The CORE study—An adapted mental health experience codesign intervention to improve psychosocial recovery for people with severe mental illness: A stepped wedge cluster randomized‐controlled trial

Abstract Background Mental health policies outline the need for codesign of services and quality improvement in partnership with service users and staff (and sometimes carers), and yet, evidence of systematic implementation and the impacts on healthcare outcomes is limited. Objective The aim of this study was to test whether an adapted mental health experience codesign intervention to improve recovery‐orientation of services led to greater psychosocial recovery outcomes for service users. Design A stepped wedge cluster randomized‐controlled trial was conducted. Setting and Participants Four Mental Health Community Support Services providers, 287 people living with severe mental illnesses, 61 carers and 120 staff were recruited across Victoria, Australia. Main Outcome Measures The 24‐item Revised Recovery Assessment Scale (RAS‐R) measured individual psychosocial recovery. Results A total of 841 observations were completed with 287 service users. The intention‐to‐treat analysis found RAS‐R scores to be similar between the intervention (mean = 84.7, SD= 15.6) and control (mean = 86.5, SD= 15.3) phases; the adjusted estimated difference in the mean RAS‐R score was −1.70 (95% confidence interval: −3.81 to 0.40; p = .11). Discussion This first trial of an adapted mental health experience codesign intervention for psychosocial recovery outcomes found no difference between the intervention and control arms. Conclusions More attention to the conditions that are required for eight essential mechanisms of change to support codesign processes and implementation is needed. Patient and Public Involvement The State consumer (Victorian Mental Illness Awareness Council) and carer peak bodies (Tandem representing mental health carers) codeveloped the intervention. The adapted intervention was facilitated by coinvestigators with lived‐experiences who were coauthors for the trial and process evaluation protocols, the engagement model and explanatory model of change for the trial.


| INTRODUCTION
Internationally, mental healthcare policies are replete with references to embed coproduction and codesign with service users in the design, planning and delivery of programmes. [1][2][3][4][5][6][7][8][9] The 2017 United Nations Special Report identified coproduction as fundamental to mental health service participation to reach the highest attainment of physical and mental health. 10 This is coupled with consensus for recovery-oriented mental health services that consistently facilitate psychosocial recovery as a subjective, ongoing process that encompasses spiritual, social, psychological and cultural dimensions for individuals. [11][12][13] Engaging mental health service users is central to the enactment of recovery-oriented systems and to ensure that participation in service design, planning and delivery holds personal and social meaning for individuals. Despite a growing evidence base that supports an association between engagement leading to improved patient experience, clinical effectiveness and patient safety, 14 a more recent review of service user participation in mental healthcare planning and programmes to improve experience and service effectiveness found that exclusion continued to be the norm rather than the exception. 15 There is also cautiousness emerging about engagement as the next big blockbuster drug for healthcare and the driver for improved health outcomes, quality and safety and reduced healthcare costs. 16,17 In the attempts to implement systematic approaches to engage people who access mental healthcare services and to ensure that the engagement methods do foster shared power and decision-making, interest has grown exponentially in participatory methods such as codesign and coproduction. In the last decade, a rapid evolution of studies labelled as codesign, coproduction, coinnovation and cocreation has occurred in healthcare quality improvement. This evolution has contributed to what has been called a Participatory Zeitgeist, where participation using codesign and coproduction has become the spirit of our contemporary times, but not without conceptual and definitional challenges and a need for robust evaluation. 18 For example, the extent to which engagement using codesign leads to recovery-oriented service delivery, individual empowerment or improved health outcomes in mental health services is yet to be determined.
One quality improvement, participatory method to engage service users, carers and staff in service design where experience is central is experience-based codesign (EBCD). EBCD aims to improve experiences of services by working in partnership with staff, service users and carers on areas for change. 19,20 EBCD has been implemented to improve service experiences and outcomes in head, neck, breast and lung cancer services, gynaecology and colorectal settings, stroke and rehabilitation, emergency, to end of life and intensive care units and to a much lesser extent in mental health. 21 While organisational improvements such as operational efficiencies, interpersonal dynamics of care, increased communication, team relationships, patients feeling listened to and reduced complaints have been documented, 22,23 no studies have examined the impact of codesigned improvements on systems or service levels or on individual health outcomes. No randomized-controlled trial (RCT) study designs have been used to date to test this.
To address this gap and to identify the benefits or otherwise of an adapted EBCD method for recovery-oriented mental health services and improved psychosocial recovery of service users, the CORE study tested the effectiveness of mental health experience codesign (MH ECO). A stepped wedged cluster randomized-controlled trial (SW-CRT) was conducted in nonclinical (psychosocial recoveryoriented) Mental Health Community Support Services (MHCSS) in Australia. CONSORT guidance for reporting SW-CRT was followed. 24 Our primary participants were people living with severe mental illness (SMI was defined as including psychosis, schizophrenia, bipolar disorder, major depression and other disorders such as personality and eating disorders), with carers who had a family member engaged in services and staff. 25 The secondary outcomes were improved quality of life for people living with SMI and carers, and changes to recovery attitudes from staff and the recovery-orientation of services. This paper reports on the trial outcomes for people with SMI only; carer and staff outcomes are reported separately. 26 PALMER ET AL. before the final follow-up period of data collection was completed. 28 The intervention, MH ECO (explained below), was directed at the service level for improvements, so a cluster design was determined to be the most appropriate. Four large mental health organisations (two nongovernment providers and two community health centres) were partners in the trial and the intervention was delivered to nine teams across these four organisations in metropolitan, outer metropolitan and regional locations. Teams were randomly allocated (three at a time) to different start dates 9 months apart (see Table 1).

