The Let's Talk toolkit: developing a theory-informed complex intervention to improve nurse-patient therapeutic engagement on an acute mental health ward by employing Experience-based Co-design and the Behaviour Change Wheel

Lack of high-quality nurse-patient therapeutic engagement is a longstanding problem on acute mental health wards, with a dearth of interventions to address this. A collaborative, theory-driven approach to developing and implementing complex interventions is more likely to be effective and sustainable. This paper describes an integrated co-design-behaviour change approach which developed the Let’s Talk intervention toolkit for improving the quality of therapeutic engagement on acute mental health wards. We describe a theory-driven approach to co-designing an intervention by adapting and integrating Experience-based Co-design (EBCD) with the Behaviour Change Wheel (BCW). Fieldwork was informed by the results of a systematic integrative review and was guided by our co-design-behaviour change approach. It was conducted over 12-months at an acute mental health organisation in England. We undertook semi-structured interviews with 14 service users (seven of which were lmed), two carers and 12 clinicians, 80 hours of non-participant observations and a facilitated intervention co-design process that involved two feedback workshops, one joint co-design workshop and seven small co-design team meetings. Data analysis comprised the identication of touchpoints and use of the BCW and behaviour change technique taxonomy to inform intervention development. theory-driven co-design-behaviour both participatory methods and behaviour change enhances a robust behavioural change theory to guide the development of complex intervention. participatory enhances BCW out a practical guide on how to meaningfully involve service users and other stakeholders when designing complex implementation interventions.


Contributions To The Literature
There is ongoing debate as how to better engage service users, carers and clinicians in intervention design, with recognition that many challenges in improving healthcare services are behavioural in nature To our knowledge, our study is the rst to combine and implement a new theory-driven co-designbehaviour change process with service users, carers and clinicians in a mental health setting We offer tools that others may use to implement this change process in their settings We describe the behavioural mechanisms behind a complex behaviour change intervention to improve the amount and quality of nurse-patient therapeutic engagement on acute mental health wards Background Nurse-patient therapeutic engagement can broadly be described as the use of verbal and non-verbal interchange to improve a service users' mental health (1,2). Lack of high-quality engagement on acute mental health wards is strongly associated with increased rates of self-harm, violence, aggression, absconding and poor perceptions of inpatient care (3,4). Engagement may initiate and enhance the therapeutic relationship (5), which arguably has the greatest impact on treatment outcomes, over and above the speci c interventions provided (6,7). However, nurses report high levels of acuity, reduced workforce, competing administrative duties and the nebulous nature of engagement as reasons for not engaging with service users (8,9,10). These factors also have a negative impact on nurses' job satisfaction (11), increasing the likelihood of burnout and leaving the profession prematurely.
Policymakers, researchers and patient advocacy groups globally have emphasised the importance of engagement in practice (12,13,14). However, lack of quality engagement is a longstanding, complex problem (15,16) and few nursing interventions to improve engagement are reported in the literature.
One such intervention, predominantly implemented in the United Kingdom (UK) is Protected Engagement Time (PET). During PET, nurses devote a speci ed amount of time to regular engagement sessions with service users (17,18). PET originates from the Refocusing Model, which was a comprehensive series of interventions to improve inpatient services and reduce work strain on staff (19,20). The Refocus Model brought about improvements to the quality of care, staff sickness and costs, rates of absconding and self-harm (19). Following this, PET was adopted as a standalone intervention by mainstream policy (e.g. 21), which resulted in its top-down implementation in many mental health services across England. Subsequent evaluations on both adult and older adult mental health wards found that whilst PET attempts to address nurses' opportunities to engage, it does not account for wider considerations about what is done within the engagement sessions (22,23,24). This may be because PET was intended to be used alongside other interventions, and its use as a standalone intervention stemmed from an atheoretical, common sense approach to implementation.
