Integration, discussion and conclusions

Introduction

Research and researchers represent a rich source of empirical, theoretical and methodological knowledge that can help health and care organisations with some of their most pressing challenges. There is, however, a persistent and well-documented disconnect between such research-informed knowledge and the arrangements for everyday care. Addressing this has challenged policy-makers, academics and practitioners for several decades.180 As a result, many parts of the health service are perhaps not as informed by up-to-date research-based knowledge as they might be, and there are many, as yet unanswered, questions that could be addressed by careful study.

In many health systems, considerable money and effort have been invested in the creation, synthesis and promulgation of clinical effectiveness evidence (e.g. through NICE guidelines), but health services research is far broader than this. It encompasses, for example, quantitative and qualitative data on health needs, evidence on the lived experience of health conditions and the care provided, and theoretical and empirical insights into the costs and consequences of particular ways of organising and delivering services. This diversity of topics, methods and theory, alongside the often contextualised nature of interpretation, makes the ready translation and transfer of knowledge from health services research especially problematic.

In consequence, many strategies have been developed to try to improve the connect between health services research (in all its diversity) and the health system, or components thereof. This project, Embedded, has explored, in depth, one of these approaches: the embedding of experienced researchers in service teams for sustained periods, with the explicit goal of creating new and actionable knowledge in context. Central to this is the locating of embodied research expertise in close physical proximity to the sites of action for research-informed knowledge (‘researchers-in-residence’).

In this concluding chapter, we first locate the embedded approach in the context of other strategies aimed at increasing the usefulness of health-related research (and the literatures surrounding these), and we outline the arguments underpinning the embedded approach specifically. We then revisit the aims and objectives of this project, noting progress towards those goals and any areas of deviation from our original plans. Subsequent sections reflect on the findings from this programme of work, highlighting the key insights and contributions, before drawing out some of the implications for future developments in embedded research. In outlining the contributions of this work, we have been guided by the framework for reporting new knowledge about how to improve health care, Standards for QUality Improvement Reporting Excellence 2.0 (SQUIRE 2.0).181 The importance of PPI is also brought to the fore, together with an analysis of its often rather limited application in embedded research initiatives to date. Hence, we make some suggestions for more active involvement in the embedded schemes of the future.

We begin by exploring the range of approaches used to address the knowledge–practice gap that emphasise closer engagement between research and service actors, before homing in on the rationale for embedded research: knowledge co-production.

Promoting better research and better research use

Making best use of health services research is about getting the right research done (work that focuses on service-relevant issues) and getting the research that is available to have influence and impact through changes to services for the improvement of care. Both of these require close collaboration with service personnel and service users if they are to be achieved.

Many approaches to these challenges have been explored (Box 4). Some have been about strengthening practitioners’ research skills, building local research capacity and developing greater research awareness.182,183 Others use brokers or intermediaries to bridge the worlds of research and practice, realising that active processes of facilitation work better than passive dissemination.3,182 More recently, we have seen initiatives putting patients and service users at the heart of research teams,70,184 and the development of research–practice partnerships, which can exist in many guises.183,185 Finally, taking the idea of ‘sustained interactivity’ further,9 we find initiatives that seek to ‘embed’ researchers in service settings, where they can broker external research-related knowledge, create and co-produce new knowledge in situ and build the capacities to negotiate and apply such knowledge to improve care.158,184,186

Many structural innovations have emerged as a means of addressing the disconnect between research and practice, including:

• Embedding research skills – research awareness has long been a part of the professional clinician’s repertoire, and many initiatives have sought to bolster research skills in the clinical setting by, for example, support for clinician–researcher hybrid roles.173
• Brokering expert knowledge – a variety of agencies and individuals have emerged that actively seek to facilitate the importation of research findings and evidence summaries into the care setting.3,173
• Developing locality-based R&D units – locally situated collaborative research infrastructures can help support better targeted applied research and undertake brokering activities.176
• Supporting research–practice partnerships – sustained collaborative partnerships between academic institutions and service delivery organisations can allow shared problem-solving and greater receptivity towards research findings.174
• Creating learning health systems – building analytic capacity into service teams so that they can access and apply evidence and better exploit insights from routine data has been promoted to spur care transformation.175,177
• Bringing patients centre stage – using patients as active partners and participants in research teams has a strong and growing tradition in ensuring the relevance and applicability of applied research findings.70
• Embedding researchers for knowledge co-production – situating experienced researchers in service settings as active members of the service team is a relatively novel, yet growing, approach to addressing the research–practice disconnect.158,175

Although each of these models attempts to mobilise knowledge by closing the gap between the researcher and practitioner communities, there are also many differences, for example the individuals involved (e.g. researchers, health professionals, managers, service users); the nature of the interactions and the settings where these take place; the nature and extent of collaboration, and the type of activities that support this; and the character of the boundaries that need to be bridged (cultural, organisational and epistemic). Each of these can vary hugely, and their details are rarely well described in accounts of various models.