| Settings
Nine teams delivering psychosocial recovery programmes as part of commissioned MHCSS were recruited across four large service providers in Victoria, Australia. Organisations had delivered services to approximately 14,000 people in any given year at the time of recruitment in 2013. Support included daily living skills development, recovery planning and facilitation of social and community participation to people living with SMI in community settings. At the time of trial commencement, an outreach model of individual support was implemented shifting away from on-site, group models of service.
The goal of MHCSS programmes is to support psychosocial recovery and deliver recovery-oriented mental healthcare.
The most available data on MHCSS service recipients showed people lived with between one and four complex factors, which included social isolation, activities of daily living, issues related to unresolved trauma, treatment-resistant symptoms, extensive time to maintain levels of functionality with little improvement in functionality over time, chronic physical health problems, difficulty maintaining medications, problems with intellectual disability/cognition, alcohol use and drug use. MHCSS staff were not responsible for clinical assessments, though they engaged with clinical care providers for updates and information sharing.

| Service recipients
Eligibility for people living with SMI to be recruited to the study followed service eligibility criteria set by the government funder.
MHCSS service users were characterized as having enduring psychosocial disabilities and long-term impairments from mental illnesses that range from diagnostic names such as bipolar disorder, schizophrenia, psychosis, chronic depression and anxiety to obsessive compulsive disorders and other personality disorders. Inclusion criteria to services were as follows: aged between 16 and 75 years and a psychiatric condition (bipolar disorder, schizophrenia, psychosis, major depression, severe anxiety, personality disorder, posttraumatic stress) that results in persistent impairment and substantial reduction in psychosocial functioning for communication, social interaction, learning, selfcare and self-management affecting social and economic participation. Carers were eligible for recruitment to the study if they were a family member, friend or in a caring relationship with the person living with SMI; carers did not need to be matched. People with SMI and carers were not recruited if they could not understand spoken English and were unable to complete the two-stage consent process (outlined in the published protocol) or were not in receipt of services from a participating team. The rationale for exclusion of people who could not understand spoken English was due to the primary and secondary outcome measures not being appropriately translated for the cultural communities in question, interpreter availability and the challenges presented by multiple languages with an interpreter within the face-to-face codesign sessions.

| Intervention
The adapted MH ECO intervention 18,28 is described in the published trial and nested process evaluation protocols. 28   Arm is the allocation of group of clusters/individuals. Three clusters were randomized to each arm. delivered training for codesign preparation and all codesign meetings. Responses to telephone interview questions were analysed by group using the Leximancer software analytics programme. Leximancer organizes prominent concepts discussed in text within themes into a thematic map. 29 Thematic maps were reviewed to F I G U R E 1 Adapted mental health experience codesign intervention F I G U R E 2 Illustration of emotion mapping and synthesis of service stories for presentation to collaboration groups to formulate codesign objectives identify the most commonly shared negative touch points per group and textual responses examined to understand nuanced meanings.
These touch points were then explored in focus groups with service users, carers and staff (held separately) as neutral statements to develop a deeper understanding of people's experiences. Emotion mapping was completed using brighter colour post-it notes to represent strong feelings and pale colours to represent less strong/mild feelings (the written feedback that participants provided could be either positively or negatively framed; the focus was on emotional connection to the touch point).
Emotion mapping and how this was used to identify shared patterns across groups is presented in Supporting Information Appendix 1. Once the shared touch points for improvement areas were determined in a cluster, a summary of the service stories was provided to the Collaboration Group (detailed in Figure 2) as a short report for development of the codesign objective.
All participants received training before codesign meetings (  Participants were asked to complete a structured questionnaire that included the 24-item Revised Recovery Assessment Scale (RAS-R) as the primary outcome measure. 30 The measure for psychosocial recovery for service users was identified in a small pilot week-in-the-life diary and a social network map. In addition to evaluation feedback at the end of every collaboration or codesign meeting, a subsample of participants (service users, carers and staff) was interviewed for the nested process evaluation. These interviews were conducted face-to-face and/or by phone; the analysis of these data will be presented separately in the nested process evaluation for the trial.

| Sample size
Sample size was powered for at least 80% to detect a standardized effect size of 0.35 for psychosocial recovery RAS-R (primary outcome) between the intervention and control phases for a fixed cluster size of 30 people with SMI from nine clusters at each of the four follow-up times (see Table 2 in the full published study protocol). 28 Power was determined in a simulation study that Interim analyses and stopping rules were not required.