In response to PET's limitations, a Swedish study developed the Time to Talk (TT) intervention (25). TT is a form of PET, theoretically informed by two studies of everyday life on acute wards (26,27) and the Tidal Model -a holistic model of nursing care that promotes the exploration of service users' own narratives (28). In a qualitative evaluation of TT (29) service users reported that clinicians were more engaged after TT was implemented; however, their quantitative evaluation found no improvement in the quality of engagement as measured through the Caring Professional Scale (30). This mirrors evaluations of PET (22,23). Although PET and TT address nurses' opportunities to engage, they may not compensate for wider de ciencies in service provision such as poor supervision, clinical skills, and personal motivations (22,24), and neither were collaboratively developed with input from service users, carers and clinicians.
To better understand and enhance nurse-patient engagement in practice we previously conducted a systematic integrative review to develop a conceptual model of engagement (31). For high quality engagement to occur, the model suggests that nurses must employ techniques that encompass ve "Principles of Engagement": 1) understand the person and their illness; 2) facilitate growth; 3) therapeutic use of self; 4) choose the right approach and, 5) emotional versus restrictive containment. The model drew upon behaviour change theory (32) to show that engagement is broadly in uenced by both the service users' and nurses' capability, opportunity and motivation to engage. To address the limitations of previous interventions, we propose a collaborative, theory-driven approach to co-designing a complex intervention to improve the amount and quality of engagement on acute mental health wards. To do so, we have drawn from our model of engagement described above and adapted and integrated two existing approaches: Experience-based Co-design (EBCD) (33,34) and the Behaviour Change Wheel (BCW) (32).
EBCD is a form of participatory action research which draws on user-centred design and user experience to improve healthcare services (33). The structured EBCD process, detailed in a freely available online toolkit (35), aims to meaningfully engage service users, carers and clinicians throughout a co-design process using observations, interviews and facilitated workshops. The Behaviour Change Wheel (BCW) and accompanying Behaviour Change Technique Taxonomy (BCTTv1) has amalgamated 19 behaviour change theories to create a framework that guides intervention development (32). It follows three phases: 1) understand the behaviour; 2) identify intervention options and 3) identify intervention content. At its core, the model suggests that capability, opportunity and motivation interact to create behaviours (COM-B) (36). The Theoretical Domains Framework (TDF) (37) is aligned in the model to the COM-B components and both are linked to nine intervention functions. The BCTTv1 is a taxonomy of 93 behaviour change techniques (BCTs). These 93 BCTs can be matched to the intervention functions to identify suitable behaviour change techniques (BCTs), which make up the active ingredients of an intervention (38,39). Figure 1 maps the BCW phases, methods and tools to the phases, methods and tools of EBCD and provides a theory-driven basis for the co-design of behaviour change interventions.
Healthcare research and policy now recognise the importance of both co-designing interventions and using a robust theory to guide intervention development (40,41), but to date very few studies report on how to co-design complex healthcare interventions using a theory-driven approach. Currently there are no published studies that develop interventions using EBCD informed by the BCW. In response, we demonstrate the implementation of a theory-driven co-design-behaviour change process ( Fig. 1) that was used to develop a complex intervention toolkit for promoting nurse-patient engagement on acute mental health wards. We aim to: Describe a collaborative, theoretically driven approach to co-designing complex interventions; Demonstrate the implementation of this process to share learning with others; Present a toolkit to enhance therapeutic engagement on acute mental health wards.

Design
The study was guided by the UK Medical Research Council (MRC) complex intervention framework (42) and was theoretically-driven by the content illustrated in Figure 1. The co-design process is reported in accordance with guidance for the reporting of intervention development studies in health research (GUIDED) (43) (Additional le 1).

Setting
The study was conducted with service users, carers and clinicians from one XX (blinded for peer review) National Health Service (NHS) Foundation Trust in England. The intervention ward has 18 beds and treats adults (18-65) experiencing an acute phase of severe mental illness. Service users are predominantly detained under the Mental Health Act (44) and the nursing team works shift patterns from 0730-2130 or 2100-0800. This project began in April 2018 and complemented other organisational improvement work to re-implement PET.

