Co-producing research on psychosis: a scoping review on barriers, facilitators and outcomes

Abstract

Introduction

Co-production is a collaborative approach to service user involvement in which users and researchers share power and responsibility in the research process. Although previous reviews have investigated co-production in mental health research, these do not typically focus on psychosis or severe mental health conditions. Meanwhile, people with psychosis may be under-represented in co-production efforts. This scoping review aims to explore the peer-reviewed literature to better understand the processes and terminology employed, as well as the barriers, facilitators, and outcomes of co-production in psychosis research.

Methods

Three databases were searched (MEDLINE, EMBASE, PsycINFO) using terms and headings related to psychosis and co-production. All titles, abstracts and full texts were independently double-screened. Disagreements were resolved by consensus. Original research articles reporting on processes and methods of co-production involving adults with psychosis as well as barriers, facilitators, and/or outcomes of co-production were included. Data was extracted using a standardised template and synthesised narratively. Joanna Briggs Institute and the AGREE Reporting Checklist were used for quality assessment.

Results

The search returned 1243 references. Fifteen studies were included: five qualitative, two cross-sectional, and eight descriptive studies. Most studies took place in the UK, and all reported user involvement in the research process; however, the amount and methods of involvement varied greatly. Although all studies were required to satisfy INVOLVE (2018) principles of co-production to be included, seven were missing several of the key features of co-production and often used different terms to describe their collaborative approaches. Commonly reported outcomes included improvements in mutual engagement as well as depth of understanding and exploration. Key barriers were power differentials between researchers and service users and stigma. Key facilitators were stakeholder buy-in and effective communication.

Conclusions

The methodology, terminology and quality of the studies varied considerably; meanwhile, over-representation of UK studies suggests there may be even more heterogeneity in the global literature not captured by our review. This study makes recommendations for encouraging co-production and improving the reporting of co-produced research, while also identifying several limitations that could be improved upon for a more comprehensive review of the literature.

Introduction

The involvement of service users in mental health research is increasingly recognised as best practice [1,2,3]. A previous review by Jo Brett and colleagues [4] identifies numerous positive outcomes of user involvement in health and social care research, across all stages of the research process [4]. In the early stages of agenda-setting and research planning, these include: identification of new research gaps and generation of further research questions, ideas, designs or proposals; improvement in the cultural equivalence of research tools and ethical considerations of trial design; and even the development of new medications. In terms of methodology and data collection: better application of the concept of informed consent and interpretation of information for participants; improved study design, including decision-making on endpoints, time of recruitment and selection of outcome measures; and more sensitivity to the general research climate. During write-up, dissemination and the implementation of results: opportunities to voice concerns about the interpretation of results and how implementation might be affected; and additional assistance in disseminating results to key stakeholders.

However, the definition of service user involvement is broad, and examples can range from tokenistic efforts, to sharing all decision-making and control, to entirely user-led research [5]. Indeed, Brett et al. [4] highlight that much of the research that claims to involve service users is limited to single-stage consultation, whilst user-led or collaborative efforts are more likely to promote involvement across all stages of the research process [4]. Co-production is one approach to collaboration “in which researchers, practitioners and the public work together, sharing power and responsibility from the start to the end of the project, including the generation of knowledge” (INVOLVE 2018, p.1) [6] (see Fig. 1 for summary of co-production process). The arguments for co-production typically fall into four categories: substantive (improving the quality of research), instrumental (greater impact), normative (related to the accountability of funders, addressing discourse between civic or public), and political (equality) [7]. With various stakeholders involved in the research process, co-production is particularly valuable as a tool to improve the relevance of and buy-in to academic research [8].

Fig. 1

Co-production: philosophy, principles and outcomes [6, 9,10,11]

Although there are several frameworks and guidelines for co-producing research, there is no single formula for doing it, and this approach is not always straightforward. Perhaps as a result, researchers have been criticised for using the term “co-production” loosely to encompass a variety of different approaches to user involvement, some of which do not necessarily promote an equal sharing of responsibility between service users and researchers throughout all stages of the research cycle. Several barriers to co-production have also been identified. For example, co-production is often time- and resource-intensive, requiring relationship-building among diverse stakeholders—and the requisite exchange of favours and conflict resolution often involved in relationship management—with no guarantee of a good outcome [7]. Even where the co-production of research is successful, the translation of this research into practice can easily be derailed by poor organisation and planning, resulting in the disappointment and dissatisfaction of stakeholders involved [12, 13].

