Effective implementation of change in healthcare organisations involves multiple professional and organisational groups and is often impeded by professional and organisational boundaries that present relatively impermeable barriers to sharing knowledge and spreading work practices . Professional boundaries in healthcare are reinforced by historically determined power and status differentials between healthcare professionals  whereas organisational boundaries could be caused by the government’s imposition of divergent performance frameworks upon organisations that are expected to collaborate, with explicit incentives for collaboration frequently being absent . A complementary view of professional and organisational barriers to knowledge sharing is suggested by practice-based theorists who maintain that knowledge is localised, embedded and invested in collective practice and see boundaries as inherent sociocultural differences between distinct collective practices underpinned by shared language, meanings and ways of doing things [4–6]. According to this tradition, the effect of these differences is dual: they can lead to innovation, learning and cross-fertilisation between practices, on the one hand, and to separation, fragmentation and disconnection, on the other [7, 8].
One of the practice-based theories specifically exploring boundaries between different sets of practice is the communities of practice (CoP) approach developed by Jean Lave and Etienne Wenger [9, 10] and applied to the analysis of learning, practice, meaning and identity in various contexts, including organisational studies  and healthcare [11, 12]. CoPs are work-related communities of individuals created over time through sustained collective pursuits of shared enterprises [4, 10, 11]. Shared knowledge, practice and identity produce boundaries between CoPs, at which they differentiate themselves from and also interlock with other communities, forming complex social landscapes of practice. An inherent feature of the CoP landscape is actors’ multimembership in various CoPs which can be crucial for bridging boundaries between communities [10, 13]. It should be noted that CoP boundaries do not usually coincide with the organisational ones. On the one hand, organisations represent a multiplicity of subcultures and could at best be seen as ‘constellations of interconnected practices’  or ‘communities-of-communities’ . On the other hand, some of the CoPs may cut across organisational boundaries, with their members potentially transferring knowledge between organisations [5, 14].
At the same time, CoP boundaries are often seen as a reproduction of professional boundaries, with the possibility of multiprofessional CoP formation in healthcare being contested. For example, in their study of eight National Health Service (NHS) innovations, Ferlie et al.  show that CoPs in healthcare are predominantly uniprofessional, tend to seal themselves off from neighbouring professional communities and are highly institutionalised, which enables a relatively easy flow of knowledge within these CoPs but causes the ‘stickiness’ of knowledge across boundaries and hence retards the innovation spread. They argue that great effort is needed to bridge professional boundaries and create a functioning multiprofessional CoP because uniprofessional CoPs tend to defend their jurisdictions and group identity. The construction of such a CoP was observed by Ferlie and colleagues only in one of their primary care cases, where professional boundaries could be successfully bridged because GPs and practice nurses shared common values; participation in change was incentivised; established systems for interprofessional dialogue were deployed; and basic cognitive assumptions of professional groups remained unchallenged.
A more optimistic view on multiprofessional CoPs is presented by Gabbay and le May  in an ethnographic account of knowledge sharing in primary care, which describes a system of overlapping ‘communities of general practice’ existing in primary care organisations. In addition to the uniprofessional ‘coffee room GPs’ community which is at the centre of their analysis, the authors also identify several ‘specialist’ CoPs located within the same practice, some of which are multiprofessional; wider (and looser) CoPs external to the practice (e.g. a group of fellow managers for the practice manager or a network of old colleagues for practice doctors); and, interestingly, a multiprofessional CoP that evolved from formal practice meetings and included almost all of the staff in the practice. Another example of a multiprofessional CoP operating in primary care is provided by Hudson , who describes a multi-agency team working at the interface of district nursing and social work and argues that the promotion of shared values and socialisation to an immediate work group can override professional or hierarchical differences amongst staff and lead to the formation of a multiprofessional and multi-organisational CoP.