Participants
The co-design team was recruited through: A convenience sample of service users and carers via: 1) face-to-face contact and posters at community mental health teams (CMHTs) and 2) face-to-face contact and email at service user advocacy groups connected to the participating NHS organisation; A whole population sample of clinicians on the participating ward were invited to take part via presentations, posters, email, and face-to-face meetings.
XX screened all potential participants, speci cally looking for those who had, or had cared for somebody who had at least one inpatient admission at the organisation but was not currently experiencing a relapse. Eligible individuals were then guided through a written informed consent procedure. Figure 2 shows the recruitment process by type of participant and workshop attendance through the EBCD process. A total of 35 members were recruited to the co-design team including 15 service users, two carers, 10 registered mental health nurses (RMNs), four health care assistants (HCAs), three psychological therapies clinicians and one student nurse. Just over half of the co-design team were female (54%) and just under half were from a Black, Asian and minority ethnic background (49%). Participants' ages ranged from 18-64 years.
Data collection methods and processes were aligned to the EBCD phases and BCW stages contained in Figure 1 and informed by the aforementioned integrative review (31) these included non-participant observations and semi-structured interviews to gather service user, carer and clinician experiences, and feedback and co-design workshops to facilitate development of the engagement toolkit.
Non-participant observations and semi-structured interviews XX conducted 80 hours of non-participant observations on the acute ward between the hours of 0730-1500 or 1330-2130, Monday through Sunday. Observations were performed in 15-minute intervals, beginning with the rst nurse encountered and continued until all nursing staff had been observed. Fieldnotes were guided by Tyson and colleagues (45) and documented patterns of nurse-patient behaviour, nurse-patient dynamics, tone of voice, body language, potential in uences on engagement and general ward atmosphere.
XX also interviewed 14 service users, two carers and 12 clinicians on a one-to-one basis at a location of their choice including university premises, o ces at CMHTs or by telephone. All interviews were audio recorded and seven service user interviews were lmed in keeping with the EBCD approach. Interviews lasted between 30-80 minutes. A topic guide was followed, informed by our review (31), the nonparticipant observations and the COM-B/TDF domains (37,38). Interviews addressed participants' experiences of engagement, barriers and facilitators to engagement, and clari ed assumptions made from the observations. Full details of the non-participant observations and semi-structured interviews, including the inductive analysis of data to identify 'touchpoints' (emotionally signi cant points) of importance to the co-design team, are reported in a separate paper (in preparation). A secondary deductive analysis of interview data, which is reported in this paper, was also undertaken to identify barriers to engagement. Deductive codes were based on the COM-B and TDF components of the BCW which were used as an a priori framework to analyse and thematically organise interview data. XX independently coded and themed the data using this framework. Extracts from both the lmed and audio-recorded interviews were also edited into a trigger lm that was used to stimulate discussion at the feedback and co-design workshops.
Feedback and co-design workshops Touchpoints and themes were shared at separate service user/carer and clinician feedback workshops and at a joint co-design team workshop. This ensured validity of the analysis, facilitated the joint selection of target behaviours based on the touchpoints, and allowed intervention options and content to be agreed. Seven co-design team meetings were also established to work on speci c priority areas.
Consensus was reached through facilitated discussions and consensus building exercises including emotional mapping (46) and a nity grouping (47).
Input was also sought throughout the co-design process from two mental health patient and public involvement (PPI) groups based at the participating organisation. An advisory group consisted of a service user representative, one clinician and clinical academic experts in (a) the EBCD methodology and (b) therapeutic engagement, respectively. The service user representative co-facilitated the feedback workshops with XX (a mental health nurse trained in the application of the EBCD and BCW approach) who also facilitated the joint co-design and co-design team workshops with the assistance of another nurse researcher trained in the BCW approach. Three co-design team members wrote re ective accounts of their experiences of the co-design process and are co-authors of this paper.