Some groups also have more opportunities for involvement in research than others. In particular, people with severe mental health conditions like psychosis may be under-represented in co-produced research. Psychosis is an overarching term for a number of symptoms, including hallucinations, delusions and thought disorder, which characterise several mental health disorders, such as schizophrenia [14]. A review by Woodall et al. identified several barriers to participation in mental health research among people with schizophrenia: fear, misunderstanding and mistrust of research, the medical establishment and medical interventions; the burden of participation coupled with insufficient remuneration, logistical difficulties and competing obligations; stigma, isolation and general lack of motivation; and severity of symptoms [15]. Any of these barriers could equally apply to co-production, which requires an even more intensive level of engagement than participation as a research subject.

Rationale

While previous systematic reviews have investigated co-production in mental health research [3, 11], the transferability of their conclusions to research involving people with psychosis may be limited. This scoping review aims to explore the peer-reviewed literature as a starting point to develop more robust methods for systematically reviewing and appraising co-produced research on psychosis. The specific objectives are to document the terminology, methods and outcomes reported, as well as the barriers and facilitators to successful co-production in this context.

Methods

Eligibility criteria

Table 1 summarises our eligibility criteria for this scoping review. We included original research studies on psychosis (i.e., not protocols, reviews, commentaries, etc.) published in peer-reviewed journals that involved adults with psychosis, regardless of study design. Where study participants and/or the service users involved in the conduct of the research included a mix of adults and adolescents, we excluded studies where the mean age was under 18 years. As this is a scoping review, we employed a broad definition of psychosis. We primarily used the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria for Schizophrenia Spectrum and Other Psychotic Disorders and the International Classification of Diseases, Tenth Revision (ICD-10) definition of psychosis, which includes schizophrenia, schizotypal and delusional disorders. However, we also included studies where individuals were not directly diagnosed with a psychotic disorder, but were reported as having key symptoms such as delusions, hallucinations or thought disturbances, etc. In addition, we considered studies that failed to report on diagnosis or symptoms, so long as the studies themselves were investigating psychosis and we could reasonably infer from the text that at least some of the users involved came from this clinical population.

Table 1 Summary of inclusion and exclusion criteria

We included papers in which the authors outlined a process of co-production and reported on barriers or facilitators to co-production, and outcomes of co-producing research. We anticipated that terminology and methodology might vary in practical application and therefore used a broad definition for co-production, characterised by the involvement of service users across multiple stages of the research process, starting from the beginning of the study. This allowed for a greater number of potentially relevant studies to be considered. We included papers that explicitly described methods of co-producing research on psychosis. Where this was unclear (for example, where a study did not describe its approach as “co-production” or similar), we referred to the five INVOLVE (2018) principles to make our screening decision: sharing of power, including all perspectives and skills, respecting and valuing the knowledge of all those working together on the research, reciprocity and building and maintaining relationships (see Additional file 4: Appendix D).

Information sources and search strategy

The search strategy was developed with oversight from qualified librarians at King’s College London and the London School of Hygiene and Tropical Medicine. Search terms included subject heading and string searches for concepts related to psychosis and co-production, respectively (see Additional file 1: Appendix A for PsycINFO search terms). Subject headings were tailored for each of the three databases searched: PsycINFO, MEDLINE and EMBASE [16,17,18]. The literature search was limited to studies published in English. No other restrictions were applied. Searches were conducted on 24th June 2020.

Screening and study selection

Results from each database were exported into EndNote X9 [19], where duplicate studies were identified and deleted. Title, abstract and full-text screenings were completed independently by 2 screeners using the Rayyan web platform [20], and any discrepancies in screening were resolved by consensus. The four-phase PRISMA diagram summarised the study selection process (Fig. 2) and the PRISMA checklist was completed (Additional file 5: Appendix E).