Partnerships are central to most of these models, but the structures and mechanisms for promoting and supporting such partnerships can also vary. Dimensions of effective research partnerships that have been highlighted in the literature include a sharing of power and influence within the partnership, relationships based on strategic programmes rather than individual projects, the desirability of working on common areas of interest, the need for early demonstrable advantage to the partners to justify their efforts, the importance of demonstrating a genuine commitment to partnership over time, the importance of researchers valuing a broad notion of ‘knowledge’ and ‘evidence’, and the need for practitioners to value scientific evidence.187–192

Much of the supporting literature57,193 has focused on the attributes needed for partnership roles and on the consequences for individual post-holders. Building on this, reviews by Bullock et al.12 and Phipps and Morton194 have considered the characteristics and capacities of individuals who seem to be effective at working in partnerships across the academic–practitioner boundary. These include good communication and negotiation skills, flexibility, the ability to engender trust, entrepreneurship, creativity, and a sensitivity to context.

Overall, many different models for better connecting research and practice through partnerships have emerged, and there is a need to deeper explore how these are put into operation and their potential. What has been published to date mostly comprises descriptions of specific initiatives or wish lists for the personal attributes of their protagonists, but little in the way of support for more structured analysis across schemes or practical guidance about set-up and management. There is a clear need, therefore, to develop better theoretical and empirical underpinnings for the different models, and to use these to help the construction of practical tools and resources.32,38

Alongside these varied approaches, and the accompanying research effort to tease out their roles and contributions, we can see a growing number of local initiatives that deploy embedded researchers. The distinctive features of this approach are elaborated on in Box 5.158,184 For those who judge the conventional research approach to be a distant and tardy process, insufficiently focused on utility, embedding researchers is intuitively appealing; consequently, the approach is growing in popularity in many health sectors. Indeed, we found around 50 such schemes in the UK NHS (see Chapter 3). Nevertheless, even prior to this most recent work, it was clear that challenges remain, chief of which were the problems of dual identity or affiliation, the potential for conflicting objectives between multiple institutions and the challenge for the embedded researcher of retaining a ‘critical’ academic perspective.158

Embedded researchers, or ‘researchers-in-residence’:

• are affiliated to both an academic institution and a service-based host organisation, spending considerable time in the latter
• try to develop long-term relationships with staff in the host organisation, and are seen as members of the service team capable of bringing new expertise and experience to group discussions
• import and interpret research-based knowledge and/or co-produce with local teams new research-based knowledge that responds to the needs of the host organisation
• help to build research capacity and capability in the host organisation.

Recognising that embedded researcher initiatives were gaining traction in the UK and elsewhere, this project placed such schemes at the centre of its research focus. Before recapping on the study aims and objectives, and then setting out the key findings of the work, we first briefly rehearse the logic of knowledge co-production that lies at the heart of such embedded models.

From knowledge transfer to embedded co-production

The increase in interest in embedding researchers reflects, at least in part, a recognition of the importance of ‘co-producing knowledge in context’ as a means of crafting more applicable research knowledge that is more likely to be actioned. Rather than research and practice being construed as separate processes operating sequentially (research fed ‘into’ practice), the production and utilisation of research evidence are seen as complex, interdependent and situated activities (research ‘in and with’ practice).11 The emphasis is on dialogue about how research-informed knowledge is shaped and understood, on encouraging productive ongoing relationships, and on creating environments conducive for change.14

This reframing suggests that researchers need to be willing to actively negotiate knowledge among managers, practitioners and (crucially) patients and service users; they need to recognise and accommodate power differentials in who dictates how knowledge is defined; and they must adopt a pluralistic orientation to knowledge to achieve change.26,195 Co-production models of knowledge mobilisation are based on the assumptions that research knowledge usually needs to be adapted if it is to have impact, and that all participants need to be involved in its creation, interpretation and use.

Participatory approaches, such as co-production through embedded research, attempt to mobilise knowledge by encouraging greater interaction and dialogue between researchers, practitioners and people who use health services. Beyond these basic principles, there is great diversity over, for example, the individuals involved, the academic interests or disciplines engaged, the level of embeddedness of individuals in the practice setting, the structural arrangements for such embedding, the nature and degree of co-production, and the type of activities that take place. Until recently, there has been little in the way of systematic examination of either the nature or the impact of the different approaches to embedding, and this study was an attempt to address such knowledge gaps.

Study aims and objectives

As detailed in Chapter 1, the Embedded project aimed to increase the influence of health services research on decisions about the improvement and redesign of NHS services by:

• developing insights and understanding about the nature, challenges and effectiveness of co-production initiatives in which researchers are embedded within service settings
• producing practical guidance on the design and implementation of embedded models of co-production for managers and clinicians in the NHS, their academic partners and people who use services.

In addressing these aims, the project focused on the following specific objectives:

• to review the theoretical and empirical health services, management and organisational literature relevant to embedded research initiatives and knowledge co-production, and identify the relationship(s) between the two (workstream 1)
• to gather examples of embedded models in operation around the UK’s health services and public health sectors, focusing on examples where embeddedness and co-production co-exist, and to describe the features of these models, including their history, context, participants, scale, scope and content (workstream 2)
• to undertake in-depth case studies in four of the examples identified, in order to understand their mechanisms, effectiveness and challenges (workstream 3)
• to provide resources aimed at assisting the recruitment of embedded researchers, alongside recommendations and guidance for their training and development, customisable for the different ways in which embedded co-production may be framed and specified, to allow those interested in developing and using such approaches to understand the design choices they face (workstream 4).