Results
Here we present our theory-driven approach to co-designing the Let's Talk complex intervention toolkit. Our ndings are organised under the three stages (and eight constituent steps) of the BCW guide, as shown in Figure 1.

Stage 1: Understanding the behaviours
Step 1: De ne the problem in behavioural terms Through previous research (5), our integrative review (31) and initial discussions with our PPI, advisory groups and the clinical service lead, modern matron and divisional medical director at the NHS organisation, the behavioural problem was de ned as the absence of high quality nurse-patient therapeutic engagement on acute mental health wards i.e. not using the Principles of Engagement identi ed in our review.
Step 2: Select target behaviour(s) In keeping with the EBCD methodology, it was important to understand how service users and staff typically experienced engagement prior to the identi cation of relevant areas for behavioural change. Through observations and semi-structured interviews, the research team identi ed 28 touchpoints. Some examples of touchpoints were 1) I was left on my own and ignored; 2) my care was robotic and 3) As a nursing team we need to create better bonds with service users (full results in preparation to be published elsewhere).
To ensure credibility, the touchpoints were discussed during two facilitated feedback workshops -one for service users and one for clinicians. In an emotional mapping exercise, participants were encouraged to identify improvement priorities based on their touchpoints and assign associated behaviours (see Additional le 2 for breakdown of touchpoints into improvement priorities and associated behaviours). Participants then ranked their improvement priorities in a dot voting exercise and chose four priorities to take forward to the joint workshop. The service user and clinician priorities are shown in Table 1. Table 1 -Service user and clinician priorities for change At the joint workshop, facilitated discussion encouraged participants to consider the potential impact, likelihood of change, spill over effect and ease of measurement of all the improvement priorities and associated behaviours. An a nity grouping exercise was conducted and through this, four shared improvement priorities were identi ed and agreed: Improve communication with withdrawn people Nurses to help service users help themselves Increasing nurses' con dence when interacting with service users Improve team relations and ward culture.
Step 3: Specify target behaviour(s) EBCD focuses on identifying participants' improvement priorities as a way of bringing about change that is meaningful to service users and clinicians (35). We used the BCW to examine each of the four joint improvement priorities. At the joint workshop, the co-design team formed into smaller groups with equal numbers of service users and clinicians. Each group completed a written exercise where they examined the joint priorities and associated behaviours in terms of who needs to perform the behaviour, what the person needs to do differently to achieve change and when, where, and with whom they will do it ( Table  2) (See Additional le 3 for example of written exercise).

Table 2 -Speci cation of behaviours for joint improvement priorities
Step 4: Identify what needs to change From our review and semi-structured interviews with service users, carers and clinicians, the research team identi ed 26 barriers to engagement and mapped them to the COM-B/TDF domains. The barriers were discussed with participants at the feedback workshops to ensure credibility. At the joint workshop participants matched the barriers to their four joint improvement priorities. The barriers related to each COM-B component are discussed below, with the corresponding TDF domains presented in parentheses.

Capability
Participants agreed that nurses often had limited knowledge and inadequate training in therapeutic engagement techniques (skills and knowledge): "Although I've been doing this for almost ve years it's like sometimes with certain patients you just don't know what to say…I wish there could be some training to understand that stuff." -RMN6 Nurses also felt that the very nature of having a mental health problem could make it di cult to engage, and while service users agreed that their mental illness and medication effects could negatively impact engagement (memory/attention/decision process), they were able to describe helpful engagement techniques that nurses could employ, even with the most acutely unwell people. This further highlighted the need to improve nurses' engagement skills: "Sometimes you have a lot more patients who are unwell or sometimes they're less unwell, so engagement uctuates week on week from that point of view" -RMN2 Opportunity It was felt that there needed to be a cultural shift on the ward and within the organisation so that nursepatient engagement activities were supported and valued in the same way as other tasks such as hourly observations or administrative duties (social in uences): "It was a numbers game, everyone's taking handover, another one's doing checks, some are on break…in an ideal world allocate friendly HCAs just to sit with patients." -SU7 There was unanimous agreement that lack of resources negatively impacted on nurses' ability to engage therapeutically: "The problem for me lies on the number of staff, that is not enough…" -C1.