Fig. 2

PRISMA Flow Diagram

Data extraction, analysis and interpretation

Data was collected using a data extraction form covering participant and study characteristics and participatory research terminology, methods and outcomes (See Additional file 2: Appendices B, Additional file 3: Appendix C). While the INVOLVE (2018) principles were used to gauge initial eligibility during the screening phase [6], these include general, broad recommendations that are challenging to evaluate critically; hence, INVOLVE’s key features were used during data extraction and synthesis to assess further the extent to which service users were reportedly involved in the research process [6]. The eight key features address factors like continuous reflection, joint ownership of key decisions and a commitment to building relationships (see Additional file 4: Appendix D for the complete list). Each of these features was double-rated as fully present, partially present, not present, or impossible to determine from the included text (see Table 2).

Table 2 INVOLVE key features

Results were synthesised following Popay et al.’s (2006) guidance on narrative synthesis by summarising, tabulating and grouping findings, then describing these narratively [21]. To aid in the interpretation of findings and draw conclusions and recommendations for co-production, co-authors from diverse perspectives (e.g., high-income country vs. low-/middle-income country, female vs. male, professional vs. lived experience of psychosis) participated in a series of six virtual discussion sessions and provided critical input into manuscript drafts.

Quality appraisal

The quality of the included qualitative and cross-sectional studies were assessed using the Joanna Briggs Institute (JBI) Tools [22,23,24], and in the absence of an appropriate JBI tool we evaluated descriptive studies using the AGREE (Appraisal of Guidelines for Research and Evaluation in Europe) guidelines [25] (see Table 3). All the quality appraisals were completed independently by two reviewers, and any discrepancies were resolved by consensus.

Table 3 Co-production Components

Results

A total of 1243 records were identified through the database search. After duplicates were removed, 978 remained, and the titles and abstracts of these records were screened alongside the pre-specified eligibility criteria. This resulted in a decision to exclude 901 records. The remaining 77 records underwent full-text screening, with an additional 62 records excluded. Fifteen studies were included in this scoping review.

Characteristics of included studies

Study designs

There was a mix of qualitative research (n = 5), descriptive studies (n = 8) and cross-sectional quantitative studies (n = 2). Several of the studies had multiple components to them and varied in sample size at each stage. The sampling strategy differed across the studies; however, two of the studies did not specify their sampling strategy, and four studies did not include any participant characteristics. Table 4 provides more detailed information on type of study and other key characteristics, where available.

Table 4 Study Characteristics

Study participants

Most of the studies occurred in high-income countries (most frequently within the United Kingdom), and only two studies included low- or middle-income countries (LMICs). In five of the 15 included studies, the participants’ ages were not specified; however, age ranges were indicated in other ways. For example, Kristensen et al. (2018) stated that the target group for the co-designed patient-reported outcome measure was patients 18 years or older [26].

Involvement of service users

Among the service users involved in the studies, the most common diagnosis was schizophrenia. Service users commonly had roles as members of an expert advisory group or “steering committee”. The collaborative research team was often comprised of service users, healthcare professionals, and academic researchers. A few studies involved service users both as part of the study team and as part of a steering group. Carers were also involved in three studies [27,28,29]. One study had a carer researcher in addition to a patient advisory group [29].

Co-production components of included studies

The terminology used to describe user involvement varied significantly between the studies, though each of the included texts recognised the expertise that service users contribute to research. The methods of service user involvement employed also varied widely: four of the studies sought consensus from service users (participants and/or members of the research team), five consulted service users or service user reference/steering groups, three utilised collaborative workshops with service users, and four had service user researchers.

When the INVOLVE key features were applied to assess the level of co-production across the included studies, it was notable that all of the studies except Higgins et al. [27] failed to report ground rules at the onset of the research project. Roughly half of the studies discussed partial ownership of the research and reported a flexible process, which enhanced relationships and created opportunities for growth and development on the part of both the researchers and service users. All of the studies acknowledged the value of lived experience in research and most of the studies used continous reflections and ongoing dialogue to support an iterative process of collaboration. Table 2 assigns a colour-coded summary to each of the INVOLVE key features, indicating whether the feature is at least partially (yellow) or fully (green) present, not present (red), or whether it is impossible to determine from the texts (no colour). Table 2 further organises the included studies by the terminology used to describe the co-production approach (i.e., “co-”, versus “participatory”, “service user researchers”, “experts by experience”). These categories are also described narratively, below.