Progress against objectives

All of the aims and objectives as set out previously were met in full, with one exception covered shortly. Chapters 2 and 3 provide a full account of the two literature reviews (on knowledge co-production and on embedded research initiatives, respectively). The findings from the scoping review of UK health-related embedded research initiatives were used to augment and enrich the literature-based accounts of such schemes, and, as such, feed into the derivation of the 10 domains that were used to characterise embedded initiatives (see Chapter 3). The four in-depth case studies are presented descriptively in Chapter 4, and analytically in Chapter 5. Finally, findings from across all four workstreams are brought together in Chapters 6 and 7 to provide the promised practical guidance on the design, implementation and management of embedded initiatives in UK health settings. Subsequent sections in this chapter reflect on these findings in more depth.

The one aspect of the overall aims to which we were able to make only indirect contributions was on the issue of scheme effectiveness. As expected, published data (in the literature reviews) and systematic analyses (in the scoped extant initiatives) were very scant. Thus, we were much more able to provide rich accounts of the nature and challenges of embedded co-production than we were to provide robust evidence on their effectiveness (or otherwise). That said, the more structured ways of describing and analysing embedded co-production enabled by our analysis point to many areas where scheme rationale and coherence can be assessed, as well as contributing to the capacity for shared goals and understandings across diverse stakeholders. Such indirect assessments of effectiveness are likely to prove of great utility while we await further robust studies of scheme outputs, impacts and outcomes.

Deviations from the original proposal

Deviations from the funded protocol were minimal, and all workstreams unfolded largely to plan. There were some minor methodological adjustments to the case study data gathering, including less reliance on diary methods, but these were inconsequential (see Chapter 4 and Appendix 3). It also became clear that the extent of knowledge co-production in the embedded schemes was often difficult to discern, and so our focus tilted more towards embeddedness per se, rather than co-production (although obviously these two are often linked). Finally, there were some adjustments to the tools, resources and templates developed to support practical scheme development: recognising the sheer diversity of schemes in operation and planned, these now provide guides and customisable resources rather than definitive items (see Chapters 6 and 7).

One area where progress was limited was our goal of creating an active network for those engaged in developing or running embedded research initiatives. In general, this proved relatively muted in operation, for what appeared to be two main reasons: first, the sheer diversity of approaches to embedding researchers sometimes precluded shared interests, terminology and concerns; and, second, potential network participants consistently noted that interactions depended largely on saliency (i.e. engagement was sporadic depending on what participants needed to know at specific points in time). Despite initial enthusiasm from participants, and several attempts to create and facilitate the network, including experimenting with different technologies (e-mail discussion lists, Google Groups and WhatsApp chats), it never reached self-sustaining status.

The only other significant deviation from the original proposal occurred towards the end of the project (March 2020), when some of the engagement activities (notably, the final workshop) had to be reworked in the light of the COVID-19 crisis. However, it proved possible to achieve the aims of these activities through other means, for example by tapping into professional networks, having virtual conversations with the more active members of the embedded network and attendees at prior workshops, and drawing on the materials produced at workshop 2. Hence, the COVID-19 disruptions did not hamper the overall achievement of the project objectives.

Reflections on the findings

In the following sections, we reflect on the findings in the light of the aims and objectives as set out previously. The two literature reviews (augmented by data from the scoping review) provided clear and structured language that has allowed us to describe and disentangle diverse approaches to both knowledge co-production and the nature of embedded research initiatives. In addition, the scoping review provided a snapshot of the landscape of health-related embedded research initiatives in the UK in 2018, and contributed to a deeper understanding of their design contours and operational challenges. The in-depth case study work that followed, in four divergent examples of embedded researcher schemes, provided additional insights into the dynamics and life cycles of such initiatives. Finally, integration between the emergent findings from each strand of the research and the engagement and influencing work of workstream 4 led to a range of tools and resources for embedded scheme designers. Each of these sets of findings is now elaborated in turn, with a concluding discussion on the role of PPI in embedded research initiatives.

A language for knowledge co-production

In recent years, attention has turned to how closer relationships and partnerships can be assembled that enable researchers and non-researchers (including policy-makers, practitioners, service users, carers, the public) to create knowledge collectively. For those crafting policies, designing and managing services and delivering care, co-productive relationships around knowledge offer the prospect of more relevant, timely, contextually sensitive (and, therefore, usable) knowledge directly applicable to organisational needs.33,196 For researchers, knowledge co-production is seen to offer easier access to research sites; stronger and more effective research collaborations; more engaged scholarship;167 and greater engagement, influence and reward for their work. In addition, there is the possibility of achieving and evidencing ‘impact’ with academic work, neatly fitting with the increasing rhetoric and use of metrics around this topic.197 For similar reasons, knowledge co-production also clearly holds some attractions for funding bodies.127

Co-production, then, has risen quickly to prominence as an approach to knowledge-making. Promising to address many well-worn problems of creating actionable knowledge that is fit for purpose, it is invested with great hopes, alongside some emerging fears and disputation,82,198 across many disciplines. Our review and synthesis of the burgeoning literature (see Chapter 2) offers a means of exploring and disambiguating the conceptual underpinnings in a way that also provides practical tools and resources for collaborative conversations.

