University of Birmingham Participatory research meets validated outcome measures

Funding for care service research is increasingly subject to the satisfaction of two requirements: public involvement and adoption of validated outcome tools. This study identifies competing paradigms within these requirements and reveals significant challenges faced by researchers who seek to satisfy them. The focus here is on a study co ‐ produced between academic researchers and people with experience of adult social care services. It examines to what extent research studies can conduct high ‐ quality public involvement and genuine co ‐ production of knowledge, whilst attempting to produce quantifiable outcome scores. Findings add to debate around how to incorporate diverse perspectives in research, which may draw on incommensurate accounts of validity and reliability. Findings also highlight constructive attempts by academic and co ‐ researchers to make the combination of approaches work in the field. These small scale acts of researcher agency indicate some scope to combine the two approaches in future research studies. However, conclusions foreground the importance of broader awareness of how tensions and power imbal-ances related to this combination of approaches play out in social policy research practice.


| INTRODUCTION
In recent decades, two approaches to social policy research have become prominent in England: an emphasis on validated outcome tools to assure quality across health and social care (Department of Health, 2016 and public involvement through co-production (Department of Health, 2007, 2012Simmons, 2011). These trends are visible internationally, as supranational initiatives focus on harmonising approaches and expanding the evidence base about outcome measures and involvement/co-production (Brett et al., 2014;Cochrane Consumer Network, 2017;ICHOM, 2017;Science Business, 2017). At first glance the connection between the two priorities seems aligned. Within political rhetoric, greater use of outcome tools, especially where they are person centred, helps to achieve the goal of embedding patient and public perspectives. However, philosophical inquiry into the process of co-production warns of fundamental tensions that arise from attempts to combine the diverse traditions of political, organisational, professional, and public domains (Bevir, 2013;Carr, 2007).
Health and social care researchers are actively negotiating one such tension as they manage competing agendas of involving the public and producing types of knowledge that are considered useful for policy planning audiences.
The prominence of co-production and outcome measures in policy discourse makes them both attractive to research funders. But how realistic and desirable is it to combine these elements within research studies? Does inclusion of prespecified outcome measures limit emergent public perspectives and threaten to make public involvement more tokenistic? Can these approaches be effectively combined, or must they sit alongside each other in separate work packages with no attempt at integration, in the same way that mixed methods studies can fail to integrate qualitative and quantitative findings (Woolley, 2009)? This article draws on our experience of undertaking co-produced social care research to explain how different participants perceive these tensions in practice. The co-researchers referred to in the study are people who have used adult social care services and their carers. Our aim is to make explicit tensions that are rarely shared given that funders and applicants have a common incentive to sustain the claim that these forms of knowledge can be productively combined.
Evaluation is "concerned with passing judgements about the worth of an initiative based on a critical assessment of available evidence" (Sullivan, 2011, p. 499). In England, the New Labour governments established large-scale evaluations as a key practice (Powell, 2002;Sullivan, 2011), on the assumption that measurement tools could readily give answers to the effectiveness of even very complex interventions. This is a practice that has been continued by subsequent U.K. governments and is common in many other advanced democracies. The evaluation model, with its commitment to surfacing "what works," feeds the performance measurement and regulation regimes of public services (Jones, 2011). Sullivan draws attention to the ways in which these approaches have "privilege [d] evaluation as technical expertise-endorsing its capacity to generate 'truth' through the deployment of ever-more sophisticated methods and models…" (Sullivan, 2011, p. 508).
Co-production is a form of public involvement that can be defined as "joint working between people or groups who have traditionally been separated into categories of user and producer" (Durose, Needham, Mangan, & Rees, 2017, p. 135). Internationally, the concepts of public involvement and epistemologies of coproduced research are still evolving. Practices vary greatly and are often guided by competing justifications, traditions, and discourses (Rose, 2017;Tritter, 2009). Recent NIHR INVOLVE guidance highlights shared power and responsibility as a key feature in the co-production of research "an approach 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" (NIHR INVOLVE, 2018, p. 4). The "public" in this setting may refer to members of the general public, patients using health service, and/or people using social care services.
Co-production can be a goal in itself and/or a means to enhance the quality of public services. Most coproduction literature focuses on policy and services, although more recently, there has been a growth of literature on the co-production of research between academics and people with lived experience of services (Beebeejaun, Durose, Rees, Richardson, & Richardson, 2014;Flinders, Wood, & Cunningham, 2016;Verschuere, Brandsen, & Pestoff, 2012). Literature reflecting on patient and public involvement in the health system contains calls for a more critical approach, which is sensitive to policy and organisational contexts (Learmonth, Martin, & Warwick, 2009;Madden & Speed, 2017).
In the social policy field, researchers have drawn attention to the epistemological difficulties of using participatory approaches, such as co-producing research with public involvement. Richardson (2013, p. 483) explores the tension between positivism and participatory research, suggesting that the latter has "tended to align with methodologies which de-privilege the idea of objective evidence of policy effectiveness." Crompton, Waring, Roe, and O'Connor's (2017) work on deliberative priority setting involving professionals and the public highlights different sources of knowledge and the privileging of "technical" over "social" data. This article builds on these insights. We focus particularly on the tensions between co-producing research with the public and evaluation methodologies incorporating scored outcome measures. In line with the concept of decentring (Bevir, 2013), we explore how two elite narratives of research commissioning (you must utilise qualitative and quantitative methods; you must involve the public and people who use services) are interpreted and made to work by frontline academics and co-researchers.

