Leadership , service reform , and public-service networks : the case of cancer-genetics pilots in the English NHS

In attempting to reform public services, governments worldwide have sought to effect change through policies aimed at both transforming structures of public service provision and facilitating the agency of public servants working within these. Various obstacles have been found, however, to impede the effectiveness of such efforts. In this paper, the authors examine the role of organizational networks and distributed leadership—two prominent policies aimed at structure and agency respectively—in the establishment and consolidation of service reform in the English National Health Service. Using a comparative case-study approach, they contrast the trajectories of two attempts to introduce and gain acceptance for service reform, noting important differences of context, process and outcome between the sites. The findings indicate the importance of dispersed, as well as distributed, leadership in achieving change in a networked public-service setting. Effective leaders may indeed achieve change through the structures and processes of the network. However, the coexistence alongside the network of other organizational forms constrains the ability of leaders to achieve change without complementary action beyond the boundaries of the network.


Introduction
Across the economically developed world over the last 10-15 years, governments have pursued increasingly interlaced strategies aimed at reforming public service provision. Transformative efforts to ‗Reinvent Government' (Osborne and Gaebler 1992) in the United States (US) and ‗modernize public services' (Newman 2001) in the United Kingdom (UK) are exemplary of policies seeking to improve both the effectiveness and efficiency of public-service delivery. In pursuit of this aim, policies addressing the structures of public-service delivery-the way public services are organized-have accompanied policies addressing the agency of public servants-the skills and attributes of the individuals employed in public services. Often, these two approaches to public-service reform are seen as mutually dependent. Organizational reforms may give rise to new roles necessitating new skills sets; the broadening of public service staff's skills may in turn allow them to act more dynamically and autonomously in identifying opportunities for service improvement and acting upon them.
This apparent synergy between structural reconfigurations and upskilling individuals as agents of reform, however, is not always so straightforward in practice. Many commentators have noted the contradictions inherent in certain organizational changes within the public services, and the inhibitive effect these have on the possibilities for collaboration and joined-up service provision in reformed public services. In particular, there is the well documented tension between hierarchies, markets and networks. The collaborative network, as an organizing principle for public-service provision which might bridge gaps between state agencies and assist in addressing social problems that defy the efforts of a single agency (Jennings and Ewalt 1998;Provan and Sebastien 1998;Huxham and Vangen 2000), clashes with traditional top-down hierarchical organizational forms, and with the increasingly prominent role of markets and quasimarkets in national contexts subject to the Anglo-American influence (Hood et al. 2000;Ferlie et al. 2003;Currie and Suhomlinova 2006). In the UK, a policy focus on networks has coexisted with the reinforcement of a hierarchical performance-management regime and the increasing marketization of public-service provision, mirroring developments elsewhere (Kickert and Koppenjan 1997;Tenbensel 2005).
In this paper we empirically examine the degree of convergence (or otherwise) between a particular structural reform-the centrally mandated network-and a particular facet of individual agency-leadership of change-which have both been central to government reform policies, in the UK and elsewhere. Networks and leadership have been advocated as means of reinvigorating state-run services, especially in the US, the UK, and developed-world countries subject to Anglo-American influences such as Canada, Australia and New Zealand O'Toole 1997;Shortell et al. 2000;Agranoff and McGuire 2001;Hennessey 1998;Kakabadse et al. 2003). In theory, leadership and networks should complement each other, with the looser organization and less hierarchical ordering logic of the network allowing leadership of change, distributed among network members rather than led from a single organizational apex, to flourish. This would affirm the findings of Denis et al. (1996) on the potential of distributed leadership in complex organizations with plural, ambiguous aims (cf. Bryson and Kelley 1978). In practice, as we set out in the sections that follow, this may not be the case. Moreover, as we explore in the empirical section of this paper, in the ever more complex and unstable context of networks in the National Health Service (NHS) in England, 1 both the possibilities of distributed leadership and the barriers it faces are increasing. Drawing on a wider comparative study of the implementation of pilot cancer-genetics services in the NHS, we focus on the divergent courses of two case-study sites in which similar service reforms were introduced, highlighting the importance of differences in network organization and leadership style, and both the synergies and incongruities between these. Our analysis suggests that in itself, the structural reform of the network does little to achieve public-service reconfiguration. It does, however, create a ‗space' within which a certain enactment of leadership might be effective in linking diverse actors across organizational boundaries in pursuit of joint agenda. To the extent that the network competes with other organizational forms and modes of governance, however, this space is a constricted one; consequently there are limits to the effectiveness of the leaders and their agency, at the points where the network overlaps with other organizational forms (hierarchies and markets), which have their own norms of influence and authority which differ from those which are effective within the network.
We continue first by reviewing the policy and academic literature on networks and leadership, before describing the specific empirical field in which our study was located. We then present our methods and findings, before discussing the theoretical and policy implications of these.

Networks and leadership: theory and policy in the public services
The increasing attention paid to the role of networks and leadership in public administration has its roots in the waves of reforms to public services over recent decades often grouped under the heading ‗New Public Management' Peters and Pierre 1998;Hood and Peters 2004). Though all characterized by efforts to improve the effectiveness and efficiency of public services, these reforms were heterogeneous in nature. The evolution of the approach, exemplified in approaches such as ‗Reinventing Government' and ‗modernization' that aim for a more totalizing, transformative effect on public services, has precipitated an increasing need for new organizational forms and new forms of leadership. As Ferlie et al. (1996: 13) have it, this model of New Public Management that emerged from the 1990s onward moves away from hierarchical forms and leadership approaches towards -new management styles, such as management by influence; an increased role for network forms of organization; stress on strategic alliances between organizations as a new organizational form.‖ Collaborative organizational forms and associated new forms of leadership have gained currency among governments of many developed-world countries (Scharpf 1993;Provan and Milward 1995;Rhodes 1997;Agranoff and McGuire 2003;Hall and O'Toole 2004;Meier 2004). In the UK, the publicservice network has been seen as a means of ‗joining up' service provision even as hierarchy and market have remained, and indeed been strengthened, as modes of governance. Though the notions of the intra-and inter-organizational network are founded in the private sector (see Nohria and Eccles 1992;Reed 1992), there is a growing literature on their role in public-service provision (Provan and Milward 1995;O'Toole 1997;Peters and Pierre 1998;Rashman and Hartley 2002). Private-sector networks are largely explained in terms of reduced transaction costs and interdependence between firms, but for public-service organizations, network-level outcomes for clients are additionally seen as a motivation for integration (Provan and Sebastian 1998). In the literature, moreover, networks are seen as more flexible and adaptable in the face of organizational turbulence, as opposed to hierarchies which orient towards stability (Thompson et al. 1991), though empirical evidence for this is not strong (Bate 2000;Rashman and Hartley 2002).