What we found, from reading across the literature and developing a framework analysis (see Chapter 2), was that issues and approaches in knowledge co-production could be thought of in five main domains of meaning: politics, knowledge, identity, space–time and aesthetics. Drilling down, each of these domains yielded two subthemes to provide a finer-grained analysis of the issues. Furthermore, we found that each of these domains could be overlaid with a graded spectrum of co-production ‘styles’, from relatively conventional research processes, through provisional, moderate and committed, to, finally, radical knowledge co-production. The exploration of meanings in each of these domains and across these spectrums, and the surfacing of very divergent exemplifications of co-production, provide some clear conceptual guidance and more precise language to aid in the disambiguation of both the rhetoric and the practice of knowledge co-production.

Disentangling distinct conceptual concerns in this way will, we hope, lead to further discussion and insights on the nature and role of co-production processes, helping to prompt more focused empirical and evaluative work. More pragmatically, we hope that the framework we have presented will be used by those involved in embedded initiatives (and, indeed, in knowledge co-production achieved through other means) to articulate their understandings of their own and each other’s approaches. This, in turn, may help to clarify, and support discussion of, the implications of their respective stances, and so facilitate better communication around the difficulties and challenges of such collaborations.

As well as aiding conceptual clarity and being of practical use, our analysis moves to forestall a growing risk that knowledge co-production could become unattractive as a result of its complexities, ambiguities and difficulties, which can all be germinal points for misunderstandings, disappointments and disillusionment.198 Debates about knowledge for practice and policy are, we think, well rehearsed,4 and new grounds for conversation are needed if we are collectively to create a more progressive and enlightened culture around knowledge production and use. In short, we must be able to make the most of different possibilities for co-producing knowledge, and the first step to capitalising on them is to find ways to think and speak about them with greater clarity.198 The work presented in Chapter 2 begins the task of arraying these different possibilities so that they can be systematically and transparently explored. All of this has great relevance to embedded research initiatives, many of which have co-production at their heart (at least rhetorically, if not always so obviously in practice).

Crucially, our work in the review and synthesis was not an attempt to solve the problems of co-production, or even necessarily to promote its application (although it is clear that we think it has some potential). In exploring the growing literature, we were acutely aware of the frequently aspirational and rhetorical uses of language in this field. During our analysis, we sought to lay out constellations of ideas and themes, with their associated tensions, overlaps and consequences, without lapsing into judgement or advocacy. The value of this analysis lies, we believe, in its capacity to provide a useful framework for more thorough conceptual work and better communication over practices in this area.

Overall, knowledge co-production can take many and diverse forms, from approaches more akin to conventional collaboration to those with much more radical intent. To date, many embedded research initiatives do not appear to have fully understood or engaged with this diversity and there may be advantages in doing so using the conceptual groundwork we have laid out in Chapter 2.

The landscape of embedded research initiatives in the UK

Embedding researchers in service delivery, public health or service commissioning settings is a key means of developing knowledge co-production. Knowing that embedded research initiatives (or ‘researcher-in-residence’ schemes) were springing up in UK health settings made an analysis of such schemes a timely proposition. The combination of literature review and scoping of extant schemes allowed us to develop a bounded account of what constitutes embedding and a structured way of describing the components of such schemes (see Chapter 3), even when the contours of these components in practice are more implicit than explicit (such as addressing power dynamics).

Reading across the 45 embedded researcher schemes uncovered in the UK (those that met our criteria as set out in Chapter 3), a number of features of this landscape became clear. First, schemes were in place across the UK, and were located in primary, secondary and community settings. It was clear that this approach to the knowledge–practice puzzle held some wide and enduring attractions. Second, the scale of embedded research initiatives was highly variable, ranging from single short-term projects to longer-term programmes of work or strategic partnerships, sometimes involving multiple embedded researchers. Relatedly, some of these initiatives were funded for fixed periods, some had secured sequential funding as they evolved, and others were in receipt of more stable ongoing support. Third, around half of the schemes used university-held contracts for their embedded researchers, around one-quarter used NHS contracts, and joint appointments were relatively unusual (only about 10% of schemes). Finally, it was clear that some initiatives were carefully planned from the start, but many more were evolving and emergent, with changing intentions, structures and processes.

The formalised accounts in the published literature (which tend to focus on a relatively established or successful subset of schemes) were usefully enriched by the intelligence gathered on the wider hinterland of schemes in the UK. From the scoping work, we gained sight of internal scheme documents, and interviewed embedded researchers and scheme managers, to gain insights into the intricacies of scheme design and delivery. This work, when combined with the literature review, led to the exposition of 10 domains (grouped under the headings of scheme intents, structures and processes) that provided a comprehensive way of articulating the contours of embedded researcher initiatives (see Chapter 3). What was clear from this analysis was that there were many different ways of constructing embedded research initiatives, and that there was real diversity of approach across the field, as well as a frequent reshaping of approach in ostensibly a single initiative over time.