This created an untherapeutic ward environment where "professionals would run around like mad rabbits not giving any attention to the patients." -SU2 (environmental contexts and resources).

Motivation
Nurses felt that they could not always trust all members of their team to carry out the job in the right way. This created a feeling of helplessness for some nurses, which deterred them from engaging therapeutically (beliefs about capabilities): "I became very aware that when there is an incident, I'm left on my own…I stopped trusting the team…I couldn't rely, therefore I needed to take a step back from the patients." -HCA2 Service users were also deterred from approaching nurses for engagement because they felt nurses often did not understand their problems or would punish them if they asked for therapeutic engagement too often (beliefs about consequences): "I kept myself to myself because even when I asked for simplest of things I was made to wait for ages so I would get frustrated, but if I showed frustration no doubt that would be on my notes and I would get set back." -SU4 As well as issues of trust, the ward staff felt as though their team were transient, with many longstanding nurses leaving to work elsewhere. This led to a lack of shared responsibility. Therapeutic engagement could easily "fall through the cracks -HCA1" and when poor quality engagement was witnessed, it was rarely followed up by a senior member of the team. This made some nurses feel they could not be bothered to engage: "I mean to put it blunt; I know it sounds really bad…I can't be bothered." -RMN5 There was also a blurring of professional roles, where although nurses knew they should engage, they left it to other professionals such as the occupational therapist or activities coordinator: "I can completely understand why nurses want separate roles because they would say you don't do our job so why should we do yours, but I do take people out on escorts and I do blur the boundaries there." -PT1.
When asked to give examples of nurse-patient engagement, many service users spoke about engagement with professionals other than nurses. This shows both the lack of engagement from nurses and the di culty service users have in delineating between the nursing role and the role of other health professionals (social/professional identity).
There was a general sense from nurses that therapeutic engagement "didn't always help people" -RMN8 (optimism). This led some nurses to feel anxious about engaging therapeutically, particularly when they felt they did not have the required skills. When this was coupled with feelings of frustration at the perceived lack of managerial support, nurses reported feeling drained, burnt out and demotivated (emotions): "One of the biggest problems is the management style which on paper, yes, it seems to be doing everything right, but in practice they have a very poor relationship with their staff and that does impact on performance…I just feel like no one cares about you, so why give up your time?" -RMN3 Stage 2: Identify intervention options Step 5: Identify intervention functions PPI and advisory group meetings highlighted that some of the terminology used to describe intervention functions would not be suitable to use with our participants. Words such as "coercion" can have negative connotations to mental health service users. Instead, practical examples that captured the essence of each intervention function were provided to participants at the joint co-design workshop. In a written exercise they were encouraged to use these examples to think about intervention functions that could address their four joint improvement priorities. Where possible we modelled these examples on illustrations from interviews with service users and staff. Where this was not possible, we developed examples from the BCW book (32) ( Table 3). Table 3 -Practical examples of behaviour change wheel functions given to co-design team Participants identi ed ve intervention functions that were relevant to bringing about the desired change. These were 1) training; 2) education; 3) enablement; 4) coercion and 5) persuasion. Through discussions with senior management, the research team also identi ed restriction as a relevant function. The links between the COM-B/TDF domains and the intervention functions are shown in Table 4.
Step 6: Identify policy categories The BCW includes policy categories which may help to support the delivery of an intervention. At the outset, changing policy was not a primary aim of this study; the research team did not engage in this step with participants. However, through discussion with senior management the research team identi ed communication/marketing, guidelines and social planning as potentially relevant to facilitating our intervention on other wards in the future.