Networks have had a particular appeal for policymakers under ‗new Labour' governments in the UK since 1997, as an adaptable organizational form which combined the dynamism of the market with the co-ordination of the hierarchy in delivering the ‗modernization' of public services. By facilitating joined-up working across professional and organizational boundaries, and by assisting the sharing of knowledge, good practice might be translated into rapid service improvement, spread across the public services. As we will see in the next section, though, the reality of the implementation and practice of organizational networks is not quite so straightforward.
Alongside this focus on the reconfiguration of organizations and the relationships between them, efforts at public-service reform have also been based on complementary attempts to transform the workforce and its capacities: empowering public-service employees to act intelligently and autonomously within the enabling structures of networked organizations to effect change. The ‗Reinventing Government' initiative of the Clinton-led administration of the US is exemplary of the expectations placed on this synergy, as a means of -[changing] the culture of our national bureaucracy away from complacency and entitlement toward initiative and empowerment.‖ 2 In the UK, new Labour's ambitions for ‗modernization' similarly envisaged a public-service workforce empowered to act intelligently and autonomously to effect change within the new organizational forms (Currie et al. 2005).
Within such policies, ‗leadership' has been an ongoing emphasis. Research on leadership in the private sector has tended to give precedence to transformational leadership, embodied in a single, heroic figure at the apex of an organization, as both a subject of study and a prescription for effectiveness (Bryman 1992;1999). Such norms and ideals of leadership have to some extent permeated public-service leadership policy (Currie et al. 2005), but there is recognition also (among academics and policymakers) of the limits to this model in the more pluralistic and political context of the public services. For Heifetz (1994), leadership theory premised on the idea of a single inspirational leader is inapplicable to the public services, and attempts to conceptualize public-service leadership in this way serve only to perpetuate this myth. Rather, leadership is about ‗adaptive work' which requires buy-in and enactment by individuals at all levels of an organization. In the policy and practice of leadership in the public services, then, there is an onus on distributed leadership, -from apex to front line‖ (Hartley and Allison 2000, 39), even as power and responsibility are formally concentrated in the roles of the heads of publicservice organizations. Arguably, the need for distributed leadership is all the more pronounced in public-service networks, since -the notion of a leader with a hierarchical relationship to followers does not apply in collaborations, so the potential for exercising ‗decisive leverage' by virtue of a formal position is reduced‖ Vangen 2000, 1167;cf. O'Toole 1997). Whilst the need for an understanding of leadership as distributed within public-service organizations, and especially networks, is accepted, exactly what constitutes distributed leadership is less obvious, and the subject of ongoing clarification in the literature. Van Wart (2005) offers a typology of distributed leadership, variously emphasizing the deliberate delegation of leadership from the top down through ‗substitutes for leadership' and ‗superleadership', and the more ‗organic' growth of ‗self-leadership' and ‗self-managed teams'. In all cases, though, shared leadership ultimately requires both a willingness to cede leadership to others on the part of organizational heads, and the capacity of other actors to take it on (van Wart 2005, 372-3). As Crosby and Bryson (2005, 29) suggest, -potential for effective leadership lies alike with those who do and do not have formal positions of power and authority. Indeed, this view of leadership may be most useful in reminding those with little authority how powerful they can be through collaboration […] and in reminding those in a supposedly powerful position just how much they rely on numerous stakeholders for any real power they have.‖ Thus even if we accept Bennett et al.'s (2003, 7) notion of distributed leadership as -the product of concertive or conjoint activity, […] an emergent property of a group or network,‖ it is important to recognize that such emergence occurs within the boundaries set by other parties, including those in whom leadership is traditionally concentrated. Furthermore, distribution of leadership emphatically does not imply equal spread of leadership among all parties. Power and influence may remain concentrated in certain ‗nodes' of leadership, who -lead up and out rather than down‖ (Crosby and Bryson 2005, 32). This is, indeed, the finding of Vangen and Huxham (2003), who find that the leadership qualities required by network managers are as much about ‗thuggery'-manipulation and politicking-as about facilitation and mobilization. These contours of distributed leadership-in particular, its relation to traditional top-down authority, and its potential concentration in new nodes of power-are important to acknowledge in an understanding of the role of leadership in networks. As with transformational leadership, distributed leadership is seen in the literature as having the potential to improve organizational performance (Rainey and Steinbauer 1999;Denis et al. 1996;Pearce and Conger 2003).
Government policy in England has emphasized the centrality of leadership in the modernization of all fields of public services (Cabinet Office 1999). Calls for individualistic leadership from the top and distribution of leadership co-exist in government policy, so that in health, for example, programs supporting transformational leadership from organizational heads are accompanied by efforts to empower staff and -lead change through people‖. 3 Together, structural reforms such as the introduction of networks, and reforms aimed at harnessing the agency of staff, such as leadership development, are seen by policymakers as having the potential to contribute positively to reforms in the US, the UK and elsewhere. Furthermore, they are seen not just as complementary, but to some extent as mutually dependent. Without equipping and empowering staff with the leadership and other skills to inseminate, manage and adapt to change, organizational reforms to public services are bound to fail. To this extent, then, the logic of these approaches to public service reform is evident. As we see in the next section, however, the practice is not so clear cut.