This research provides then, for the first time, a robust theoretically and empirically informed tool to describe and analyse embedded research initiatives. Such a tool can be used for various purposes: for research, to delineate embedded researchers as interventions that can be compared and evaluated; for practice, as a design aid to the development of new schemes (see the extended account of such a development in Chapter 6, and later discussion in this chapter); and for management of extant schemes, by aiding dialogue between stakeholders. It was also clear from the later case study work that significant tensions could arise between and within stakeholders in embedded research initiatives; the tool described here can provide a useful framework for exploring (if not always resolving) such tensions.

Finally, the work in understanding the landscape of embedded research in UK health settings underpinned the workstream 3 case studies in various ways. It aided the identification of potential study sites: the scoping work enumerated all the sites known to exist and provided sufficient information on the intent, structure and processes of those sites to allow for purposeful selection (see Chapter 4). Moreover, the 10 domains surfaced through the literature review and scoping interviews provided a means for structured engagement with stakeholders: shaping the interview schedules (see Appendices 4 and 5), the use of the partner identification tool (see Appendix 1) and the focus of site observations (see Report Supplementary Material 4). Ultimately, the data analysis of the case accounts (both descriptive and analytic) drew on and expanded the insights articulated in the earlier workstreams; it is to this that we now turn.

The dynamics of embedded research initiatives

We purposively selected four embedded researcher initiatives (denoted as Bridgetown, Coxheath, Porter and Evansville). These showed marked differences in the embedded researchers’ backgrounds, motivations and practices, yet a common thread was a desire to ‘make a real-world difference’, to get research ‘out of the ivory tower’ and to improve local services. Using the design domains surfaced in workstreams 1 and 2, we also saw much diversity in the scheme intentions, structures and processes. Taken together, these provided many insights into the wide range of embedded tasks and activities undertaken in the roles.

There were common themes across the schemes in their basic intentions. The embedded researchers (and members of their networks) aimed at mediating between different forms of knowledge, negotiating organisational, cultural and epistemic boundaries to promote co-produced, shared and actionable understandings. There were also significant differences between schemes (indeed, sometimes within schemes, e.g. in Porter). The degree to which knowledge co-production was evident varied considerably: even though these cases were all selected in search of co-production, the approaches deployed were more usually provisional or moderate (see Chapter 2) with more radical intents towards knowledge co-production largely absent. The degree to which active co-production figured was often dependent on the framing and structuring of the embedded researchers’ roles: less bounded roles usually offered greater scope than those roles with narrow and concrete goals.

Often there was no single or sustained intent, at least explicitly. Intentions evolved or fluctuated over time, sometimes only crystallising as the scheme matured. The intents/purposes were also influenced by key partners, not only at the design stage, but as schemes unfolded. The individual skills and motivations of the embedded researcher were also influential, especially in shaping some of the changing goals and increasing ambitions of the schemes.

Most intentions (as articulated by the embedded researchers, but also by their service partners and associated networks) were related to creating situated knowing by brokering external knowledge and combining that with locally collected data, building research capacity (to shift capabilities from academic institutions to front-line staff), and enhancing a reputation for research-informed care. Because of this range, the embedded researchers were often left to juggle, prioritise or reconcile competing and changeable expectations about what being embedded actually entailed. As a result, embedded initiatives (and especially their embedded researchers) needed advocacy at strategic levels: people who could ‘see the big picture’, shape the narrative and influence expectations. Embedded researchers also needed support (from academic partners) to maintain critical distance and objectivity, with local ‘capture’ being an occupational hazard.

A common theme under ‘intentions’, often raised obliquely, was the need to address power dynamics. The issue of power was seen as a potential minefield and was rarely addressed directly in the articulation of scheme intents. Nevertheless, a restructuring or rebalancing of power across different stakeholders was often implicit, not just in scheme intentions, but also in how these intentions were put into operation. However, understandings of power were, perhaps unsurprisingly, fairly rudimentary, and this remains an underdeveloped aspect of embedded research initiatives and of knowledge co-production in general.

Looking across these insights, it is clear that the intentions of embedded research schemes are far more complex than initially might be assumed. The diverse and dynamic nature of these goals makes for very different schemes under the broad rubric ‘embedded research’, and this diversity is seen equally often in the assembled structures through which embedded research is to be delivered.

Core to the operationalisation of an embedded research scheme is the creation of opportunities for sustained interaction between the embedded researcher and actors in the non-academic setting. Practical matters such as contractual arrangements, line management and physical location all contribute to proximity, visibility and perceived contributions. Such structural arrangements also have implications for more nebulous concerns, such as a sense of belonging or the maintenance or blurring of professional identities. There was much ‘hidden labour’ by embedded researchers in negotiating these concerns, and a real necessity to draw on wider networks and senior advocates to help achieve ‘buy-in’, ‘open doors’ or ‘provide air cover’.

