinder them from embarking on the deprescribing journey with their patients. In Australia, the fee‐for‐service (FFS) model (i.e., Medicare) provides better remuneration for shorter consultations, which may disincentivise GPs from the lengthy conversations that deprescribing can require [103, 104]. FFS models are known to create a tension between prioritising the volume of consultations with quality of care and exists in many countries [105, 106]. Whilst some countries (e.g., United Kingdom, Netherlands, New Zealand) have moved toward blended funding models that include capitation (i.e., payment per patient not per service) and pay‐for‐performance (i.e., for meeting quality indicators) [107], the pay‐per‐performance funding available in Australia is inadequate and not often used by GPs [108]. This means that several of the first‐priority strategies identified in this paper (especially extended consultation times) have limited feasibility in Australian primary care without policy change.

Enhanced training and access to evidence‐based clinical tools are needed to initiate and sustain deprescribing efforts; however, education alone is likely insufficient to change GP behaviour [109]. So, whilst essential, to be effective, education needs to occur alongside system‐level supports [110]. For example, multidisciplinary support, through mechanisms like case consultations, will enhance confidence and informed deprescribing decisions so that GPs enact newly acquired knowledge. In this study, GPs agreed that the most important way to promote clinical behaviour change is by having systems and structures that reinforce BZRA deprescribing as a priority. Not just to adequately remunerate the clinical intervention but to also provide procedural guidance at the clinic‐level, allowing GPs to deliver consistent care and communication to their patients, with the support of their peers. At the individual‐level, capability building through the provision of practical resources, education and coordinated clinical support are key.

7. Strengths and Limitations

This study has several notable strengths. The modified Delphi method provided a systematic and iterative approach to consensus‐building with input from a heterogeneous sample of GPs. The study also benefited from high levels of GP engagement, with participants providing thoughtful and detailed feedback throughout the process that enriched the findings. Likewise, the inclusion of a panel of GP advisors improved study rigour, streamlined the participant's overall experience and enhanced the potential for translation into actionable, real‐world recommendations. However, some limitations must be acknowledged. The a priori decision to conclude the Delphi process after three rounds, although made to minimise participant burden, may have limited the opportunity to reach consensus or confirm stability on two strategies. Additionally, it is possible that not all implementation strategies identified as useful in the Australian sample of GPs will be translated directly to other countries, health care settings or different healthcare providers (e.g., pharmacists) so this study requires replication.

8. Future Directions

Findings underscore the critical importance of co‐design with PCPs to ensure research recommendations are contextually relevant to support adoption. Future research should also involve people prescribed BZRAs in the design of deprescribing strategies to ensure they are underpinned by the principles of person‐centred care, prior to evaluating feasibility through implementation trials. Assessing the perceived effectiveness of the identified implementation strategies among other primary care health professionals (e.g., pharmacists, nurses) is also warranted to support multiple pathways for deprescribing and patient care. Critically, future implementation efforts must be multipronged to address both systemic and prescriber‐level needs concurrently to support behaviour change and overcome persistent barriers to BZRA deprescribing.

9. Conclusion

Present findings suggest that BZRA deprescribing in primary care may be more constrained by system‐level barriers than GP knowledge or willingness. Consequently, strategies focused solely on GPs will likely be insufficient in changing practice without enabling structures such as appropriate remuneration, clinic‐level workflows, and access to peer and multidisciplinary support. Scalable implementation will require co‐designed, multilevel initiatives aligned with policy and practice infrastructure.

Author Contributions

Conceptualisation: EO, PS, AC; methodology: EO, PS; data curation: MP; data analysis: EO, MP, KG; writing – original draft: EO, MP; writing – review and editing: PS, EO, MP, AC, KG, MH, ST, HW, CA, MT, NZ, AM; visualisation: MP; funding acquisition: EO, PS; project administration: PS, EO, MP, KG; supervision: PS.

Funding

This work was supported by the Department of Health, State Government of Victoria, ADRIA Grant #101.

Ethics Statement

Ethics for this project was granted by the Deakin University Human Research and Ethics Committee: HEAG‐H 202_2023. All participants were required to read a Plain Language Statement and provide their informed consent before participating in this study, including their preference to be listed in acknowledgments.

Conflicts of Interest

EO works at Reconnexion, a state‐funded service in Victoria, Australia, providing specialist treatment for benzodiazepine withdrawal. This study was undertaken as part of the Benzo PACED project, with funding from the Victorian Government via the Alcohol and Drug Research Innovation Agenda (ADRIA) provided by the Victorian Alcohol and Drug Association (VAADA) (Grant ID: #101). MP and KG report that financial support was provided by this grant. This funding source was not involved in research design, data collection, analysis or interpretation, writing of the manuscript or decision to submit for publication.

Data Availability Statement

The data that supports the findings of this study are available in the supplementary material of this article.