Networks and leadership: obstacles and impediments
In implementing networks in the public services, certain practical difficulties come to light. First, and most obviously, the resilience of the hierarchy and the increasing role of the market in many national contexts mean that networks have coexisted alongside two other organizational forms, with certain consequent tensions. Top-down performance management and quasi-market-based competition have the potential to drive a wedge between individuals and organizations in a way that is inimical to the collaborative ideals of the network Currie and Suhomlinova 2006). To the extent that their performance is managed according to their core duties, and judged in competition with their colleagues, professionals will tend to orient towards their own silos rather than reach out to foster collaborative relationships which may not benefit them. The same is true of organizations (Hood et al. 2000;McNulty and Ferlie 2002;Ferlie et al. 2003) and is reflected in various national contexts where network-based reforms coexist with other organizational forms, such as Australia (Ryan and Walsh 2004), the Scandinavian countries (Christensen et al. 2007), the Netherlands Brandsen and van Hout 2006) and the US (Provan et al. 2004). Competition between organizations in contexts governed by market or quasi-market relations, as exhibited for example in managed care networks in the US, will have a similarly divisive effect (Shortell et al. 2000), with the risk that agencies relapse into relationships based on contracting rather than collaboration (Brandsen and van Hout 2006;Keast and Brown 2006).
Furthermore, many public-service fields, such as health and education, are professional bureaucracies in which these kinds of difficulties are compounded (Riccucci 2005). Each profession has its own pervasive socialization, which might work against a network orientation in a number of ways: a professional knowledge domain that is esoteric, to some extent tacit, and difficult to share (Currie and Suhomlinova 2006); a career pathway that tends to be based on specialization rather than broadening of knowledge (Nancarrow and Borthwick 2005); and a place in a well established and institutionalized hierarchy of professions and subprofessions (Bate 2000). Consequently professions have long been acknowledged as averse to managerial interventions (e.g. Lipsky 1980;Riccucci 2005). Notably for our empirical field, physicians are often characterized as the quintessential autonomous profession (e.g. Riccucci 2005), and tend to be individualistic in their practice and attitude towards integration (Shortell et al. 2000). Furthermore, the status, power, and expertise of the medical profession mean that any set of organizational reforms requires the compliance of physicians for its effectiveness (Greco and Eisenberg 1993;Shortell et al. 1998). This is arguably all the more so in the NHS, where in contrast to the relative erosion of medical power and autonomy in the US (Hafferty and Light 1995), physicians have retained an authority that gives them power of veto over NHS administrators in regard of any organizational reconfigurations that they do not welcome (Harrison et al. 1992;Bate 2000).
Moreover, there is the risk that, in insisting that organizations adopt a more networked form, policymakers undermine the very essence that makes the network a functional organizational form in the first place. If they rely for their effectiveness and dynamism upon their voluntaristic, informal and organic nature, then efforts to impose networks may have perverse consequences. In particular, as Kickert and Koppenjan (1997: 39) point out, in networks, -government is not the single dominant actor that can unilaterally impose its will. Hierarchicial, central top-down steering does not work in networks.‖ Rather, governance mechanisms must fit within the network structure (Bruijn and Heuvelhof 1997). Yet the ongoing influence of top-down performance management regimes, so strong in countries such as the UK where vertical accountability remains characteristic of public services, may militate against this need, to the extent that the network may become little more than an instrument of surveillance and performance management, to the detriment of collaborative efforts by network managers and members (Addicott et al. 2007).
The necessity of more lateral means of governing and guiding networks towards common aims creates particular challenges for the leader within the public-service network. The widely noted ambiguity of objectives, power and accountability that characterizes the public sector as compared to the private sector (Heifetz 1994;Denis et al. 1996; is compounded, as we have seen, by the network. This suggests a need for the distribution of leadership across individuals in collaborating organizations. However, organizational objectives are frequently set by stakeholders external to networks, such as policymakers. This complicates any such effort to pluralize change agency and distribute leadership, since actors across networks will be constrained in their actions by the parameters set by these external stakeholders (Currie and Lockett 2007), which may not encourage collaborative activity. Furthermore, the hierarchical order of professions-in health, for example, the ongoing subservience of nursing to medicine, at least in the view of many physicians, and the endurance of intra-professional hierarchies and power relationships (for example between surgery and medicine) (Currie and Suhomlinova 2006)-is also likely to, at best, complicate any effort to distribute leadership beyond traditional leaders.

Networks and leadership: transforming structure and agency?
The preceding section indicates the difficulties of establishing both networks and corresponding modes of leadership in public services characterized by the co-existence of other organizational forms and various tensions between professions and policy aims. Ostensibly, distributed leadership and networked organizational forms seem well suited to each other, but the complications of the public-services context mean that their effectiveness and synergy in effecting service reform are compromised. Equally, however, each policy might have the potential to surmount the difficulties presented by the contingencies of the public services, by reconfiguring power relationships to foster a more dynamic organization than one governed by hierarchy or competition.
What is suggested in particular is the need for an enactment of leadership that seeks to engage a multiplicity of stakeholders, facilitating and consolidating change by concentrating on processes and outcomes together (Huxham and Vangen 2000;Vangen and Huxham 2003). Even as this suggests a distribution of leadership, it also implies personal investment by those charged with network leadership tasks: as Vangen (2000, 1171) have it, -it is paradoxical that the single-mindedness of leaders appears to be central to collaborative success.‖ At the same time, though, in a professional bureaucracy characterized by looseness of accountability and practitioner autonomy, there is a need to engage professionals for organizational change to be achieved and consolidated (Shortell et al. 2000;Riccucci 2005). Any change needs to ensure that -organizational players are empowered enough to be able to avoid the frustration and chaos that leads to the reassertion of pervious forms and processes‖ (Ferlie et al. 1996, 14).