The structural arrangements, and the softer consequences of these noted previously, often went through a recognisable trajectory: an initial incubation period exploring the possibilities; a degree of formalisation as the embedded researcher came into post; developmental work to ‘flex’ the arrangements as the scheme bedded in and responded to local contingencies (not least of which are the skills and aptitude of the embedded researchers themselves); and, finally, some consolidation and addressing of sustainability issues, or (for fixed-term initiatives) a managed completion and attention to legacy issues.

The dynamic and contingent nature of these structural arrangements, although a key element of the model, poses real challenges: when schemes are (explicitly or implicitly) in a process of ‘becoming’, there is a real need for ongoing dialogue among all stakeholders to keep coherence and alignment. Changes of personnel in the service setting, and evolving knowledge needs, reinforce the necessity of structured engagement and shared understandings. The insights gleaned from the literature reviews and the empirical work, and the tools and resources laid out in Chapters 6 and 7 (to be discussed shortly), could be seen to assist such processes.

Whatever the structural arrangements, embedded research initiatives delivered value through the activities and processes that they pursued. These varied dramatically, but included (to different degrees) activities aimed at research capacity development, activities related to research itself and activities in support of knowledge utilisation and service improvement. Either directly related to these ‘goal-supporting’ activities, or more indirectly necessary to build visibility and support for the initiative, were diverse activities aimed at creating, maintaining or strengthening networks, communications and relational bonds. Such relational processes were at multiple levels, including the strategic (securing senior management and clinician buy-in), cross-professional (building support for new research or improvement projects) and peer to peer (maintaining academic support networks and searching for informal support from people in similar roles).

The wide range of supporting processes and the emphasis on relationship-building often placed significant strains on embedded researchers. This draws attention to an apparent paradox at the heart of embedded research schemes: the relatively simple idea of placing research expertise at the sites where research-informed knowledge is most needed is gainsaid by the complexity of the structural and processual arrangements needed to support such placements. This further reinforces the need for training to support an embedded researcher’s own capacity development, and organisational strategies to ensure that appropriate support functions are in place. Again, the tools and supporting materials described in Chapters 6 and 7 may have a productive role to play.

Given the wide-ranging (and sometimes implicit) nature of embedded scheme intentions, and the varied and sometimes nebulous nature of the processes used to achieve these, it is unsurprising that most schemes struggled to articulate or evidence success. Although the enthusiasm for, and belief in the potential of, such schemes was widely evident (and there is a strong theoretical rationale based on critiques of ‘knowledge transfer’), systematic (as opposed to perceptual or anecdotal) evidence of effectiveness was harder to come by. Moreover, the scope for assembling such evidence varied as a function of the boundedness and definitional clarity of the schemes. Initiatives that were relatively narrowly defined (in terms of scope, activities, skill set, expectations, etc.) tended to have concrete deliverables that were more easily achieved and acknowledged. Embedded roles with more ambition and broader goals (such as increasing absorptive capacity or changing the culture around knowledge and collaboration) had more ambiguity and fewer boundaries, and, therefore, were much harder to assess, monitor and manage.

Overall, the case accounts and the insights drawn from them add nuance and depth to the 10 domains of embedded research uncovered through the literature work and scoping review. By themselves though, such insights sit more in the research domain, perhaps not so readily accessible to scheme protagonists. A core plank of the Embedded project was to ensure that robust insights rigorously derived through research were made widely available through practical tools and resources, and it is to these that we now turn.

Tools and resources to support embedded research schemes

As befits a project concerned with knowledge co-production, we adopted a collaborative, creative and engaged approach to translating the research insights into practical tools and resources (see the account of the overall engaging and influencing strategy in Chapter 1, the specific account of tools and resources creation in Chapters 6 and 7, and the recruitment resources pack in Chapter 7 and on the project web page [www.journalslibrary.nihr.ac.uk/programmes/hsdr/165221/#/ (accessed 18 February 2021)]. This led to a design framework (with visual metaphors, dialogic questions and interactive web resources) and various other supporting materials (e.g. an introductory animation, case studies of existing schemes and a recruitment resources pack). We anticipate that, collectively, these tools and resources will support the development of new embedded research schemes, their ongoing management and evolution, and (importantly) renewed efforts to examine their effectiveness both formatively (as part of within-scheme learning) and summatively (as part of a wider research effort to assess impacts).

In terms of developing new schemes, much of the literature on embedded research highlights the importance of agreeing and communicating the intention, structure and processes of initiatives early and clearly.65,162 Such calls tend to be born of the frustrating experiences of embedded researchers who are often faced with the need to manage competing demands and expectations and respond to the differing needs, priorities and values of the individuals and organisations involved. These dynamics were also readily seen in our case studies. We suggest that many of these frustrations and tensions stem from the difficulty of identifying and discussing the multiple features of an embedded research initiative, especially at its inception. A further layer of tension is introduced by the difficulty of reconciling some aspects of embedded research with one another (e.g. an initiative premised on disrupting power dynamics will need to pay particular attention to the contractual arrangements, as these may work with or against existing power relations). We propose that our framework provides a way of assessing the internal coherence of planned initiatives and the extent to which the various features have been considered and discussed sufficiently by those involved.

