2.1 Identifying the research question

A preliminary literature search was undertaken to identify existing reviews and to assess the extent and volume of current evidence to inform this review's aim. The literature search methodology adhered to the inclusion criteria outlined in Table 1, ensuring a comprehensive selection of relevant studies for analysis. No studies were identified that examined the daily operations of RCs and/or initiation and implementation of co-production practices.

Studies were excluded if they focused on a specific context within the RCs setting, that is, dementia and homelessness. Given the Australian focus of this review, the timeframe (2013–2023) was selected to correspond with the establishment of the first RC in Australia in 2012.

2.2 Identifying relevant studies

Seven databases were searched for published, peer-reviewed literature: Embase, CINHAL, ProQuest, Medline, PsycINFO, Emcare and Scopus. A grey literature search was conducted via ProQuest Dissertations and Theses and Google Scholar, using the same search terms (Table 2).

2.3 Selecting studies

Identified citations (n = 4122) were extracted into EndNote 20 and duplicates were removed (n = 1125). Article titles and abstracts (n = 3002) were screened. The remaining articles (n = 79) were retrieved in full and imported into Rayyan. Full-text screening of the remaining articles was undertaken³⁵ by two independent reviewers. A third reviewer was engaged where there was any disagreement to achieve consensus. Following a full-text review, 70 records were excluded. Manual hand searching was conducted of all reference lists of included studies to identify additional literature (Figure 1).

2.4 Charting the data

Selected articles were charted using a descriptive-numerical approach³⁴ (Table S1; Box 1).

2.5 Collating the data

Following the development of the data extraction table, key content for further description was transferred into a Word document and narrative descriptions were created. Each of the included papers was discussed in detail with the entire research team and reviewed during consultation.

2.6 Fidelity criteria

Studies were assessed to determine the reporting of fidelity criteria underpinning RC operations. A yes/no checklist was used to identify the inclusion of the criterion. Please refer to Table 6.

2.7 Consultation

Consultation via semi-structured interviews using Microsoft Teams was conducted with authors of the included articles, providing methodological rigour in validating findings.³⁴ The corresponding authors of each included article (n = 8) were invited via email to participate in an interview. The interviews explored the author's understanding of recovery, co-production and the requirements for the successful implementation of RCs. Narrative summaries and quotes from the consultation were used to illustrate the findings.

3.1 Overview of characteristics of the included studies

Eight studies^{2, 19, 22, 27, 28, 36-38} were conducted in the United Kingdom and two in Australia.^{9, 21} Five studies^{9, 19, 21, 22, 27} were situated in the mental health sector and three^{2, 37, 38} did not specify sector. King and Meddings³⁶ indicated a variety of sectors, with the majority located within the healthcare sector (52%). Zucchelli and Skinner²⁸ reported healthcare settings, nominating mental health as the most appropriate sector for RCs to be located.

3.2 Sample

Four studies^{19, 22, 27, 37} operated out of Mental Health Trust bases, which are organisational entities within the UK National Health Service responsible for mental health service provision. Two studies^{9, 21} operated within existing clinical mental health services, the remaining four^{2, 28, 36, 38} did not specify their RC location.

Six studies^{9, 19, 21, 22, 27, 37} examined the perspectives of RC staff and students, while one study² included RC managers only. King and Meddings³⁶ used existing RC networks such as the Recovery College International Community of Practice and ImROC, while Zucchelli and Skinner²⁸ gathered insights from students, staff and other Trust staff involved in RC operations. In the remaining study Meddings et al.³⁸ gathered reflections of stakeholders involved in developing a RC pilot. Study sample sizes ranged from 11³⁷ to 63 participants.² Three studies^{22, 28, 38} did not state their sample size.

3.3 Study design

Four studies^{2, 19, 27, 37} used purposive sampling and one employed snowball sampling.³⁶ The remaining studies (n = 5) did not report sampling methods.^{9, 21, 22, 28, 38} Most studies used qualitative methods (n = 6).^{19, 22, 27, 28, 37, 38} Three studies^{9, 21, 36} used mixed methods (e.g., interviews with a survey) and one study used a quantitative hierarchical cluster analysis.² Five studies^{2, 9, 19, 21, 27} documented receiving human research ethical approval.