In other words, the structural change of the network is reliant on the agency of leadersdistributed or otherwise-in making it functional, since -formal bureaucratic structure, as defined by rules and regulations, has little impact on the behaviors of front-line workers‖ (Riccucci 2005, 64). For all the collaborative rhetoric of the network form, then, it will have little tangible impact on practice in the absence of effective leadership. In making this rhetoric reality, leaders can, however, draw on the resources of the network: in the language of Vangen (2000, 1171), the -leadership media‖ of -structures, processes and participants.‖ It is through the properties of the network that leadership might be enacted, and although these may well be outside the control of any one leader, they are also to some extent malleable. Structures, for example, -normally emerge out of the practical reality of the tasks that they tackle‖ Vangen 2000, 1167), so that even within an overarching organizational structure, certain ‗substructures' might be assembled in pursuit of the leader's task. Similarly, O'Toole (1997, 48) suggests to network managers the possibilities of a number of routes to influence: -find ways to shift network membership toward more supportive coalitions; locate key allies at crucial nodes; try to alter agreements between the parties to heighten program salience; and buffer wellfunctioning arrays to limit uncertainty and complexity.‖ Crosby and Bryson (2005) indicate the potential for influence in networks through ‗forums', ‗arenas' and ‗courts', each subject to their own rules of conduct and therefore amenable to particular enactments of leadership. These are the concrete opportunities within inter-organizational collaborations for leaders to be effective, for example through visionary leadership, -creating and communicating shared meaning‖ in formal and informal network forums that inseminate ideas, and political leadership, -making and implementing decisions‖ in the legislative and executive arenas that determine formal policy (Crosby and Bryson 2005, 35).
Whilst the nature of the structures, processes and participants of the network will vary, and with them the opportunities for influence, what these ideas do encompass is an emergent image of what the effective leader within a network might look like, in terms of skills, style and personality. The remainder of this paper considers this question, as well as the more general question of the interaction between the structural reform of the network and the agency of leaders in achieving change. In the next section, we introduce the specific field in which we conducted our fieldwork, and describe the service reform that the pilots studied attempted to introduce into this context.

Cancer networks in the NHS, modernization and service reform
Although in the UK networked public services are often characterized by commentators as being a particular feature of policies of new Labour governments since 1997, the form has earlier roots. Reflecting wider trends for networked organization in the policies of various developed-world countries, UK public services were already moving towards networked governance in the early 1990s under the previous Conservative administration (Newman 2001). To this extent, networks can be seen as a logical development of the New Public Management policies prevalent across developed-world governments, especially those subject to Anglo-American influences O'Toole 1997). In the UK, health policy was an early adopter of the network principle, with ‗managed clinical networks' introduced in a variety of specialties, including cancer, in the NHS from the mid-1990s, mirroring trends towards networked healthservice delivery in Europe (Wijngaarden et al. 2006) and the US (Shortell et al. 1994).
Though funded from general taxation and offering universal access to British citizens, the NHS is far from a monolithic public-service organization. In common with other fields of public-service provision, the introduction of the internal market to the NHS in the early 1990s created an increasingly fragmented organizational scene. From a system based on geographically determined organizational divisions, the NHS was subdivided into purchaser and provider ‗trusts', each effectively an organization in its own right, with autonomy over its human resources policies, contracting arrangements, and finances, and the potential (in theory) for bankruptcy (though the political unacceptability of hospital closures ensured that in practice this was rarely threatened, with mergers and takeovers a more palatable alternative). As with public-service networks elsewhere in Europe and the world , then, the network represented one means of mediating such organizational fragmentation: effectively, managed clinical networks in the NHS were to be inter-organizational in character (between autonomous NHS trusts), albeit technically within a single overarching organization, the NHS.
In relation to cancer, the 1995 Calman-Hine report envisaged new service structures in England and Wales -based on a network of expertise in cancer care reaching from primary care through Cancer Units in district hospitals [acute secondary-care hospitals] to Cancer Centres [specialist tertiary-care facilities]‖ (Department of Health 1995, 7). The vision here was of a network that would coexist with the emerging internal market in healthcare, by encouraging competing providers to work together with purchasers to reduce duplication of services and promote clearer patient pathways. Good practice was to spread across the network through leadership from the medical lead in collaboration with colleagues across the network. In the interests of better clinical outcomes, though, concentration of expertise was also on the agenda, with cancer units needing a critical mass of patient throughput to maintain surgical subspecialization, and expertise in rarer cancers and specialist diagnostic and treatment processes to be concentrated in cancer centers. Consequently, while they were intended to be collaborative ventures between the different partners in the networks, ‗big players' in the networks-hospitals with much to gain or lose from the concentration of resources-have in many places come to dominate their agenda over smaller hospitals and primary care organizations. From the start, cancer networks were intended to promote both service rationalization and collaboration-with all the potential tensions this could create. However, as Addicott et al. (2007, 95) point out, in recent years the role of cancer networks has been significantly managerialized, becoming -largely focused on the structural configuration of services, performance targets and workforce planning.‖ In four of the five cancer networks they study, a preoccupation with meeting topdown performance-management agenda meant that collaboration was limited.
Despite their government-driven mandate, cancer networks mirrored the function of two rather different network forms in the US. First, they reflected the role of integrated networks of health-care providers that emerged in response to the market pressures created by the advent of managed care, as a means of rationalizing services and offering cost-effectiveness through economies of scale (Shortell et al. 1994). Second-and as Addicott et al. (2007) point out, in tension with the first role-they also serve a function similar to that of the community clinical oncology programs and special populations networks funded by the National Cancer Institute, with their aim of fostering knowledge-sharing relationships between research institutions and community-based intervention, education and research projects (McKinney et al. 1993;Baquet et al. 2005).