Moreover, because many embedded schemes evolve over time, the framework provides a means of revisiting design considerations in an ongoing manner, hopefully contributing to better communication and alignment across stakeholders. By focusing not just on the main design parameters (intentions and structures), but also on the more dynamic array of processes (what embedded researchers actually do), it should allow better dialogue between embedded researchers and other stakeholders as schemes mature or become reconfigured. Local learning too may be facilitated.

Many embedded research initiatives appear to lack clear mechanisms for learning and evaluation. This is perhaps unsurprising, given that the difficulty of evaluating such complex and relational knowledge/practice initiatives is well documented.199,200 It is, however, becoming increasingly accepted that evaluating complex social initiatives depends on the ability to articulate and link the intended outcomes, activities and resources associated with the initiative.199,201 This can inform, not only summative, but also developmental approaches to evaluation, which may be particularly valuable given the fluid and emergent nature of embedded research initiatives. We suggest, therefore, that our tools, resources and supporting materials can provide a basis for articulating an embedded research initiative’s intended outcomes, activities and resources and for building a coherent logic model to guide evaluation and learning. Such evaluative work is needed not only to support scheme evolution, but also to create an evidence base to underpin the selection of embedded approaches in the first place.

In these ways, we hope that our tools and resources can make a substantial contribution to the maturation of the field. Moreover, all the tools and resources described in Chapters 6 and 7 (and discussed here) will remain available67 hosted through the Research Unit for Research Utilisation179 at the University of St Andrews (project co-applicants and report authors VW and HD are co-directors of this long-established research unit). In this way, we hope to ensure the enduring accessibility and ongoing application of the insights detailed in this project report.

Reflections on patient and public involvement in embedded research

Data from the scoping of UK embedded research initiatives, the findings from the literature review, and the detailed work in the four case studies presented a picture of very patchy PPI. There was a wide range of PPI activity in the initiatives that we explored, ranging from a ‘patient panel’ supporting the overarching initiative (seen in Evansville; see Chapter 4) through to minimal involvement, with PPI described as not needed or not useful. In general, PPI was most often seen at project level (consistent with current research norms), but was rarely seen at the level of the embedded initiative itself. That is, patients and the public were far less likely to be involved in conversations about the overall design of embedded co-production than they were to be engaged in specific research or improvement projects. Even when PPI was seen, there were many different approaches to the role of lay representation, including drawing on existing NHS arrangements or using local charities to act as liaison. Moreover, the style of engagement also varied, from active participation in project work to more passive engagement, such as simple consultation about patient information.

Throughout the Embedded project, and as they learned about existing PPI arrangements in embedded schemes, our own project PPI group met to explore what a more effective and comprehensive set of arrangements for PPI in embedded research initiatives might look like. This group (see Acknowledgements) had wide and diverse experience of PPI in health services.

First, the group explored and differentiated active from passive PPI, under the three headings of intent, structure and process surfaced by the review and scoping exercise (see Chapter 3). As shown in Table 17, a spectrum of PPI arrangements can be envisaged from more passive to more active. Relating these features of PPI to the design concerns for embedded research initiatives allows us to see how building in more active PPI begins at the design stage. Thus, the passive/active ‘anchors’ in Table 17 provide a framework for integrating PPI concerns into the design discussions.

Discussions in the Embedded project PPI group concluded that the limited and patchy engagement of lay perspectives in many embedded research initiatives was a missed opportunity, but that there were few easy answers and no ‘off-the-shelf’ models. In particular, the group recognised the complexity and challenge of bringing together researcher perspectives with health service perspectives, and acknowledged the essentially relational nature of embedded co-production. This challenge can be encapsulated as that of building more complex and multifaceted relationships that include patients and lay representatives in equal dialogue with embedded researchers and service actors. That will require firm commitment as an intent, with redesigned structures and clearly set-out processes: using Table 17 in tandem with the embedded design framework (see Chapter 6) to build in deeper PPI. In addition, the group identified the need for more innovative recruitment to PPI roles to ensure that those participating truly reflect the diverse populations beings served.

These discussions in the Embedded PPI group led to the sketching out of a more radical vision for PPI in embedded research initiatives (Table 18). Again, drawing on the design framework (see Chapter 6), this sets out some core presumptions that would underpin more radical and embedded PPI. These could be seen as touchstones for discussions, or as more concrete goals for redesigned PPI in embedded initiatives.

In conclusion, the lack of nationally available models of PPI for embedded co-production has led to diversity of practice and sometimes a lack of focus on this crucial part of knowledge co-production. The work of the Embedded PPI group and the insights generated from the research data have been combined to sketch out a more radical vision for PPI in embedded initiatives, with Tables 17 and 18 offering a route map through discussions for implementing that more radical vision.

Future research

As befits the pragmatic intent of this project (supporting the development of embedded knowledge co-production), we have focused here on the practical insights and outputs from this project, and their potential application in the field, especially in the NHS. However, it is also clear that there are more academic contributions to be made emerging from the extensive empirical work. These will probably explore issues such as identities, power and liminality, as well as the means of accommodating across diverse epistemologies. That is, the material gathered across this project has the potential to contribute to academic and service innovation debates much broader than those limited to embedded research initiatives.