Stage 3: Identifying intervention content and implementation options
Step 7 & 8: Identify behaviour change techniques and mode of delivery Rather than provide participants with a long list of BCTs, the written exercise at the joint workshop encouraged them to design intervention strategies they thought relevant to each of the four priorities and its in uencing factors. The research team retrospectively assigned BCTs to the participants' examples and selected further BCTs and intervention strategies not identi ed during the joint workshop. These were the basis for the development of the rst intervention prototype.
The prototype was further re ned through an iterative process of email exchanges, telephone calls, a PPI meeting, seven small co-design team meetings and nally presentation of the work at an organisation wide acute care forum. As per the BCW guide (32) the APEASE criteria (affordability, practicability, effectiveness/cost effectiveness, acceptability, side effects/safety and equity) were used in an adapted form (see Additional le 4) to stimulate discussion and ideas. These criteria ultimately informed the choice of intervention strategies for each improvement priority.
Fourteen BCTs were considered relevant to the Let's Talk intervention toolkit. Table 4 shows the link between each phase of the behaviour change intervention design process, the 14 BCTs and the intervention strategies and modes of delivery which resulted from the co-design process.
The Let's Talk toolkit consisted of four main components, linked to the co-design team's four joint improvement priorities: A 30-minute training lm for nurses, delivered by service users and carers to be shown to nurses at the start of the intervention. Service users and carers discuss good and bad engagement techniques and personal accounts of their experiences of engagement whilst an inpatient, structured by our model of engagement.
An illustrated workbook called My Conversation Companion which includes guided exercises that nurses and service users can do together to help structure therapeutic conversations.
Signs attached to the outside of service users' bedroom doors to enable them to indicate, with a sliding panel, whether they would like engagement time or not. The signs are linked to the hourly nursing observation record, where each hour nurses will be required to record if a service user has requested engagement and if that request has been ful lled. "Missed engagement" will be handed over at each nursing shift with the expectation that it is ful lled that day. Observation records will be audited each month and feedback given to the nursing team. Additionally, an illustrated sign on the inside of service users' doors will encourage service users to use the signs if they want to engage.
Changes to nurses' daily routines, for example during handover, time will be made to check-in with the nursing team and offer additional support to any team member that needs it that day. Additionally, quarterly facilitated workshops will bring clinicians and service users together to discuss, re ect and improve practice.
Through discussions with the chief nurse, assistant director of nursing and divisional medical director and presentation of the work at an acute care forum it was agreed that the Let's Talk intervention would support the relaunched implementation of PET within the organisation. Discussion with participants revealed that they supported this and considered some form of PET essential to support nurses to use Let's Talk in practice. See Additional le 5a and 5b for the toolkit.

Discussion
The delivery of high-quality nurse-patient therapeutic engagement is a complex issue that requires input from service users, carers, clinicians and researchers alike. Interventions to improve engagement must be multifaceted and encompass service users', carers' and clinicians' capabilities, opportunities and motivations to engage. We used the methodical and evidence-based framework of the BCW to guide intervention development within a co-design process. This enhanced the process by supporting its "intrinsically desirable qualities" (48) with a robust theoretical underpinning that facilitated a full analysis of existing barriers and behaviours among its principle stakeholders. Although Larkin and colleagues (49) suggest that it may be unrealistic to expect co-design participants to generate solutions to longstanding problems within a short space of time, supporting participants' ideas with a systematic and methodical theory of behaviour change may help mitigate that limitation.
Recent literature encourages a systematic, comprehensive and transparent approach to intervention development (44). However, many behaviour change interventions are poorly de ned and do not use consistent language to describe their mechanisms of action (50,51) making it di cult to pinpoint what did and did not work, which also reduces the ability to compare such interventions (37). The BCW enabled us to understand and describe the mechanisms of action behind Let's Talk, which is likely to both improve its effectiveness (52) and enable us to review and re ne intervention targets after preliminary testing.