| PARTICIPATORY RESEARCH AND OUTCOME MEASURES AS COMPETING AGENDAS
Tensions between participatory approaches and evaluation tools are best characterised by examining the knowledge they seek to produce. Here, we highlight two distinctive barriers: emergent concepts versus prescribed measures and the democratisation of knowledge production versus the use of specialist tools. Those with experience of undertaking co-produced research often present the inevitable messiness, clashes, and risks as the very essence of "what makes it so fresh and innovative" (Flinders et al., 2016, p. 261). For Filipe et al. (2017, p. 1), the capacity to form original combinations of participants, each embodying unique knowledge, promises genuinely novel content-"co-production can be understood as an exploratory space and a generative process that leads to different, and sometimes unexpected, forms of knowledge, values, and social relations." In contrast, validated outcome measures provide standardised approaches to measure the value and effectiveness of services. Internationally, there has been a growth in the development of subjective well-being measures, which allow the development of comparable indices of quality of life (QoL; Austin, 2016;Bache & Reardon, 2013;Bache, Reardon, & Anand, 2016). In health and social care, these are considered to be a leap forward from approaches that measure easily quantifiable outputs (number of care visits a day and number of surgeries performed) and are increasingly visible in health and social care research and evaluation (Couzner, Ratcliffe, & Crotty, 2012;Glasby, Allen, & Robinson, 2018;Jones et al., 2013). The value of subjective outcome measures is associated with their ability to put a score on the more intangible features of well-being (Caiels, Forder, Malley, Netten, & Windle, 2010).
Examples include EQ-5D to measure health-related QoL (Gusi, Olivares, & Rajendram, 2010), the General Health Questionnaire to measure psychological well-being (Goldberg, 1992), WEMWBS for mental well-being (Tennant et al., 2007), and English Adult Social Care Outcomes Toolkit (ASCOT) to measure care-related QoL (Caiels et al., 2010). These outcome measures have been widely used for audit and strategic decision making by policy-makers and local service planners. ASCOT, for example, is a central measurement tool in the English Adult Social Care Survey sent out to people who use publicly funded care services by the Department of Health.
The formulaic, scored conception of QoL-and the emphasis on fidelity to the protocol in gathering the dataseen in these kinds of outcome measures contrasts with the notion of the democratisation of patient and user perspectives. Rights-based approaches to involvement highlight the importance of emergent contributions of knowledge gained from open dialogue between those that traditionally hold power and those that have been passive (Ellins et al., 2012). To this end, Williamson (2008, p. 102) identifies guiding principles of an emancipatory patient movement as "… unmasking of new issues … and the gradual acceptance of some of the ideas … that activists work to promote." For a research project, to do justice to the epistemological commitments of both the emergent and the foreclosed requires more scrutiny than it has received to date.
It is important not to simply conflate good participatory research practice with qualitative research. There are examples of quantitative research that pride themselves on being participatory and co-produced, especially in the U.S. tradition of community-based participatory research and citizen science (Minkler & Wallerstein, 2011;Thornhill, Loiselle, Lind, & Ophof, 2016). Likewise, some qualitative research in an ethnographic tradition has relatively little interest in giving voice to the people it researches. This paper looks beyond a qualitative/quantitative discussion examining in more detail exactly how the public has been involved in evaluation of public policy and services and how this worked alongside the inclusion of outcome measures specific to the health and care field. Another consideration that moves us away from a qualitative/quantitative discussion is the complex, hybrid nature of health and well-being outcome measures. Although their domains are scored and quantified, the holistic design of domain types and tools used to gather these person-centred data do not sit neatly in a quantitative research camp.
The outcome measure discussed in this article is the ASCOT, which is designed to measure social care-related quality of life (SCRQoL). The Personal Social Services Research Unit (PSSRU) at the University of Kent developed the tool and makes it freely available for academics to use. Its emergence as the national standard in measuring SCRQoL, and the scope to compare ASCOT outcomes from an individual study against wider benchmarks, makes it an attractive and commonly used option for social care evaluation (Makai, Brouwer, Koopmanschap, Stolk, & Nieboer, 2014). The dataset allows regional comparisons of how well populations are being served to achieve SCRQoL (NHS Digital, 2017).
Many validated outcome measures have themselves been co-produced with the public. The acceptability of outcome measures within the overall research design has also been negotiated with input from the public for some research. The ASCOT measure had elements of public involvement and testing. The tool was developed by PSSRU with input from people with lived experience of care, and the ASCOT team has produced different versions of the tool to make it appropriate for people in different care settings. In this article, we use it as an exemplar of a validated outcome tool, and the tensions that arise when they are used in coproduced research, and not as a critique of ASCOT itself.