This was the context faced by the pilot program of cancer-genetics services cosponsored by the Department of Health and Macmillan Cancer Support, 4 which sought to improve provision in this field by introducing a structured care pathway for those with a suspected family history of cancer. The Harper report (Department of Health 1998) identified inequalities, inefficiencies and poor practice in cancer-genetics provision, and suggested a division of labor and flow of knowledge between primary care, cancer units and cancer centers which mirrored the set-up of cancer networks. This would result in smoother patient experiences, and the elimination of poor practice, such as inaccurate risk assessments resulting in unnecessary screening (if genetic risk is overestimated) or inappropriate reassurance (if underestimated). An expert group convened by Macmillan and the Department of Health operationalized this in a framework known as the ‗Kenilworth model', and in turn, this was adapted to local conditions and piloted in seven locations across England. With endorsement from an expert committee comprising policymakers, hospital clinicians, family physicians (GPs), nurses, service users and others, it had credibility and a degree of ‗top-down push', though it was not mandated; rather, the expectation was that each pilot would shape the care pathway in accordance with the particularities of its own locality.
In principle, then, the Kenilworth model represented just the kind of improvement to and rationalization of service organization that cancer networks were designed to help to achieve and reproduce. In terms of innovation diffusion, the overall setting augured well. With the change agency of Macmillan and the Department of Health behind it, the innovation of the Kenilworth model seemed to have some leverage behind it, and addressed an identifiable problem (Rogers 2003). Given the preceding discussion, however, there are clear obstacles to the operation of networked collaboration, and to leadership, that might confound uptake and establishment of the new care pathway. In the remainder of this paper, we compare the trajectories of two of the Kenilworth pilots, considering the roles played by leadership and by the cancer networks in which they were located in their course. The degree of success of the two pilots was divergent, as assessed in terms of two key outcomes: throughput of patients, and whether the model was sustained beyond the pilot period. ‗Derton' had a throughput that was several times that of ‗Nottley'; the project (and therefore the Kenilworth-based care pathway) in Derton was sustained with local money following the pilot period, while Nottley's project and pathway were not. 5

Methods
These two cases were among 11 that were selected as case studies for a wider evaluation of a program of pilot genetics services in the NHS. This program encompassed 28 governmentfunded pilots working in various clinical areas, including the seven cancer-genetics pilots working to implement the Kenilworth model in different localities. The pilots followed a genetics white paper (Secretary of State for Health 2003), which set out the aim of ‗mainstreaming' genetics knowledge and practice in everyday NHS provision. The pilots were to be one means of achieving this ambition.
The brief of the evaluation was to identify the barriers and facilitators faced by the pilots in realizing this aim. In pursuit of this, a comparative-case-study approach was adopted, using qualitative methods to facilitate intra-and inter-case analysis to illuminate key differences of context and mechanism that give rise to differences of outcome. This enables generalization through process analysis and the development of theory (Eisenhardt 1989;Yin 2003). More generally, where quantitative approaches permit the statistically based identification of apparent correlations between cause and effect, qualitative methods such as these are particularly useful in elucidating the mechanisms which give rise to such correlations (Lee 1999;Silverman 2004). As such, they are well suited to the study of processes within organizations (Pettigrew 1997;Langley 1999): in this case, the establishment of service reform within a network and the enactment of leadership in facilitating this. Our study thus addresses the gap identified by Bryman (2004) in relation to research on leadership, by adopting a qualitative approach which does not simply attempt to replicate quantitative research in identifying the key variables on which effectiveness rests, but rather focuses on the mechanisms in given contexts (such as the network) that give rise to certain outcomes.
Our selection of 11 case studies from 28 funded pilots was theoretically guided (Eisenhardt 1989). The intention was to select a sample that was representative of the variation present among the 28 pilots in terms of a number of variables. These included clinical field, host organization (primary care organization (PCT), ‗mainstream' clinical department in a hospital, specialist genetics centre), profession of pilot lead (medical geneticist, mainstream physician, GP, nurse, manager), characteristics of the area served (urban/rural, socioeconomic profile, ethnic profile), and so on. Potential cases were assessed by members of the research team on the basis of documentary materials provided by the pilots, and preliminary interviews with pilot leads in every site. Following this, the research team selected its sample of 11, which included four (of a total of seven) cancer-genetics pilots aiming to implement the Kenilworth model. The focus of this paper is on the pilots in Derton and Nottley, which were both based in hospitals and served areas covered by their cancer networks. The other two cancer-genetics cases were considerably smaller in scale, each serving the population of the single PCTs which hosted them, and therefore engaging less with the wider cancer networks. Whilst the issues relating to cancer networks in this paper are specific to Derton and Nottley-since no other pilot among the 11 cases was embedded within a network in this way-other findings presented (for example, the range of leadership behaviors exhibited) have much in common with the wider findings in the other casestudy sites.
In each case-study site, our research included a number of qualitative methods: in-depth interviews with pilot stakeholders, participant and non-participant observation of meetings at pilot and program level, and documentary analysis. The interviews covered a number of issues, grouped under the following headings: collaborative working, information technology, human resources management, leadership, user involvement, knowledge management, organizational culture, policy. Whilst the interview schedule included a number of possible questions under each heading, in line with in-depth interview methodology, the interviewers tended to use these as prompts only, deviating from the list of questions to discuss issues as they were relevant to respondents and their organizational contexts, but seeking to frame questions as openly as possible so as not to ‗lead' interviewees (Silverman 2004). In the course of observational research, we paid particular attention to exchanges of knowledge between involved parties, and barriers identified in the course of meetings to collaboration and service reform in their ‗mainstreaming genetics' ambitions. Documentary analysis encompassed bid documents, reports to sponsors and evaluation reports, and focused on the same issues as those explored through interviews and observation, with particular attention paid to change through time.
Some 88 interviews have been conducted across the 11 sites, including 12 each in Derton and Nottley. Interviews were tape recorded and transcribed in their entirety, and lasted between 40 minutes and three hours. In Derton and Nottley, similar respondents were interviewed, reflecting the similar aims and organizational contexts of the two pilots. They included pilot leads (medical geneticists in both cases), nurses and genetic counselors involved in delivering the pilots, primary-care based practitioners such as GPs, managers and clinicians within the cancer networks and the host hospitals, and service users involved as members of the projects' steering groups. Data drawn from observation and documentary-analysis activities both informed these interviews and complemented them, by illuminating the approaches taken to establishing service reform within wider networks, and the barriers encountered in attempting this.