“Co-production”, “co-design” or “co-creation”

Although assessing their rigour is challenging, due to lack of methodological justification and data triangulation, the five studies included in this category described a long-term collaborative process referred to as either “co-production”, “co-design” or “co-creation”. These studies demonstrated a flexible approach with ongoing dialogue, continuous reflection, and a commitment to fostering relationships. However, none reported that ground rules were established at the onset of the research, even though this is one of the central features of co-produced research according to INVOLVE [6]. Sin et al. used participatory methodologies described alternately as co-production and co-design in their report, illustrating the interchangeability of terms used to describe user involvement [28]. The researchers used participatory design workshops, iterative consultations with stakeholders, and an Expert Advisory Group (EAG) to design an eHealth intervention. In each workshop, all members reportedly assumed equal decision-making roles and contributed their respective strengths and expertise. Similarly, Terp et al. aimed to improve co-design methods by conducting workshops so that young adults with schizophrenia could become active participants in mental health services [30]. Their workshops utilised various tools, including non-digital materials and metaphors, to encourage active participation. Larkin, Boden and Newton conducted experience-based co-design events to improve healthcare services by enabling service users, carers, and staff to collaborate [31]. Likewise, Kristensen et al. used an iterative co-design method, where the output was based on a co-creation process between patients and healthcare professionals [26]. The Patient Peer Board (PPB) participated in workshops and worked individually, in pairs, and in groups facilitated by an interdisciplinary steering group to reach a consensus. Finally, Realphe et al. invited service users to provide feedback on designing and delivering an online platform for early psychosis to encourage participation from hard-to-engage service users [32]. The authors used an iterative five-stage co-design process with service user involvement throughout all stages, in which the feedback from one session informed the next.

“Participatory”

The six studies included in this category also addressed the INVOLVE [6] principles for co-production but identify their approach as “participatory research”. Higgins et al. included focus group discussions to gauge information regarding the co-design and co-facilitation of the education program. Core values were agreed upon at the project’s outset, making this the only study to report ground rules, and indeed the only to satisfy all eight key INVOLVE features. Schneider et al. also used a participatory research approach in which service users were involved throughout, yet only partially achieved several key INVOLVE criteria. The research team agreed that the findings would be disseminated through a theatrical performance as per the participants’ preferences and published in an academic journal to reach a larger audience [33]. The service users would be listed as co-authors but not involved in the writing of the research publication. The authors, however, acknowledged this project’s benefits in generating knowledge and offering a transformative experience for those involved.

Participatory research studies that involved service user researchers often achieved at least partially the INVOLVE features of joint ownership, personal growth, development, and ongoing dialogue, though there was not always sufficient information reported to make a full assessment. Csipke et al. referred to an adaptation of the “SURE’’ model, where service user researchers were involved in surveying the literature, performing the data collection, and undertaking a considerable part of the data analysis. One of the researchers had experience with the service under investigation, which provided greater insight [34]. Pelletier et al. employed participatory action research, which relied on close collaboration and consultation with an advisory board and the co-research team throughout all stages of the research [35]. The emphasis was on joint participation between non-academic members and the co-research team, where those involved immediately benefitted from the research and knowledge generation process. Additionally, the authors of the research publication included a physician with lived experience. Moreover, Tischler et al. [36] acknowledged the difficulty of establishing collaboration in practice due to the service users’ relative lack of research experience, which required support from the research team [36]. While service users contributed to the study design, they were also participants in focus groups, blurring the lines between research collaborator and research subject. Finally, Susanti et al. explored patient and public involvement (PPI) as a tool to strengthen the Indonesian health system [29]. The PPI advisory group consisted of individuals who either had lived experience of psychosis or cared for someone with a diagnosis of psychosis. Although the PPI advisory group was consulted throughout the entire research process, involvement was perhaps not sufficiently collaborative to qualify as co-production. One carer-researcher co-facilitated focus groups together with two mental health nurse academics and contributed to the analysis of transcripts. All the researchers were supervised by academics overseas in the United Kingdom.