| COMBINING PARTICIPATORY RESEARCH AND OUTCOME TOOLS IN THE FIELD
From 2013 to 2015, Needham, Allen, and Hall (2016) led an ESRC-funded project on the relationship between organisational size and performance in adult social care. Eighteen people with lived experience of care were recruited as co-researchers, to undertake semistructured interviews and administer the ASCOT survey with people who use care services. These co-researchers were older people, adults with learning disabilities, and people who self-identified as carers. In total, 106 people who use services and carers were interviewed, and 95 of these completed the ASCOT survey. Ethical approval was granted by the national Social Care Research Ethics Committee (ref. 13/IEC08/0029).
The study incorporated an evaluation of public involvement, undertaken by two colleagues who had not been directly involved in other stages of the research process. Findings of the public involvement evaluation have been published separately and consider the methodological difference that a co-produced research approach makes to the interview content and process (Tanner, 2018).
The rationale for public involvement was both rights based and instrumental. We recognised that involvement of people with lived experience is an essential part of ethical social care research and a recognition of a historical legacy of "epistemic injustice" (Fricker, 2007). The instrumental rationale was to improve the validity and relevance of the research findings. According to Edwards and Alexander (2011, p. 274), "The 'privileged access' … afforded to peer researchers is a tool for rapidly generating trust and accessing information." They can draw on "acceptance, trust and empathy, which … is important in informing the construction of research knowledge." Rather than merely servicing research through data collection roles, the study aimed to encourage the co-researchers to have an active role in the interpretation of data and to communicate directly with research, health, and social care audiences. We aimed to align our approach with a wider concept of co-production in public services, where "individuals, communities and organisations have the skills, knowledge and ability to work together, create opportunities and solve problems" (NDTI & Helen Sanderson Associates, 2009, p. 3).
The research design also incorporated ASCOT, as a means of assessing the extent to which outcomes are realised for people who use care services. We recognised the value of standardised quantitative measures such as ASCOT, to enhance the reliability of our findings and allow them to be compared against those of others. Such tools also have a currency in the policy world, and we were keen that our research would be used by policy-makers and practitioners. There are a number of ASCOT versions available, such as self-completion questionnaires, interviews, and observation tools. This study used the interview version (INT4) that asks people to score themselves against four levels of need: ideal state, no needs, some needs, and high needs, relative to domains such as "how much control you have over daily life." Respondents also have to score themselves in a hypothetical scenario that services are taken away, so there is a way to measure the contribution that care services are making to meeting those needs.
The co-researchers led interviews with people using services, having attended 2 days of training in their local area. These peer-led interviews incorporated a semistructured qualitative interview phase followed by administration of ASCOT. The initial intention had been for co-researchers to lead both the qualitative interviewing part of the research encounter and the collection of ASCOT data with the academic researcher on standby at the back of the room in case the co-researcher wanted support. However, during the co-researcher, training it became evident that co-researchers were not comfortable about using ASCOT in addition to qualitative interviewing, for reasons discussed in more detail below. A compromise was agreed that the co-researcher would lead the main part of the interview, handing over to the academic researcher to ask the ASCOT questions.
Changes to a research design after piloting are of course common. However our experience and discussions with co-researchers suggested to us that there was something incompatible in the two approaches that we were seeking to bring together. Below, we draw on data from the evaluation of the co-researcher involvement to explore two explanations of why it was difficult to use ASCOT in such a context: The first relates to practical considerations, which potentially could be overcome with additional time and resourcing; the second relates to epistemological tensions, which are much more difficult to resolve.