We undertook an iterative analysis process, rereading and coding transcripts, notes and documents, generating themes, and cross-checking these through discussions between authors. Thematically related parts of the embedded analysis in each data source were grouped together. Each of the three authors had engaged in fieldwork, and all three engaged in analysis. In Derton, GPM developed an initial analysis which was checked by GC; in Nottley, RF developed an initial analysis which was checked by GPM. We discussed the coding of transcripts, document and meeting notes with each other, ensuring inter-researcher reliability of interpretation and enhancing analysis. Subsequently, the analysis agreed across the authorial team for each case was considered against the over-arching research questions. As a means of triangulating this analysis, findings were presented to both the commissioners of and participants in the research, providing a check on the authenticity of our analysis (Yin 2003).
In presenting qualitative analysis, a number of approaches are possible, each seeking in its way to produce a summary that is both comprehensible and explicitly derived from primary data, and thus amenable to critical appraisal by the reader (Eisenhardt 1989). Our approach in this paper involves presenting the narratives from the two cases examined in turn: in-depth descriptions that deploy the primary data to illustrate the dimensions of leadership in the establishment of service reform they emerged in the cancer networks of Derton and Nottley, which reflect many of the key findings from the other nine case-study sites.

Results: Derton
This pilot was led by a specialist geneticist, but he was keen to ensure that the service was delivered through and ‗owned' by the cancer network, which covered three acute hospitals (comprising two cancer units and one cancer centre) and 10 PCTs. Genetics provision here had previously been delivered from another hospital, some 40 miles from the cancer center: consequently patients and clinicians in the network had seen genetics as a remote specialty, its perceived esoteric knowledge compounded by its geographical distance. Recently, however, a new genetics ‗satellite unit' had been set up in the hospital that hosted the cancer centre, and the geneticist leading the project had been appointed to a post here with a view to improving genetics coverage in this underserved area. The prior inaccessibility of cancer-genetics expertise had led to the setting up of a plethora of local protocols within the network for dealing with suspected inherited cases of cancer. This variation, together with the arrival of the genetics satellite unit two years earlier, had created the political will within the cancer network for a more coherent way of dealing with suspected family histories of cancer, and an opportunity to do something about it. A colorectal surgeon (and, notably, the current medical director of the cancer network) described the dissatisfaction with the incumbent system: -The ones with the really high risk we would have sent to [the genetics service], obviously. But there's a lot of them who didn't fit the high-risk criteria, for example for bowel cancer, and we give them advice which-you're not quite, 100 percent sure whether it is scientifically based.‖ Colorectal surgeon Clinicians in cancer-related specialties outside genetics had taken on responsibility for genetic risk assessment. But they encountered criteria for assessing genetic risk in breast, bowel and ovarian cancer which were far from universally agreed upon, and limited access to specialist clinical knowledge which might assist adjudication. Here, then, was a potentially receptive structure in terms of the attitudes of the network's members, with political will exemplified by the attitude of the colorectal surgeon and medical director of the network quoted above.
Having taken up responsibility for genetic risk assessment reluctantly, existing services had nevertheless grown into an important part of cancer-service provision locally. Growing patient numbers and dissatisfaction over the variation in protocols saw the issue rise in importance within the cancer network, for practitioners and managers alike. Increasing numbers of patients were being admitted to screening services, without a ‗proper' (i.e. universally agreed) system of weighing the benefits and risks that might accrue from this radiation exposure. Moreover, this was without the explicit agreement of the primary-care purchasers who paid for the service, and this was a concern for the hospitals providing the service, who feared that in the absence of an agreed contract, funding might be withdrawn. As with the attitudes of clinicians, the organizational context too, then, was potentially receptive to the kind of change in care pathways augured by the Kenilworth model.
Consequently-and in contrast to the experience of Nottley below, and indeed of other case-study sites not detailed here-the cancer network was proactive in seeking to establish the pilot as an integral part of local cancer care pathways, as a means of solving these issues. The pilot's lead clinician, meanwhile, was acutely aware of the need to ensure ownership by the This is a pre-copy-editing, author-produced PDF of an article accepted for publication in Journal of Public Administration Research & Theory following peer review. The definitive publisher-authenticated version is available online at http://jpart.oxfordjournals.org/. © 2009 The authors network, given his status as an ‗outsider' associated with the distant genetics service. He sought to ensure ‗ownership' of the service by the network, and adopted a highly dialogical and diplomatic stance in putting his ideas to colleagues within the network: Lead: [Previously] there was no continuity, and no discussion between genetics and the surgeons. I think if any lesson's come out of this project, it's that this sort of very detailed day-to-day discussion underpins a successful service. If you don't have that networking, then you can't sustain a service. GPM: The discussion between geneticists and their referrers? Lead: Yes, the stakeholders, the ability to sit down with somebody and explain exactly what you want to do, and for them to be able to pick up the phone and say, ‗I've got this patient: what do you think?', any different niggles that happen. If you've got someone who's not happy, it's easy to go and see them or pick up the phone: it's just simple human communication.
Lead (clinical geneticist) Through these kinds of forums, in Crosby and Bryson's (2005) lexicon, the clinical lead provided a vision in a politically sensitive way. These dialogical, quiet leadership skills had some success in paving the way for the service reform, as confirmed by the geneticist's new colleagues: -It sounds a bit like I'm blowing [his] trumpet, but seriously, he was a knight in shining armor, and was really swamped with demands from people like me, who wanted to have this pathway set up, like, yesterday. It was very efficient very quickly. And [he] is a particularly good public-relations individual: his ability to make complicated genetics understandable is almost unique.‖ Gynecological oncologist In persuading physicians in these forums, then, a ‗quiet' approach to leadership was important, alongside firm, quantitative evidence of the potential benefits of the service reform in terms of patient outcomes and conforming to national guidance on risk assessment: -speaking their language,‖ as the pilot lead put it. In Crosby and Bryson's (2005) terminology, this was about visionary leadership using forums adroitly in order to create a compelling, shared vision of service reform, and utilizing existing -norms of relevance and pragmatic communication, media and modes of argument‖ (Crosby and Bryson 2005, 120) to good effect.