3.4 Operational and contextual information

Seven studies^{2, 9, 21, 22, 28, 36, 38} provided information pertaining to organisational factors that impact RC operations. For example, infrastructure and resource management. Eight studies^{2, 9, 19, 21, 22, 27, 28, 37} referred to infrastructure such as buildings, classrooms and other spaces where RC activities took place. Two studies^{36, 38} did not specify physical facilities.

Two studies^{22, 28} provided details of course content. Zucchelli and Skinner²⁸ referenced 37 courses split into four categories: (1) understanding mental health difficulties, (2) rebuilding your life, (3) developing knowledge and skill and (4) getting involved. One study³⁸ stated how RC courses were marketed through a course prospectus, one study² referenced the cost of developing courses equated it to 8000 pounds to develop due to the resource-intense nature of co-production. One study addressed mechanisms for the moderation of subject matter and course construction.²²

Three studies^{2, 21, 28} referenced the development of a peer workforce, senior management endorsement of the presence of lived experience in decision making and the space to build a competent and confident workforce. One study²⁸ provided details on the composition of RC staff. Four studies^{21, 22, 27, 38} highlighted the importance of educational pedagogy in the RC model. For instance, Ali et al.¹⁹ examined the creation of shared learning spaces to facilitate student learning. Similarly, McGregor et al.²² explored the role of co-production and recovery-focused approaches, which provide authentic collaborative learning opportunities.

Four studies^{2, 9, 21, 28} described the importance of dedicated funding and sufficient resources for successful RC operations. Hayes et al.² described costs associated with implementing RCs, including those associated with staffing, rent and course development. For example, Participant 3 states: ‘Like a lot of RCs who use co-production, it seems like a lot of the time is free, which is another contentious and difficult issue, people should really be paid for their time. With co-production you do have to the time to do it but also you've got to have the funds to do it properly’. Other operational considerations included the importance of policies and processes,^{22, 38} administrative support,⁹ training and development³⁸ and managing organisational cultures.³⁸

Five studies examined funding aspects.^{2, 21, 36-38} Government departments, particularly mental health departments, were commonly identified as the primary funding source.^{36, 37} Hayes et al.² reported three-quarters of the RCs in their study received funding from the NHS. Two studies^{21, 37} described pilot grants as a means of funding, while Meddings et al.³⁸ detailed funding originating from a partnership between voluntary sector organisations and a Trust. The remaining studies (n = 5)^{9, 19, 22, 27, 28} did not specify where RC funding originated. Nucelli and Skinner²⁸ highlighted the complex nature of financial sustainability and the challenges associated with securing funding.¹⁹ All studies reported a description of RCs (n = 10). Reported definitions are presented in Table 3. Of the 10 studies, nine described their educational institution as a ‘Recovery College’ and one as a ‘Discovery Centre’ (dedicated youth focus).⁹

Participant 3 reinforced this perspective noting; ‘RCs can look different and actually they should not be cookie cutters of each other, as they should be co-produced’. Further corroborating this point participant three went on to communicate awareness of RCs was relatively low outside of those that attend or work in them: ‘But I think in terms of the general public knowing about RCs, not really’ (P3).

Six studies^{9, 19, 28, 36-38} provided reasons to justify the establishment of RCs. These included: a desire for transformative change within mental health services to embrace a more recovery-orientated approach,^{19, 21, 22} a commitment to addressing the community's specific needs,³⁶ drawing inspiration from existing RCs,³⁶ providing support for individuals facing mental health challenges while tackling healthcare disparities⁹ and catalysing organisational shifts by influencing attitudes and culture.^{9, 36}

All studies referenced co-production; one provided a specific definition.²² Co-production was framed as the relationship between those with lived experience of mental ill-health and mental health professionals. Five articles^{9, 27, 28, 37, 38} mentioned that co-production involved people with lived experience working alongside clinicians. Three studies^{22, 27, 38} mentioned how co-production is applied in practice. For example, Meddings et al.³⁸ described how co-production was embedded within RC operations, from course development to training, governance and evaluation. Table 4 provides author-reported descriptions of co-production.