| RESEARCH DESIGN
The case study presented here draws on data from the evaluation of public involvement and provides insight into how potentially competing evaluation requirements are negotiated by participatory research teams. Conceptually, the analysis is guided by the principles of decentred theory (Bevir, 2013). By exploring the perspectives of participants and areas of conflict and resistance, we observe how power and influence are diffused through the participatory research team, reflecting the broader philosophical position of decentred theory (Bevir, 2013, p. 1). The analysis reveals the competing experiences, knowledge traditions, and clashes experienced by co-researchers and academics. This single-case study is designed to gain an understanding of a complex issue by watching a phenomenon develop within a defined context. In this case, the complex issue is how well the use of a standardised outcome measure (phenomenon) fitted within a participatory study that sought to co-produce knowledge with people who used services (context).
The specific questions underpinning this analysis are What was the experience of using a validated outcome measure in a co-produced evaluation of social care? What tensions between elite forms of knowledge and local traditions and resistances are evident in the beliefs and actions of the participatory research team? This explicit framework provides clarity of ontological approaches within the data analysis.

| Data collection
The case study uses two sources of data: three group interviews with 15 co-researchers and interviews with three academic researchers. All interviews were undertaken by two researchers who had not been involved in the main study, as part of the study's evaluation of public involvement. All interviews took place after the overall study data had been collected and analysed, but whilst the project was still active. In this way, we aimed to ensure that reported experiences were as full and recent as possible. Interviews were semistructured. Topic guides were designed to highlight: how co-researchers felt about undertaking qualitative interviews and ASCOT survey interviews, whether academics and co-researchers took different roles, and what factors were perceived to help or hinder the research process. Using interviewers who were not involved in the main study was recognised to be important to give the co-researchers more freedom to be critical of the process.

| Data coding
The analysis protocol was theoretically based; however, this did not take the traditional form of an initial coding frame derived from literature (Miles & Huberman, 1994). Data were analysed in a way that gave individual accounts prominence. The use of a coding frame was rejected as it would distract from developing a clear impression of individual agency, an essential element of a decentred analysis. The codes were devised to remain aligned to their original narrative meanings. To ensure a narrative approach, "sequence and consequence" were considered in the formation of codes and themes (Riessman, 2005). We not only coded to abstract themes but captured the main "stories" being told. Analysis specifically aimed to capture what research events were selected and how they were viewed by the interviewees.
Thematic coding of all data from the co-researcher and academic interviews was led by the first author. Coding and theme generation were checked by the second author, and discrepancies were discussed and resolved. The resulting themes were refined through discussion within the academic research team. The researchers involved in the analysis of the data included some of the same academic researchers that took part in the interviews, whereas none of the co-researchers were involved in analysis and writing since the project had finished 2 years earlier. This raises issues of potential bias of academics' perspectives over co-researchers' perspectives in the overall analysis.
Berger's (2015) three step approach was used to reduce the effects of researcher bias. A process of acknowledging what has been said by a participant, examining your interpretation of this (as the analyst), and reflecting on your personal viewpoint about the statement was used by the academic researchers during coding and as a shared tool in team discussion. Quotes from the interviews were selected as exemplars of the eight stories set out below. Table 1 displays eight accounts that summarise the most prominent stories told by the co-researchers and academics about their involvement in the data gathering for the study. The collective presentation of narratives reflects patterns in the data and not a claim that all co-researchers or academics had identical stories.