Although systematic, the BCW approach may be considered somewhat prescriptive. This can clash with the underlying principles of co-production and co-design, which demand democratic, innovative and creative techniques (53,54). The concept of co-production in mental health was not commonplace even ve years ago (55). Traditionally, professional knowledge had a higher status than service users' lived experiential knowledge (56,57). Despite some notable exceptions (e.g.58,59), service user participation in research was, and often still is tokenistic, with participants having little in uence over de ning the problems or required changes (57,60). It was essential that our process acknowledged, explored and addressed these power differentials so as not to reinforce these entrenched ideals.
Academic language and terminology can preserve power differentials and compromise user and clinician participation (61, 62). People who suffer from mental health problems experience effects that can negatively impact cognition and concentration, often exacerbated by medications (63). The use of overly technical language may disproportionally affect people from this group and may lead to exclusion and disempowerment (64), which mirror some of the alienating experiences faced whilst an inpatient on acute wards (e.g.65). While the COM-B model uses relatively simple terminology (37), the language used to describe the intervention functions was particularly problematic. Intervention functions such as "coercion" and "restriction" may have triggered di cult emotions for some of our participants. These words describe negative ward experiences for example when clinicians coerce service users into taking medication (66), or when liberties are restricted due to treatment under the Mental Health Act 1983 (44).
This was also true of the clinicians who participated in our study. Suggesting that they lacked "skills" or "knowledge" was likely to alienate them from the process and make them feel devalued.
To ensure delity to the underlying principles of co-design we therefore tailored the BCW approach to the needs of the co-design team. The research team found that providing practical examples of each intervention function, using language from the service users', carers' and clinicians' interviews, was a suitable way of adhering to the principles of co-design and using evidence-based theory in a nonalienating, con rmatory way. Although APEASE criteria were not considered to contain triggering terminology, some of the language was overly technical which also risked alienating co-design team members. The research team therefore translated the APEASE criteria into more accessible language. Furthermore, the co-design team were encouraged to design their own intervention content based on the behavioural analysis. The research team retrospectively assigned BCTs and con rmed these with the codesign team. This adhered to the underlying principles of co-design by foregrounding service user experience (rather than privileging academic knowledge over experiential knowledge), whilst also creating an intervention that could be clearly and methodically described through evidence-based theory and language.
Re ective accounts from three of our co-design team support the steps taken by the research team to ensure an inclusive, participatory process. Whilst the potential for experiential re ections to trigger di cult emotions was anticipated, team members' expressed anxieties were soon 'quashed' by a 'safe and secure' environment in which members 'never felt pressured or judged'. This allowed the service users, carers and clinicians 'to support each other on an equal basis and share a common goal'. The opportunity to share personal experiences emerged as an important dynamic across the three re ective accounts. It was variously described as 'a privilege', and an 'incredibly moving' and 'powerful' experience that allowed their expert knowledge to be used 'to implement new models of care and improve quality standards' that 'would make a real difference'. Consequently, these co-design team members described an 'enjoyable' and 'rewarding' process that engendered feelings of pride and empowerment. One member referred to it as a 'life changing' event that promoted self-esteem and self-awareness, and another reported the development of re ective skills. Notable also was the wider outreach and consultation that members undertook through liaison with professional colleagues, services users and carers in various institutional and community arenas, which mirrored their experience of the co-design process. This allowed stakeholders to express any concerns, ask questions and provide feedback. In turn, this led to 'product re nements to make the workbook more accessible and easier to read'. As well as personal impact, these team members described inter-personal bene ts including 'feeling (more) engaged with mental health professionals' and managing to 'engage in some really good work' with patients. They were also optimistic about the likely impact of this work moving forward. They sensed that service users and clinicians were 'inspired' by their work and believed the workbook would have 'a ripple effect…and help create a cultural change within the organisation'. Each of the re ective accounts is provided in full in Additional le 6.