The consultation suggested that co-production was a central tenet of RCs, however, participants reported a range of considerations for embedding principles in practice. For example, ‘I think people were trying to run before they could walk’ (P2), suggesting that there may be a tendency to prioritise expediency over thorough co-production efforts, without fully recognising the time and commitment required for co-production. Another participant suggested that co-production was challenging: ‘I think that's something that clinicians really struggle to get their heads around’ (P1). Finally, participants highlighted the relational aspect of co-production and the role of power and expertise when working with those with professional and personal experience. For example, ‘The idea is that you have equal input from people who are experts by training and people who are experts by experience, in all facets’ (P2).

Five studies^{2, 9, 19, 22, 37} described the concept of recovery. One study¹⁹ acknowledged various types of recovery, providing descriptions of personal and clinical recovery. Consistently, broad frames of the individual journey were applied to descriptions of personal recovery. Conversely, the frames surrounding clinical recovery included symptoms, treatment and/or cure. Table 5 presents author-reported descriptions of recovery. Consultation supported findings, highlighting the variations in interpretations of recovery suggesting its temporal nature, the contested language and the juxtaposition between the medical and social aspects of recovery. For example, ‘But like I said, recovery means different things to different people’, (P3) a sentiment echoed by Participant 2: ‘So I'm big in kind of personal recovery or social recovery, in particular as opposed to clinical recovery. So it should be people get out of it what they're looking to get out of it, you know at their particular time of their own journey’.

Seven studies cited a fidelity criterion.^{2, 9, 19, 22, 28, 36, 38} Perkins et al.¹² was cited five times^{9, 19, 28, 36, 38}; Nottingham²² was cited twice^{22, 38} as was RECOLLECT,^{2, 23, 36} while three studies^{21, 27, 37} did not cite a fidelity criteria. Table 6 shows the cited fidelity criteria.

For example, one study²² detailed fidelity criteria concerning the operation of RCs, describing how the RC met each criterion in both the results and discussion section. Another study² reported that most RCs in their sample scored high on fidelity and this was reported in the discussion. However, the consultation indicated that interpretations of fidelity criteria varied based on the local context and were not always suitable as a measure of RC success. As Participant 2 stated: ‘Those measures are a useful starting point to get people to think through these other things that we need. There are things we need to commit to and not beating ourselves up if we don't quite get there’. This participant also expressed in terms of measuring either the impact and/or success of RCs: ‘Fidelity doesn't, doesn't feel right to me’ (P2).

All studies (n = 10) reported study outcomes, from the need for more clarity around terminology¹⁹ to working through power imbalances²⁷ and expectations in co-productive working arrangements.^{9, 38}

Three studies^{2, 36, 37} reported on their limitations, including data collection taking place during COVID-19,² bias due to reliance on RC managers to self-report² and lack of generalisability due to small sample sizes.^{36, 37} More than half of the studies^{2, 22, 27, 28, 37, 38} provided recommendations for future research endeavours. Recommendations included further evaluations to determine the effectiveness of RCs,^{2, 37, 38} outcomes²² more in-depth exploration of the fidelity criteria^{2, 22} and co-production^{2, 22} as well as operational running costs and commissioning arrangements.^{2, 28, 37}

4.1 Positioning RC operations: Navigating diversity

The RCs in the identified studies were predominantly positioned within the mental health sector, for example, RCs described in the papers by Hopkins et al.²¹ and Hopkins et al.⁹ were linked to established mental health services. These findings are consistent with broader literature, where there are few examples of RCs extending their presence beyond the mental health sector. For example, King and Meddings³⁶ report only 10% of RCs are led by educational bodies. RC affiliations with the mental health sector may reflect changes in both national and international policies emphasising recovery-orientated mental health care.² However, the alignment of the RC model with the medical mental health model remains a source of tension^{20, 39, 40} as highlighted by Hayes et al.,² who found that just under half of the RCs were not affiliated with statutory health services.