Academic researchers
ASCOT is valuable as it can be linked to national datasets and financial data √ ASCOT surveys felt awkward to administer and sometimes inappropriate √ √ ASCOT surveys were not always effective for collecting useful data (especially compared to the preceding qualitative interviews)

√ √
There was not enough time for training around ASCOT √ √ It worked better for the academic researcher to undertake the ASCOT questionnaires √ √ The qualitative interview often answered the ASCOT questions in advance √ Academic researchers led decisions about how ASCOT could be best managed √ √ Only academic researchers could analyse the ASCOT responses √ √ Findings reveal that we, the academics, and the co-researchers shared most of the narratives in relation to ASCOT. More details of the findings are presented below, divided into two sections. The first brings together storylines, which emphasised practical barriers to the effective involvement of co-researchers in the ASCOT process.
The second looks at issues around the incommensurability of participatory research and outcome measures. Although the research team talked about practical issues and tensions, the findings also display how academic and coresearchers were instrumental in finding ways to make the combination of two different research methods work in practice.

| Practical constraints
This theme includes narratives: 1. "There was not enough time for training around ASCOT." 2. "Only academic researchers could analyse the ASCOT responses." 3. "It worked better for the academic researcher to undertake the ASCOT questionnaires." 4. "Academic researchers led decisions about how ASCOT could be best managed." A strong theme in the interviews was inadequate time for training. The conceptual nature of the SCQoL measure is designed to accommodate the breadth and complexity of social care generally. In ASCOT, this complexity manifests in multiple domains designed to capture different aspects of well-being and the different versions of the tool for various care settings and beneficiaries. We planned for 2 days of dedicated co-researcher training at each site, based on replication of a training model used in a similar successful participatory evaluation study. The duration of the training felt proportionate to the amount of work involved in the project (around 10 interviews followed by analysis). However, this limited the amount of material that could be covered.
The challenge of fitting all of the elements required into the time allocated for training was acknowledged across the academic and co-researcher interviews. Additional planning and delivery time was required to include coresearchers with learning disabilities, and as this quotation illustrates, the results were valued by co-researchers: Academic researchers were similarly aware of the significant time investment that would be needed to train coresearchers to understand and use specialist tools-which we as the academic team had received separate training on before the study had started. This academic researcher's account frames the problem in terms of feasibility. The academic researchers and co-researchers agreed that the academics would lead the ASCOT survey data collection. In co-researchers' accounts, reasoning revolved around a preference for a freer style of peer interviewing,

The training was an introduction to all of this and I thought it was very resourceful because
where their own way of wording questions was seen as important. Co-researchers' preference aligned with the agreed purpose of their involvement, to bring in the perspectives of people that used services and enhance rapport during data collection. The demands of the qualitative interview meant they expressed relief about sharing data collection with academics. Both accounts are visible in this co-researcher comment:

I personally found it quite draining doing the interviews because you're getting a lot of emotional stuff from people as well …, and I was relieved at the end of my part for [academic researcher] to do the ASCOT questions and to finalise the thing. (Co-researcher, older person, site 1)
This co-researcher's account provides one example of how we overcame practical constraints in ways that played to the strengths and needs of our research team members. As academics led the ASCOT surveys, coresearchers could focus solely on the often demanding role of being qualitative peer interviewers. It also offered a structure where co-researchers could hand over smoothly to another member of the team. In a similar way, the accounts of the academic researchers raised the benefits of that interviewer handover as an opportunity to capture extra data, revisiting areas that had been raised in the peer interviews and "ask some questions without interrupting the co-researcher."

| Epistemological dilemmas
This theme includes the narrative accounts: 1. "ASCOT surveys were not always effective for collecting useful data (especially compared to the preceding qualitative interviews)." 2. "ASCOT surveys felt awkward to administer and sometimes inappropriate." 3. "ASCOT is valuable as it can be linked to national datasets and financial data." 4. "The qualitative interview often answered the ASCOT questions in advance." The practical problem of combining the qualitative peer interviewing with the collection of ASCOT data could have been eased in a number of ways: for example, by increasing co-researcher training or by a repeat visit to the home of the interviewee for ASCOT data collection on a different day. A more profound limitation was the difficulty of gathering data, which seemed to prioritise different forms of knowledge, which were difficult to combine. Our case study highlights how those involved in data collection came to see participant narratives of experience as incompatible with standardised outcomes. This incompatibility is in part about the two methods producing qualitatively different data about the same topic and in part that these different accounts have no straightforward way of being combined in an overall analysis. The parallel-track rather than consolidated nature of the findings led us to question whether this is the most effective way to evaluate the impact social care services have on those who use them.
We as academic researchers and the co-researchers in our team were very aware of the contrast between the peer-led qualitative interview and the administering of the outcome measurement tool. Being "relaxed," "at ease," and "less daunting" were the ways that co-researchers typically described the interaction in the peer interview part of the session. This co-researcher describes the contrast with administering a survey:

You were speaking to the person in plain English and there was no gobbledegook for you about a word.
They understood all the questions, you got proper answers from them, that made them relaxed and then, you know, it made you at ease. (Co-researcher, site 3) Co-researchers gave less favourable accounts of ASCOT as a research tool than the peer-led qualitative interviews. ASCOT was seen to collect less rich data, to disengage participants and to cause a general atmosphere of formality associated with a lack of productive rapport. Co-researchers often found the ASCOT questions difficult to understand and contextualise. In some cases, this appeared to be about the questions being abstract in nature, making them feel complex to follow and to give a score. One co-researcher with learning disabilities described her inability to relate to ASCOT questions. It was anticipated that the ASCOT survey would be less flexible than the qualitative interview questions, and this was not viewed as a problem in the initial study design. The two research methods were seen as complimentary, with peer interviews generating in depth understandings of the value of social care services and ASCOT providing a score for this. It was during the practice of undertaking data collection that researchers started to lose confidence in whether ASCOT questions were being interpreted correctly and generating valid responses. This challenge was largely due to the juxtaposition of the two research methods, leading researchers to make stark comparisons of the relative success and ease of communication between the methods.
It was problematic in that they were very structured, rigid questions, which came at the end of what were usually quite narrative conversational interviews. (Academic researcher, site one) In some cases, the co-researchers suggested that the ASCOT questions limited effective rapport. One coresearcher highlighted the ASCOT element as being "a bit sterile." Another co-researcher suggests that the "formal" nature of the questioning shuts respondents down to the extent that interviews would have been less successful if the encounter had begun with the ASCOT survey.

(Co-researcher, older person site 3)
The interview data showed that we, the academic researchers, felt that ASCOT could be usefully aligned with national data or combined with financial data to add value to findings. However, these positive accounts tended to be followed by contrary negative points, as visible in the following statement.
One of the potential epistemological dilemmas identified was that ASCOT questions covered similar topics to the preceding interviews but framed the questions in a more formal way. Academic researchers acknowledged this tension but also felt there was scope to use the initial qualitative conversations to facilitate better informed ASCOT scores. Co-researchers often helped to elicit background information that could be used to contextualise an ASCOT question, which had been difficult for an interviewee to understand.

Sometimes when you listened to the peer interview they answered those [ASCOT] questions exactly and
that gave you a better idea … Like 'how often they are enjoying activities? What difference the service makes?' (Academic researcher, site 2) These pragmatic research practices perhaps suggest the potential for combining research methods and beginning to overcome the epistemological tensions raised. However inferring answers from qualitative data does not offer the fidelity to the research gathering protocol that outcome tools require if they are to be aggregated and comparable.