There is also an important future empirical research agenda. In order of importance to the overall aim of increasing the impact of health services research, we propose the following:

• How effective are different models of embedded research in achieving their aims? The Embedded study explicitly aimed to shine a light on the theory and current practice of embedded research, rather than to formally evaluate the outputs, outcomes, value for money and impact of embedded initiatives. The effectiveness and cost-effectiveness questions either are, or should be, asked by those responsible for initiating, funding and maintaining embedded research initiatives on the ground. There are many ways of addressing effectiveness. Given the complex, multifaceted nature of embedded research highlighted in this study, it is probable that a realist-type evaluation approach will be more useful than an experimental one.
• How does embedded research work in practice? This study has highlighted both the choices available to those who wish to use the embedded research model, and the possible ways in which embedding researchers in practical settings might add value. But more detailed process unpacking is required; for example, to unpick the role of embedded researchers in mediating different epistemic, cultural and political boundaries, such as the way stakeholders make sense of knowledge needs in different ways, and the divergent assumptions about what counts as valid knowledge and how knowledge might be used. Part of any embedded researcher’s role is to build connections between parties in ways that lay the foundations for more partnership-working, collaboration and co-production, leading to more sustainable collaborative communities. Exploring the accomplishment of these complex tasks is both necessary and valuable and could be achieved by further developing the methods used in workstream 3 of this study.
• What is the relative effectiveness of investing in the embedded research model, in comparison with investing in other ways of increasing the impact of health services research? The embedded research model is just one approach to increasing the impact of health services research. Some localities are investing in research partnership infrastructures such as local R&D units or learning systems. Others are funding placements for practitioners to spend time in academic settings, rather than placing researchers in service delivery settings. We do not currently know the relative effectiveness of these different investment decisions. Starting with comparative case study designs, this is an important research question for those investing in initiatives to increase the impact of research on practice.
• What role can patients and members of the public play in co-designing, implementing and evaluating embedded research initiatives? This study describes the significant challenges that we experienced in identifying the role of patients and members of the public in embedded research, despite our commitment to and investment in developing the PPI agenda. We developed a radical vision of what effective PPI in embedded research might look like. There is an urgent need to evaluate the implementation of different practical models of this vision, initially, at least, using an in-depth case study approach.
• What needs to be done to maximise the attractiveness to researchers of contributing to embedded research, and to minimise the risks? This study has clearly described the challenges and risks to researchers, especially early-career ones, of engaging in what is not currently a valued mainstream academic career path. Recruitment of researchers to embedded posts will remain challenging until we have a better understanding from a researcher and academic institutional perspective of the associated challenges and solutions. A qualitative case study approach would most likely add greatest value.
• Are the tools and resources developed as part of this study being used and are they useful? The purpose of this study was not only to deepen our understanding of the theory, evidence and practice of embedded research, but also to make a practical contribution to those wanting to establish embedded research schemes through the development of the tools and resources described in Chapters 2, 3, 6 and 7. Although these tools and resources were co-created with practitioners and researchers and were, in many cases, based on materials that have a track record of use in health and care settings, we do not have a systematic understanding of their use in practice. A multimethod evaluation of the tools and resources would help to maximise their contribution to embedded initiatives.
• To what extent are embedded research initiatives being used in the health and care sectors? This study identified a large number of initiatives across the country that satisfied our criteria for embedded research. We also identified the lack of infrastructure and sustained funding to support these initiatives beyond short-term projects. An ongoing assessment of the scale and scope of embedded research initiatives would build helpfully on the findings of workstream 2 in this study and enable policy-makers and research funders to maintain a strategic view of the field.

Taken together, there is a large and fascinating research agenda that opens up as a result of our preliminary attempt to gain some clarity and place some structure on the field of embedded knowledge co-production.

Concluding remarks

Embedded co-production initiatives have sprung up across the NHS in many different settings: acute care, primary care, community care and public health. There is considerable enthusiasm for such schemes, among service personnel and embedded researchers alike, and there is a robust logic that underpins their use. However, we found considerable diversity as to the nature and understandings of embedded research and knowledge co-production, a diversity and complexity that belies the apparently simple precepts.

By definition, embedding research means deploying a person (or people) to build connections between the world of research and that of practice. We found that beyond this simple and seemingly straightforward idea lies a tangled web of possibilities and challenges that have much to do with the structures, cultures and social institutions within which individual capacities and experiences are nested. A researcher may be embedded, but, in a broader sense, they also facilitate embedding by establishing connections between people, ideas, knowledge and action, with potential dividends in terms of social and institutional change. Judgements about, and measures of, successful embeddedness are likely to be highly variable between particular contexts and projects. Understandings about embeddedness and co-production are not yet at the stage that would allow detailed recommendations to be made that cut across contexts. Rather, every embedded research initiative is an opportunity to learn more about the gains that are possible and the challenges to be faced.

The insights and resources from this project are intended to support the NHS and other partners to develop such learning. These focus on providing practical support to the dialogue needed for more thoughtful schemes, and we hope that this more structured way of describing and delineating schemes will also allow for more concerted research efforts to assess scheme impacts where they count: on the service.