While some authors have argued for RCs as distinct educational entities located independent of the mental health sector, others have suggested that the medical and the RC educational model should not be seen as binary or oppositional.^{20, 39, 41} Included studies emphasised the importance of aligning RCs with educational pedagogy^{21, 22, 27, 38}; however, details on the use of educational paradigms, such as transformative learning theories, were lacking. This finding suggests that while there is recognition of the need to integrate pedagogical principles into the design and delivery of RCs programs, there are opportunities for future research on the practical implementation of these educational theories.³¹

Existing literature suggests that while RCs are framed as learning centres, they should continue to engage with mental health professionals to prevent them from losing relevance within the broader mental health sector.^{20, 39, 40} However, as evidenced in this review, attracting mental health professional expertise remains challenging^{9, 42} and specific details on the engagement of mental health professionals remain largely absent from the broader literature.^{20, 28} These findings collectively emphasise the need for a more comprehensive understanding of the operational identity of RCs. Such exploration is vital not only for addressing the ongoing challenges of engaging mental health professionals but also for enhancing the overall acceptance of RCs within the broader community.

4.2 Definitional and language ambiguity: Shaping perceptions

Definitional ambiguity related to RCs was found to contribute to a lack of clarity regarding their role and value both in the operationalisation of RCs. The included papers presented a diverse range of interpretations (see Table 3). As Collins et al.²⁰ asserts, lost meaning is a real challenge faced by RCs. Similarly, Perkins and Repper⁴⁰ have suggested that criticisms directed at RCs may stem from a misunderstanding of their mission. Remarkably, none of the included studies presented a consistent definition of co-production or recovery. This lack of consensus is notable, considering the foundational importance of both these concepts as guiding principles¹² (see Tables 4 and 5). The consequence of divergent interpretations may not only hinder the development of RCs but may also inadvertently sideline structural factors influencing individuals' access to education, recovery opportunities and the very existence of RCs.³⁷ This raises a significant question: is there a genuine gap in understanding these fundamental concepts, or does it reflect a resistance to the adoption of new approaches within this field? These findings suggest a need for improvements in definitional clarity of terms and the creation of a shared understanding.

RCs are often exemplified by their collaborative nature. Co-production has been widely recognised as the transformative element of the RC model,^{11, 30, 43} as a process that can challenge established hierarchies.⁴⁴ The reviewed studies demonstrated consensus that co-production involved service users and providers in designing, delivering, and evaluating RCs. However, evidence of this collaborative process was lacking. None of the studies provided detail on how to establish or maintain co-production, or processes to embed diverse perspectives, or respond to contexts whereby co-production maybe diminished. Moreover, the established lack of understanding of co-production, along with its multiple interpretations,³⁰ may lead to confusion about its true application and value in RCs. This finding highlights an important opportunity for research to address the gap in a consistent understanding and practice of co-production within RCs.¹⁸

4.3 Fidelity criteria: Beyond adherence measures

While fidelity criteria were acknowledged across all studies in this review, the lack of application to the operationalisation of RCs was evident. This finding was supported by the stakeholder consultation which suggested a need to shift from rigid fidelity criteria towards exploring and identifying fundamental aspects that underpin the operationalisation of RCs. The broader literature, including Hayes et al.,² highlights the complexity of RC operations and the challenges in establishing fidelity measurements.⁴⁵ The lack of consistency in the interpretation and utilisation of fidelity criteria in this review raises questions about the contextual relevance to RC processes, consistent with existing literature,^{18, 45} which illustrates an evolving field that has given rise to a multitude of diverse evaluation standards. Findings suggest that the term ‘fidelity criteria’ may lack contextual relevance to RC processes and point to the need for a more context-specific approach. For example, criteria such as social connectedness are subjective, driven by values and attitudes,^{18, 23} making measurement challenging. As suggested by Meddings et al.,⁴⁶ striking a balance between measuring aspects of RC operations required by funding bodies and commissioners and fostering a culture of service improvement, while staying true to the ethos of RCs and encouraging ongoing innovation²³ is essential. However, the lack of clarity about the status of the evidence for the fidelity measures, particularly their relationship to outcomes, remains a concern. Without understanding this relationship, it is difficult to assess the significance of the variation in operationalisation across RCs. Addressing these concerns is crucial for advancing our understanding of fidelity in RCs and ensuring meaningful measurement for all stakeholders involved.