| DISCUSSION
This case study analysis illustrates how co-researchers championed the benefits of qualitative approaches for enhancing communication and capturing rich data. Their justifications echoed the broader sentiment around the strengths of qualitative approaches, that is, "seeing the world from the point of view of the actor," increased "contextual understanding," and greater likelihood of "discovering the novel or unanticipated" (Bryman, 1984, pp. 77-78).
Co-researchers' tendency to associate limited value with the outcome tool may be the result of their lack of exposure to how outcome measures work and how findings are used once research studies end. On one hand, this signals that a potential resolution may be to include co-researchers more fully, for instance, by providing more substantial methods training. This option raises the professionalisation paradox "we need to train lay people to enable them to participate in research cooperatively, but through training they become professionalised and therefore unable to be 'lay' " (Ives, Damery, & Redwod, 2012, p. 3).
As our findings demonstrate, the academics and co-researchers made a decision collectively not to cover ASCOT methods at length. The decision was steered by consideration of the overall purpose of public involvement in this particular study and the time limitations of the training. Co-researchers were encouraged to develop the skills they had the most interest in and in doing so retain the lay perspectives that the study valued. These findings do not necessarily indicate that qualitative methods will be the public preference in other participatory studies. In part the findings display the way the co-researcher's role had been explained and understood by the whole research team (to incorporate the unique perspectives of members of the public with experience of services) and how this had influenced co-researchers methodological preferences.
The deeper qualitative-quantitative friction at the centre of these findings is the divergence between a concept of QoL captured in outcome measures and the conception, which was gained from more qualitative peer-led accounts. These findings bring specific insight from public involvement in health and social care evaluation to a broader undertaking already underway: overcoming philosophical and disciplinary tensions in pursuit of the highest quality and most useful applied mixed method studies (Brannen, 2017;Cooper, Glaesser, Gomm, & Hammersley, 2012). There is nothing intrinsic to quantifiable research tools that would prevent the public from using and valuing them, but placing them alongside a more natural, conversation-based approach to the same research question caused a tension within this participatory study.
Co-production is profoundly relational, and participatory research teams must navigate a context of power relationships between research funders, academics, and co-researchers. In this case study, researchers met problems when they tried to combine participatory and outcome approaches but overcame them through organising appropriate data collection roles, creating additional spaces to collect data, and ongoing dialogue about the nature of the data.
These examples of constructive research practice suggest that although an appreciation of the clashes between knowledge traditions is vital to anticipating practical issues, there may be routes to develop complimentary participatory and outcome-based approaches.
Combining outcome tools with more qualitative methods requires careful consideration, mindful not to compromise the recommended process for producing outcome measures or to dilute the depth of qualitative insights.
Decision-making audiences who use outcome measures at face value may scrutinise the practice of research teams as to whether their actions have invalidated the scored outcomes, for instance, by inferring answers from more qualitative data collection, rather than depending completely on the prescribed data collection protocols. A conflict of goals arises here as evaluation teams aim for their findings to be used but might at the same time see opportunities to make responses more genuine by drawing on richer data.
Data from this case study were collected before the dissemination of project findings, limiting what can be reported about which findings were most used. This research impact activity is important in the assessment of power relations within co-produced research studies. The quantification of knowledge and the state's use of "elite research" indicators to inform public policy are acknowledged forms of statecraft (Miller, 2005, p. 406). The power and politics of public service commissioners' use of outcome measures gains frequent attention. Reflecting on the use of patientreported outcome measures (PROMs), Kyte et al. (2015, pp. 5-6) acknowledge that PROMs have primarily moved into the everyday clinic environment because of "the political context and the demands of commissioners for evidence of the quality care services provide." The dominance of commissioning groups is equally relevant for social care outcomes. Outcome measures are used predominantly by one type of social care stakeholder (strategic decisionmaking audiences) yet have become a central knowledge output in social care evaluation. Gibson, Britten, and Lynch (2012, pp. 531) frame the emancipation of patient and public involvement as requiring "knowledge spaces" that "support the emergence of social networks of knowledgeable actors capable of engaging with professionals on equal terms and influencing service provision." Specialist outcome measures, and their perceived authority, threaten to further disrupt the balance of lay and expert knowledge in co-productive research spaces. A broader understanding of where knowledge tensions arise and how they can be overcome is essential to moving towards equality of contributions, the conditions essential for co-production of research across social policy, research, and evaluation.
The question of what evidence is valued in social care evaluation is much discussed (Rubin, 2015). It forms part of broader debates about the privileging of elite narratives over local forms of knowledge (Bevir, 2013). In this case study, we saw that co-researchers valued interviewee's voices and preferred the qualitative data collection that captured these most fully. Glasby and Beresford (2006, p. 268) argued that "the lived experience of service users" is too often overlooked in favour of more "objective" evidence. Inclusive approaches, such as those currently being developed by the U.K. Social Care Institute for Excellence, aim to reduce tensions between evidence traditions in the governance of social care by ensuring a broader range of evidence for economic evaluation and the systematic inclusion of service user and practitioner knowledge (Fisher, 2016). However our research warns of problems if mixed method approaches seek only to merge the two approaches rather than exploring their incommensurability. The easiest way to combine the approaches is to undertake them in separate work packages, but this leaves unexplored the tensions between the two.
Alongside the epistemological tensions, findings draw attention to a set of practical tensions. There were clear feasibility issues including not enough time and capacity to train the public to understand, use, and analyse outcome measurement tools. The specialist nature and complexity of SCQoL conceptually was one significant factor voiced by co-researchers, and the additional time and skill required to support people new to research or with learning disabilities was another. Findings demonstrate how traditional expectations for evaluation timescales can underestimate the work required to involve service uses in all aspects of research, on an equal basis with academic researches. Additional time requirements for participatory research and how academics perceive this as a risk are well documented in participatory research literatures (Blumenthal, 2011;Tuffrey-Wijne & Butler, 2010). This analysis outlines how these issues can be amplified in the field of health and care evaluation when attempting to incorporate specific outcome measures and specialist tools. Findings about power dynamics echoed those of other participation case studies, as nonacademic researchers experienced exclusion from certain aspects of the research due to their inability to engage with technical data (Crompton et al., 2017, p. 15).
Scaling these mismatched timescales and power dynamics up, we could ask how and why traditional research structures are misaligned with the realities of participatory approaches and what would be needed to rebalance this?
Organisations that fund research are still developing an appreciation of how public involvement may alter the scope of studies and timescales for delivery (NIHR, 2017). Learning from complexity science suggests that "emergence" (e.g., innovations in research questions, design, and dissemination), a desirable feature of participatory research, remains difficult for research commissioners to manage. Cox (2012, p. 587) observes that "… it is difficult to advise funders or commissioners that bids should be considered fluid.
Within our co-produced research, it is interesting to note that when tensions and challenges did arise, power over the research returned to its traditional origin, with us, the academic researchers. Academics felt responsible for directing and taking charge of the processes that were not working. Although their actions often sought to challenge conventions in order to include co-researcher perspectives more effectively, the tendency for the academic team to lead on key methodological decisions disrupted the "co" in this co-research study.
These findings intend to contribute to and open reflection of academic power and control within co-produced research, a necessary approach for improving academic endeavour in this challenging and discursive space. Restitution of academic power is also evident in the process of analysing the data presented here and writing it up as an article. Although co-researchers were involved in analysis, dissemination, and co-authorship of outputs from the project (Brant & Stevenson, 2014;Murray, Welford, & Allen, 2015), we did not have a way to resource the ongoing involvement of co-researchers in the writing of this article.