| Randomisation
A computer-generated random allocation sequence stratified by the MHCSS organisation was generated by the statistician blin- An intention-to-treat approach analysed all study participants according to the arm that the cluster was assigned to at each time point. 34 People who refused or were unable to complete follow-up questionnaires were asked to complete the primary outcome measure to minimize missing outcome data. Up to five attempts were made over the 2-month data collection period for outcome measures.
This included contacting trusted proxies who were provided by individuals at enrolment. The reasons for why individuals were lost to follow-up were recorded. Under the mixed-effects models used for the analysis, data were implicitly assumed to be missing at random. 34 In a planned secondary analysis, direct effects of the length of time for which the participants were in the intervention phase (namely, 0, 9, 18 and 27 months) at each follow-up time were estimated using a linear mixed-effects model, where the length of time exposed to the intervention was treated as a fixed effect and adjusted for follow-up time.
Estimates were reported as the mean outcome difference when exposed to the intervention for 9, 18 or 27 months compared to '0 months', the time when individuals were in the control phase and not exposed to the intervention. The log-likelihood ratio test was used to assess whether to treat the length of time in the intervention phase as a continuous rather than a categorical variable in the regression model. Treating exposure time to the intervention as continuous assumed that the increase in the intervention effect was linear with the length of time in the intervention phase. In sensitivity analyses, long-term intervention exposure was investigated using the same methods as above, but restricted to participants recruited at baseline only.          Informing and involving carers with updates about services provided to a person they care for Communication with and involvement of carers in care planning and outcomes Support options to carers for when they are unwell themselves Communication about support groups for carers directly Time to process information when first engaging at a service Implemented carer peer support workers within service delivery Provided information to staff on the role of carer workers Designed new brochure and website updated (involved carers in the design of these) Increased activities for carers and options for access to self-care programmes a Not all touch points for improvement were addressed within the codesign stage of the intervention due to trial limitations to focus on one area. mental illness. Service users also described re-design needs for physical spaces in services, and feeling unwelcome because of unfriendly or distant voicemail messages at services. In relation to this, service users wanted to see an increased use of SMS to receive appointment reminders, make changes to their own appointments or to receive information about staff absences.

| RESULTS
These improvement areas can be understood within four categories that were initially identified in the seminal head and neck cancer centre EBCD projects conducted in the United Kingdom. 22 The four categories were quick fixes-improvements that involved little or no change in everyday working practices (e.g., revising information, updating brochures); process redesigns-improvements where new in-service procedures for consent or access to services were developed; cross-service or interdisciplinary redesignimprovements involving process or structural redesign across different services to improve responses to an issue; and organisational change-improvements addressing organisational issues such as delayed receipt of results of a procedure or appointment wait times.
Two further categories were added from this trial: Technological fixes-improvements to technological components of services such as websites, internet access, social media and app provision, phone and voicemails or SMS use, and physical infrastructure-improvements where there were changes to the physical environments of services.
Overall, the most prominent codesigned improvements implemented in the trial were related to quick fixes and process redesigns. New information and welcome packs were cocreated with service users, or information about social groups was gathered and presented on websites. A commonly reported barrier for uptake of the local information, though, was that service users tended to have limited use of emails and did not read the websites or typically use social media. The lessons here might be most relevant to future digital mental health transformation and the implementation of technology-based interventions where purposive methods for engaging people will be needed.
In one or two services, attempts were made for organisational change and physical infrastructure improvements. Organisational changes that were implemented led to a coproduced camp by and for service users in one service, and in another service, a carer peer worker was arranged as a response to specific carer engagement A weakness could be that the codesigned improvements did remain mostly quick fixes and process-oriented. Recent work that has examined the experiences that people might draw from their service encounters is important to consider here as it might point to a need to widen the lens of service users engaged in sharing their stories and to identify further who is taking part in codesign. 36 There may be a need in EBCD to establish the backgrounds of service users and carers for participation in codesign, so that multiple service encounters or engagement with specific therapeutic models are explicitly sampled for more tangible effects to be realized and

| CONCLUSIONS
There have been more than 60 noncontrolled evaluations of EBCD quality improvement projects 37 where patient experience and transformations to healthcare workforce, culture, values and behaviours have been documented to improve. 22 As mental health policy increasingly advocates for coproduction and codesign approaches 37,38 to facilitate better services, improved experiences and outcomes, there is a need for controlled studies to measure impacts. Using more rigorous SW-CRT design could as- and role in providing services that can facilitate hope, meaning and empowerment. The results of this trial confirm the importance of person-centred care and recovery-oriented mental health systems, but they show that codesign on its own is inadequate. 39 The results of this trial may indicate, then, that health policy is to some degree ahead of practice particularly in terms of what we might be able to expect of codesign. Coproduction and codesign in healthcare improvement are essential to the future of mental healthcare reforms, but it is essential that these methods demonstrate how power dynamics and shared decision-making are attended to. 18 The critical message for healthcare policymakers, service delivery providers and service reformers in this current era of participation in service design and quality im-