“Service user researchers”

Three studies employed service user researchers in their collaborative approaches and met a majority of the INVOLVE key features. Neil et al. described a previous study’s collaborative process that explored the relationship between service users and researchers; however, the researchers were also the supervisors, suggesting a potentially problematic power hierarchy [37]. Similarly, Pitt et al. used a steering committee consisting of service users who guided the study design and analysis from the commencement of the research [38]. Yet the project was overseen by clinical psychologists, as there was no service user researcher in a supervisory role, which might again suggest that power differentials existed between service user researchers and other researchers. During the analysis, the researchers’ interpretative influences on the data were mitigated by service user researchers who had lived experience and by input from the steering group. Morant et al. [39] utilised semi-structured interviews developed with a mental health service user group [39]. Service users conducted the interviews and disclosed their status as service users as part of rapport-building. Morant et al. [39] were aware of the existing preconceptions of academics, researchers and clinicians, and integrated service user perspectives into the early stages of the analysis process. Service users were invited to review, discuss and modify an initial set of emerging themes; however, the amount and impact of collaboration remains unclear. Although service user researchers conducted the study, much like Neil et al. [37] and Pitt et al. [38], the process still appeared to be guided by academics.

“Experts by experience”

Although the final study explained that they involved ‘experts by experience’ in their research, they exhibited a minority of the INVOLVE features. Roelandt et al. stated that service users and carers partook in all stages of the research [40]. Although this aligns with co-production principles, it is not described further, and the extent to which users and carers participated may be limited.

Barriers and facilitators to co-produced research

Common barriers to co-production reported by the included studies were challenges with safeguarding [28, 37, 41], power imbalances [37, 39], mental health stigma [29, 31, 35], and high turnover among mental health professionals [31, 38]. Included articles highlighted that co-produced studies must pay extra consideration to recruitment strategies and study settings, aiming to mitigate barriers that may affect participants’ ability or willingness to partake in research, such as economic disadvantage or mistrust (for example of the Internet) [28, 32, 37, 41]. Another barrier identified was power differentials, namely where researchers maintain control of the process: for example, unilaterally deciding what materials to share from a data analysis and co-design event [37, 39]. Other limiting factors were existing preconceptions of user involvement and mental health stigma. For example, Susanti et al. reported patients’ past experiences with health professionals in which complaints during co-produced research were interpreted as a relapse in their condition [29]. Researchers can acknowledge these power imbalances and take active measures to ensure genuine collaboration. For instance, Tischler et al. [36] expressed the need for adequate support of service users when attending research meetings with other health professionals to empower service users to express their views [36]. As demonstrated in two studies, high turnover in mental health personnel can also impact the relationship between service users and health care professionals [31, 38]. As good communication and ongoing dialogue are essential facilitators for co-produced research, a critical barrier to the success of these projects emerged when support from personnel was discontinued [31].

Common facilitators of co-production reported by the included studies were stakeholder buy-in, effective communication [30, 33, 35], and additional support by providing and presenting materials in multiple formats and creating a safe and informal environment [28, 36]. For example, Sin et al. provided nondigital materials such as post-it notes and pens to facilitate a creative and inclusive design atmosphere [28]. This allowed all participants to engage in the co-design workshops, regardless of their technical abilities. Co-production was found to enable effective communication by offering the opportunity to resolve potential tensions throughout the process [26, 31] and empowering marginalised groups through genuine collaboration [26, 33, 38]. Further, as Realpe et al. explained, it is important that researchers communicate how the study data will be stored as this might be a particular concern for service users with a history of psychosis [32].

Outcomes

The included studies agreed that service user involvement in research provided greater insight by harnessing lived experience. Collaborative research approaches created a secure and informal environment, which encouraged mutual engagement and resulted in more in-depth understanding and exploration [26, 27, 30, 33, 36,37,38,39]. Pitt et al. explained that user-led research enabled strengthened relationships, facilitated communication, and created a unique rapport during user interviews, allowing for the collection of data that might not otherwise be accessible [38]. Involvement allowed service users to share personal experiences, knowledge and skills, and contribute to a flexible, collaborative and iterative approach [27, 32, 37, 40]. For example, Neil et al. acknowledged that the mix of knowledge and skills allowed for healthy debates and a regular review of topics; this was essential to ensuring their results were valid and meaningful [37].