Empirical studies outlined above deployed the analytical perspective on CoPs, whereby this theory is applied to the analysis of processes that take place in organic CoPs naturally emerging as a product of collective practice over a relatively long period of time. Another strand of CoP informed thinking, which has been labelled as the instrumental perspective[13, 17] is concerned with the deliberate cultivation of CoPs in order to bridge professional and organisational boundaries and enable knowledge transfer . Deliberately constructed CoPs have been shown to be effective in enhancing professional education, adoption of innovation and problem-solving [19, 20]. In service improvement, CoP cultivation has been specifically advocated as an approach useful for creating horizontal networks across organisations, promoting the sharing of tacit knowledge and achieving a better sustainability of change [21, 22]. However, emergence of genuine CoPs within quality improvement and other change implementation initiatives in the healthcare sector may be problematic due to the time-limited nature of the projects, a top-down approach to change management and preoccupation with performance measurement at the expense of human and social aspects of change [3, 22, 23]. In addition, administrative staff, nurses, medical practitioners, allied health professionals and managers have been shown to significantly differ in their conceptualisations of quality and safety, which may challenge the collaborative implementation of service improvement initiatives in a multiprofessional environment [24, 25]. Interestingly, while the instrumental perspective on CoPs sees bridging the boundaries as its target, the impact of pre-existing professional and organisational boundaries on CoP engineering in healthcare collaboration seems somewhat underestimated in the literature .
The brief outline of literature presented above points to a number of aspects requiring more empirical attention. First, boundaries to knowledge sharing in healthcare have been predominantly explored in secondary care settings [26, 27], at the interface between the primary and secondary care sectors [28, 29], and in partnerships between NHS organisations and higher medical education . At the same time, the nature and effects of boundaries existing in the primary healthcare sector may be different from those found in secondary care and intersectoral collaboration. Little is known about how complex landscapes of practice comprised of multiple CoPs influence intra- and inter-organisational knowledge sharing, especially in the process of service improvement initiatives introduced in primary care. Second, while there is a growing number of studies exploring the deliberate cultivation of CoPs in order to promote evidence-based practice, foster collaboration and achieve service improvement in healthcare, less is known about how professional and organisational boundaries shape the development of new CoPs within these initiatives, whether the manipulated emergence of new multiprofessional and multi-organisational CoPs in primary care is realistic, and how these groups relate to pre-existing organic CoPs.
In light of the above, this study is guided by the following research questions:
How do boundaries between CoPs existing within and across general practices influence the implementation of a primary care service improvement programme?
How do these boundaries affect the emergence of new multiprofessional and multi-organisational CoPs within and across primary care organisations?
Before discussing the methodology deployed to address these research questions, it is worth clarifying the definitions of boundaries and communities of practice used in this study. In line with the practice-based approach to knowledge sharing, we define boundaries as sociocultural differences between groups that may lead to discontinuity in action or interaction . Our understanding of boundaries, therefore, partially overlaps with the notion of ‘gaps’ popular in the knowledge transfer literature, where gaps are seen as ‘the network holes, spaces and missing ties that create between-group problems and opportunities for their resolution’ . It could, however, be argued that the latter approach emphasises structural and relational separation between groups that can be ‘bridged’ by ‘transferring’ knowledge from one group to another through routines, protocols and other information channels. By contrast, our understanding of boundaries as discontinuities, underpinned by differences between groups in terms of practices, identities and meanings, highlights the cultural and political nature of these phenomena, shifts the focus of analysis from ‘gaps’ and ‘bridges’ to divergent meanings, interests and cultures, and underscores the importance of reflection, learning and transformation when dealing with boundaries [31, 32].
Following Wenger’s seminal analytical text on CoPs , we define a CoP as a group of individuals created over time through sustained collective pursuit of a joint enterprise and developing mutual engagement with each other as well as a shared repertoire of meanings, routines, stories and artefacts. In line with Wenger’s theory, CoPs are also characterised by the presence of boundaries, shared identities and collective histories of learning. It should be noted, however, that we do not accept the clearcut and sometimes criticised [17, 33] dichotomy between teams and CoPs postulated in Wenger’s later writings [18, 34]. Based on our own previous empirical research , we argue that some teams can develop certain CoP characteristics (also see [36, 37]) and that in these cases the CoP theory is applicable for analytical purposes.