| CONCLUSION
This article has explored tensions within social care evaluation in which two dominant paradigms (participatory research and quantified outcome tools) are expected to coexist. Rather than compartmentalising them into different work packages, our project sought to integrate these two approaches into a genuinely mixed methods research design. The practical and epistemological issues set out here highlight the difficulties of doing this effectively. There are lessons here for researchers in planning and resourcing the training of co-researchersalthough we question whether the co-research paradigm is ever likely to have sufficient time and resource to equip co-researchers with the specialist skills, they may need to engage with more formal research tools. More plausible resolutions may result from all stakeholders gradually redeveloping their approach to knowledge, validity, and learning.
Epistemological issues are raised for academics and for research commissioners in exposing the limits of a naïve approach to "having your cake and eating it" within a single research study. The growth of more formal evaluation tools at the same time as the mainstreaming of patient and public involvement has created tensions within research projects, which are not always discussed, because academics may feel that they ought to be able to resolve these tensions themselves. Researchers, keen to access funding for their work, risk colluding with an agenda that promises the best of both worlds, without adequately critiquing the viability of this dual task. Problems and failings in research design are often hidden in end of project reports, which focus more on research findings.
Our original contribution is the finding that co-produced research is difficult to combine, both practically and epistemologically, with the quantified outcome tools that are increasingly utilised in evaluation research. The risk that public participation is tokenistic is a familiar refrain in academic writings about co-production. The broader intention of this publication is to prevent this tokenism by adding to evidence that builds a more detailed understanding about undertaking co-produced research. The contribution of this article is methodological and empirical, highlighting the experience of undertaking fieldwork with a co-research approach and contributing to the emerging literature relating to participatory approaches. The sample size of co-researchers and academics is small, and we make no claims to statistical generalisation but rather seek to disrupt the sorts of knowledge claims that can be made within research studies using these two approaches. The future research agenda is to explore whether it is possible to address the