Additionally, meaningful service user involvement addressed systemic barriers within mental health care by, for example, building skills, improving confidence and social connection, and combating stigma [29, 30, 33, 35]. As Terp et al. and Schneider et al. explained, increased ownership and pride amongst service users created a sense of meaning and purpose, and a collaborative research empowered marginalised individuals to participate in the process of change [33, 42].

Robustness of results

All included studies reported positive outcomes of involving service users in the research process, indicating some degree of consensus within this body of research. However, the quality of the study designs and reporting, and particularly the lack of methodological justification, make it difficult to assess their rigour and call the robustness and generalisability of these studies into question (see Table 5 for details). For example, although it is generally accepted that qualitative studies tend to have smaller sample sizes, none of these studies transparently discussed how sample sizes were determined. Study reports often had missing components, including information regarding the researchers’ reflexivity, inclusion criteria for the participants, and sampling strategies. In specific instances, the broad spectrum of psychosis and potential for consensus bias [26, 31], issues with recruitment strategies [39], and problems with language translation [40] raised questions of validity. Reflecting the dynamic nature of collaborative research, several stages of research were often reported in a single paper, blending different techniques of analysis, which frequently lacked clear justification.

Table 5 Joanna Briggs Quality Appraisal Summary

Discussion

This review highlights the heterogeneity of co-produced research on psychosis in terms of the terminology and methods employed, as well as the quality of the research itself, the co-production element and reporting. Although the included studies reported service user involvement in multiple stages of the research process and emphasised its benefits, few described their methods of co-production in sufficient detail. Levels of involvement appeared to vary, from service user researchers conducting the research, to steering group members providing consultation and feedback, but with control of the research remaining largely in academic researchers’ hands. Further, several studies employed service users both as research participants and members of the research team; in these instances, it was challenging to distinguish between the service users’ roles, which may have resulted in their involvement being more tokenistic than co-productive.

In their critical analysis, Tierney et al. echoed this challenge: studies employing user involvement failed to provide a working definition of involvement, methodologies often lacked congruence, and although studies cited positive outcomes, it remained challenging to assess their quality and this raised questions about bias [43]. Importantly, in their choice of terminology, Edwards and Elwyn explained that authors do not always recognise the difference between involvement by shared information and involvement by shared decision-making power, incorrectly characterising their approach as “co-production” or similar [44].

Power-sharing difficulties are a key barrier to achieveing truly collaborative research: the equality of power necessary for co-production cannot be actualised while service users are perceived as having limited capacity [45]. Previous research has detailed that positive reports of service user involvement, such as effective communication, must be balanced with an acknowledgement of barriers, such as service users feeling overburdened by their involvement [46]. Co-production is also subject to social and economic constraints, as co-producers tend to be older, female and living in urban environments [47]. Studies included in our review emphasised turnover of staff, rather than the constraints faced by individuals with psychosis, as one of the biggest barriers, as lack of continuity has a significant impact on the patient/practitioner relationship as well as the translation of research findings into practice [31, 32, 38].

Previous literature has suggested that academic researchers experience challenges when conducting a co-produced project. These challenges include building and managing relationships, defining and adapting the project’s scope, and maintaining a professional identity [48]. Past studies have indicated that co-produced research typically favours academics with specific personality traits, such as generalists with good communication and conflict resolution skills, and a creative individuals who can be flexible while maintaining research integrity and rigour [49].

When considering co-production in a mental health context, researchers must be sensitive to the stigma, discrimination and general disempowerment experienced by individuals with mental health conditions in services and in society at large. In particular, researchers must consider the power dynamic that exists due to the legal relationship between mental health professionals and patients [50].

Strengths and limitations

This scoping review utilised systematic methods for the identification, appraisal and synthesis of research in an area that has often been criticised for insufficient standardisation of terminology, processes and reporting. Further, it draws attention to an underrepresented group in co-produced research: people with lived experience of psychosis. The methods used in this study for a relatively small number of languages and databases, and for peer-reviewed journal articles exclusively, can be built upon for a more comprehensive and global systematic review of the grey and published literature from a wider range of sources (see Table 6 for recommendations).