Beresford (67) argues that frontline clinicians can also be a marginalised group whose voices are often excluded. It is also vital to consider the needs of the service provider whilst embarking on participatory work (68). We implemented several facilitative measures such as providing back ll money so clinicians could attend the feedback and joint workshops and offered exibility with participation in the small codesign team work e.g. emailing instead of face-to-face meetings and piggybacking staff meetings. However, enabling clinicians to participate equally was challenging. Unfortunately, organisational structures such as shift patterns and sta ng levels impacted on clinicians' ability to fully participate.
Regular staff meetings or re ective practice groups were also not in place. When given the opportunity to participate, clinicians were motivated, and meaningful participation was possible during the feedback and joint workshops. However, without organisational support structures to provide clinicians time to undertake the ongoing co-design work, much of the prototyping and iterative development of the intervention components were undertaken by the service users. There is a need for healthcare organisations to recon gure their services so clinicians can meaningfully participate in such endeavours and encourage a sense of joint ownership over the work.
Although the process was highly collaborative and involved service users, carers and clinicians to varying degrees, it was conducted at just one NHS site, which represents a possible limitation. Transferability of our processes to other settings cannot be guaranteed. However, to our knowledge, this is the rst time the BCW has been translated for use with participants who have mental health problems and used within an integrated co-design-behaviour change process. This new and novel approach will require further testing to ascertain whether it is suitable and translatable to other intervention development processes.

Conclusions
This paper has described the implementation of a new theory-driven co-design-behaviour change approach used to co-design the Let's Talk intervention toolkit. It offers tools that others may use, or adapt as necessary, to implement the approach in their settings. It also describes the behavioural mechanisms behind the Let's Talk intervention toolkit to improve the amount and quality of nurse-patient therapeutic engagement on acute mental health wards. Our paper makes a timely and novel contribution to further both participatory methods and behaviour change theory. The approach enhances EBCD by introducing a robust behavioural change theory to help guide the development of a complex intervention. In turn, our participatory approach also enhances the BCW by setting out a practical guide on how to meaningfully involve service users and other stakeholders when designing complex implementation interventions. Availability of supporting data All data generated or analysed during this study are included in this published article and its accompanying additional information les.

Competing interests
All authors declare they have no competing interests. Tables Table 1 -Service user and clinician priorities for change Service user priorities Clinician priorities 1) Nurse-patient communication needs to be improved 1) Improve the way we communicate with service users 2) Treat me like a human being 2) Improve the way that leave is communicated 3) Forgive and forget 3) Improve culture around response 4) Help me help myself 4) Improve the way messages are handed over within the team  Have a poster on the ward that shows people happily engaging, with a message that reminds clinicians that engagement is part of their job, it is not "slacking off" * Incentivisation (Create an expectation of reward) Offer a prize for the ward that has the best patient feedback regarding interactions ^ Coercion (Create an expectation of punishment or cost) At discharge, ask service users to provide feedback to the ward about the quality of interactions provided and hold staff accountable for this * Training (Imparting skills) Training program that enables nurses to role play with service users, so they gain skills on how to deal with service users' problems * Restriction (Using rules to reduce/increase the opportunity to engage in target behaviour) Nurses stop paperwork/admin during mealtimes and sit with service users and have a cup of tea or some food * Environmental restructuring (Changing physical or social context) Give service users cards that display different emotions and if they want to talk they can put the card on their door so nurses know to approach them * Modelling (Providing an example for people to aspire to or imitate) Have a therapeutic engagement champion who promotes engagement and helps nurses to carry out group activities with patients * Enablement (Increasing means or reducing barriers to increase capability beyond environmental restructuring) Have a ward diary for interactions that a member of staff is responsible for each shift * Key: * = example that came from participant interviews; ^ = example developed from BCW guide Social support: Checkins at handover, re ective practice workshops & Compassion Champion -as described above