Credibility with physicians, through quiet leadership and the language of clinical audit, was however only one aspect of consolidating the place of the service within the network. Equally important was winning over the nurses who had taken on responsibility for risk assessment under the extant system, on whom the pilot relied for much of its throughput. This rather different professional group necessitated a different approach, requiring that the entire pilot team be versed in the arts of quiet leadership and negotiation. So one of the pilot's nurses took on the task of delicately persuading other nurses about the improvements that the new system might bring, deploying her understanding of -nursing culture,‖ as she put it.
Thus the pilot team members adapted their leadership to the different constituencies to be persuaded. ‗Objective evidence' was a key means of convincing medical specialists. Nurses, as the delivers of the service, were generally more concerned about the effects of the new model on the patient experience. Persuading each group, though, was crucial to the pilot's success, and each group could be engaged through its own informal and formal forums, with their own unwritten rules of engagement. And once key physicians and nurses elsewhere in the network had been convinced of the worth of the new system, they themselves were likely to spread acceptance among their colleagues. A breast-care nurse specialist in one of the cancer units, who had previously taken on much of the local risk-assessment work, explained how her colleagues had had concerns that the new care pathway was -just a way of getting people out of the system.‖ Following her own discussions with the lead nurse on the pilot, she was now convinced of its worth, and had been instrumental in persuading other nurses in the unit, acting as a -link‖ between the pilot and her colleagues. In this way, the initial visionary work of the pilot team members as leaders gave rise to a more diffuse leadership, in Huxham and Vangen's (2000) terms enacted through the structures and processes of the network as well as embodied in the wider participants beyond the core leaders from the pilot team.
The success of this process also rested on the effective structure of the cancer network, recognized by respondents as being functional, democratic and consensual, at least compared with others. 6 Prior to the start of the pilot, the network had convened a ‗visioning event', attended by stakeholders (physicians, nurses, managers, patients, others) from every hospital and PCT in the network, and this was seen to have promoted the sense of ‗network ownership' so desired by the pilot lead. This, then, had provided an important forum for the development of a collective vision. More generally, the pilot lead highlighted the role of the network in establishing the service across the cancer units, citing the leadership of one network manager in particular: -When I met her, I thought, ‗Goodness, this woman is so networked!' She's been a senior nurse for years and she knows everyone. She knows the system: she's got an amazing knowledge of how to exert mild pressures here and there to achieve different aims.‖ Lead (clinical geneticist) Alongside the receptive structural context provided by the network and its forums, and the delicate, visionary leadership work of the pilot team, the pilot's establishment also rested on the slightly more directive leadership enacted by this manager. Occasionally, individual clinicians were reluctant to accept the implications of the pilot for their own practice, despite the efforts of pilot staff through quiet leadership and audit evidence. On such occasions, the consensus of the wider network, once achieved, could be deployed as a means of persuasion in its own right, as the manager explained: -The objectivity of the network is useful to say, ‗You're not supposed to be doing that kind of work on patients': we bring intelligence from elsewhere in the network.
[…] Often just by bringing that objectivity, it can be made to work, by bringing in other clinical views.‖ Cancer-network manager In these ways, the collaborative rhetoric of the notional ‗network' was made real. Where collegiate, professional leadership through the network's structures and processes did not succeed, a certain amount of more directive, manipulative, even ‗thuggish' (Vangen and Huxham 2003) leadership could fill in, using this -objectivity‖ of the network as a tool of persuasion of more reluctant actors.
The network itself, then, and the shrewd and varied leadership tactics of various agents within it, was effective at gaining sign-up willingly or reluctantly from a range of hospital practitioners. Engaging with primary-care practitioners, though, was less straightforward. Crucially, primary-care stakeholders were not such central actors in the network, with its mandated focus on service rationalization. The pilot had held various educational and publicity events for GPs and community nurses, but ensuring longer-term engagement was a challenge. Notably, there were fewer forums at which primary-care actors might be engaged; furthermore, there were fewer obvious common interests between the pilots and GPs, on which visionary leadership might draw. To this extent, primary-care practitioners were on the fringes of the network, both in terms of its ‗infrastructure' and in terms of the commonalities of interest on which it was built.
Similar difficulties were present in the pilot's efforts to obtain ongoing funding, which involved dealing with purchasers who were also based in primary care. Though it had sign-up ‗in principle' from purchasers from the start, making post-pilot financial arrangements was not straightforward. Notably, although formally included in the network, those in charge of purchasing decisions in primary care were also responsible for funding a range of services beyond cancer. Consequently, negotiations about ongoing funding were subject to a range of considerations emanating from outside the cancer network. In Crosby and Bryson's (2005) terms, this arena was one which was subject to a range of different authorities and influences, including not only competing services but also centrally determined guidance on funding priorities. Furthermore, reorganization and management changes within primary care meant that not all purchasers were willing to follow through on their original commitment to the pilot; and indeed only about half the PCTs in the network eventually funded the service and care pathway on an ongoing basis. This partial success highlights the fact that the structure of the network, and the styles of leadership this facilitates, is limited where it meets other principles of organization and governance, a theme which we will take up in the discussion. And as we discover in the next section, this space of operation can be even more constricted by the overlapping imperatives of other organizational forms.