Table 6 Summary of Recommendations

However, these methods are also subject to several limitations which must be taken into account: (1) first, as this is a scoping review, we prioritised sensitivity over specificity, and may have been over-inclusive in our search and selection process; (2) on the other hand, the age restriction employed in screening may have excluded potentially relevant studies, for example, on early intervention in psychosis; (3) the quality of reporting may limit the robustness of any conclusions that could be drawn; (4) lack of standardisation in the evaluation of co-production made it difficult to define outcomes; (5) more targeted methods may be required to correct for the over-representation of UK studies.

As described in our methods, we employed a broad definition of co-production in our eligibility assessments. Future reviews might consider whether to adopt both the INVOLVE principles and key features in screening decisions in order to reduce heterogeneity. After applying the key features during data extraction, we came to question whether several of the included studies reflected a co-production approach in practice, or just in principle.

Many studies were also insufficiently detailed, for example in describing the characteristics of the users involved in co-production as well as the research participants. Collecting and reporting this information is important not only for screening, but also as an indication of how representative a sample is and whether certain groups may be more or less likely to engage in co-production. Further, our eligibility criteria specified an age limit and a focus on psychosis. Clarification was sought from the authors; however, they were generally either unable to provide missing information or did not respond. Rather than automatically exclude these studies, screeners made judgment calls based on the best available information (for example, assuming that studies mentioning “adults” fulfilled the age criterion, even if mean age was not recorded). Subsequent reviews may be less generous in their screening decisions; however, as this is a scoping review, we felt it would be benenficial to be as inclusive as possible, in order to better understand the research landscape in this area. We would also suggest that future reviews consider removing the age restriction entirely, as this was not only difficult to apply in practice, but may have led to the exclusion of potentially relevant studies.

Inadequate reporting on methods made it particularly difficult to employ the Joanna Briggs quality appraisal tools, which focus on whether the research project is succinct, congruent and well-justified. The included studies usually involved several workshops, interviews, focus group discussions and questionnaires, often with small sample sizes and without any clear methodological justification. This presents a challenge not just for research synthesis, but for anyone seeking to reproduce a co-produced research study. Indeed, this scoping review highlights a common theme in the literature on co-production: one of the most critical challenges lies in the conceptual and methodological ambiguities for practical application [51]. There was no uniform approach to co-production across the studies included in this scoping review, which covered a wide range of different terminologies and methods, making it difficult to compare and contrast results as part of a robust synthesis.

Finally, it is important to note that this review is heavily skewed toward high-income anglophone countries—the UK specifically. A search strategy covering more involvement terms (e.g. “peer” and “stakeholder” involvement, etc.), more languages, more databases and a wider range of publication types—grey literature, in particular—could potentially uncover more co-produced research on psychosis from around the world. Involving a multilingual advisory group with expertise in different world regions would be especially helpful to improve on our methods. However, in the case of LMICs in particular, it is possible that there simply is not much relevant literature currently available. A 2016 review of user and caregiver involvement in strengthening mental health systems in LMICs identified just one example of user involvement in research (in Brazil), and this was limited to data interpretation [52]. An update published in 2019 identified another possible example in India, in which some data collectors with disabilities may have also had psychosocial disabilities, but this could not be confirmed from the text [53].

Conclusion

This scoping review explored the peer-reviewed literature on co-production in psychosis research. The terminology and methods employed in this area of research vary greatly: even when the same term is used, the methods described may differ, and vice-versa. The quality of co-produced research is also challenging to assess, both because of poor reporting and the lack of methodological justification by researchers, and because co-production simply does not follow a traditional, linear research process. Co-production is often complex, involving multiple stakeholders working together through an iterative process. Yet, there are common barriers and facilitators to co-production with people with lived experience of psychosis. Unfortunately, although co-production is based on the ethic of shared power and equal collaboration, our scoping review suggests that this is not always the case in practice. Researchers should work to address these barriers and build on common facilitators from the earliest stages of study design, to improve the chances of successfully co-producing research with this under-represented population.