Results: Nottley
This was a joint pilot led by two geneticists at neighboring hospitals, part of a large cancer network containing six hospital trusts and covering a total population of around two million (twice that of the network in Derton). The pilot involved two, largely separate, projects aimed at increasing the accessibility of cancer genetics in primary care. Here we focus on one of thesewhich used advertising in GPs' surgeries to attract self-referrals from concerned individuals, especially those from ‗hard-to-reach' ethnic-minority backgrounds-though many of the issues we raise apply to both. In common with Derton, both projects followed the core principles of the Kenilworth model to rationalize care pathways. Given the primary-care focus, however, the principal constituencies to whom they needed to appeal were quite different. The catchment of each project was also smaller than in Derton, partly because of concerns about the risk of overwhelming the genetics service with new referrals by ‗opening the door' in primary care.
As in Derton, the pilot had obtained sign-up from the cancer network, but here the network was somewhat less proactive in encouraging and spreading the work of the two projects. There was a sense that the cancer network was preoccupied with conforming to centralgovernment targets, and with the need to find funding for existing services from cash-strapped purchasers: -Because it hasn't got a definitive work stream or target attached to it, no-one has said to me in the three-and-a-half years I've been here, ‗Account for what you're doing on cancer genetics'-and they ask me that about lots of other things. I've got to have action plans on a million things, but cancer genetics isn't one of them.
[…] If I go to the [purchasers] with any service development, unless it is linked to a target or is completely unavoidable, they just say, ‗You must be mad.' What they say-which is true-is: ‗We haven't got the money to run the services we've got: why would we fund new services?'‖ Cancer network manager Compared to Derton, the network seemed much more governed by hierarchical principlesespecially conformity to centrally determined priorities-and this was reflected in the nature of the leadership enacted by this network manager and others.
The pilot thus sought to forge its own relationships with primary-care practitioners and the target communities, but with mixed success. The project successfully negotiated space in the premises of a number of GP practices to carry out its surgeries and to display advertising materials. But contact with GPs went little further than this, due to difficulties similar to those experienced by the pilot staff in Derton, around the lack of buy-in from primary-care professionals. So pilot staff decided to concentrate on self-referrals, exempting GPs from the care pathway and avoiding the need to bring them onboard. However, engaging the target communities in the pilot also proved problematic, and so efforts to increase uptake from minority-ethnic groups had only marginal success: Administrator: It's always been the aim but it hasn't necessarily come to the outcome we would like. RF: In terms of? Administrator: Because gaining access to some ethnic minorities is very difficult. I still think there's a lot of understanding yet to be achieved as to why we don't get these people coming forward. With little experience of such community-engagement strategies, inseminating interest among the public was problematic.
Efforts, then, to engage the key stakeholders-potential patients and GPs-were marginal, by accident and design respectively. This was in contrast to Derton, where the cancer network enabled the pilot to gain buy-in from clinicians within hospitals whose direct importance to the service (as referrers) matched their indirect influence (in creating a critical mass of consensus and support among the powerful constituents of the network). 7 An absence of common interest was exacerbated by the absence of the opportunities for effective leadership that the structures and processes of the network might provide. Over time in Nottley, the realization dawned that given a relatively low throughput of self-referrals, buy-in from primary-care practitioners (especially GPs, with their influence on purchasing decisions) was something that the project needed after all: GPM: Is it an issue if GPs don't know about it? Do they need to know? Genetic counselor: Yeah, they do. For any possibility of extending the service, they do. Every person we've seen has had a letter, […] copied to their GP, so the GP knows they've been seen as part of the project. But it's just along with the other thousand [things] they look at per day! That's the side that has been a little disappointing, I thought we might get more involvement and a bit of ownership of the project amongst the practice nurses, and the other staff in the GP surgeries.‖ Pilot genetic counselor The pilot staff had little experience of engagement with primary care, with its rather different set of pressures, policies and relationships. The same culture shock had affected Derton to some extent, but there the project could rely on the proactive work of the cancer network or amenable groups of hospital-based stakeholders. Within Nottley, the structure of the network did not support the pilot team's forlorn efforts at influencing practitioners. The managerial priorities of the network were compounded by the fact that primary-care practitioners, though formally part of the network, were also subject to many other pressures and priorities.
Furthermore, the pilot staff lacked experience of the competitive funding mechanisms for this kind of project, being based in a specialist genetics department which received its funding through separate arrangements. Bidding for money from mainstream purchasers was thus a foreign experience for the project lead, and the departmental business manager. So, as a cancernetwork manager explained, the pilot was -outside the mainstream cancer community‖ of the wider hospital, and therefore marginal to cancer clinicians, the cancer network, and purchasers. The manager continued that although she wanted the project to work, -in terms of my overall job [in the cancer network]-the things I am meant to lose sleep about-it is quite a small part.‖ The pilot, then, found itself falling between stools. To use Crosby and Bryson's lexicon, it was estranged from the core concerns of the cancer network, and the opportunities for effecting visionary leadership through the network's forums. Like Derton, it was also estranged from the primary-care-based purchasers. Unlike Derton, though, it lacked the critical mass of support from stakeholders within the cancer network, which had provided Derton's project with sufficient political leadership and leverage to succeed in the arenas outside the cancer network which determined decisions on ongoing funding.
Consequently, despite various applications, neither this part of the pilot nor the other obtained ongoing funding. The absence of money or political will across the cancer network led the project lead to surmise, rather gloomily, that proactive efforts to lead this kind of reform were doomed to failure. It would only be when genetics centers began to place directive limits on the numbers of referrals that purchasers would be forced, reactively, to engage with the need for reformed care pathways and effective demand management.

Discussion
The divergent trajectories of these ostensibly similar service reforms, based on a common carepathway model, can be traced to a number of interacting factors. Both sought to introduce service-delivery reforms to improve client focus, and differences in the adaptation of this ambition to existing provision were significant in this divergence in themselves. However, the divergent trajectories also related to differences in the contexts of the two pilots-particularly the networks-and the agency of those seeking to implement them-notably, the presence or absence of particular forms of leadership among pilot staff and wider stakeholders.
In both cases, the pilot staff attempted to draw on the resources of networks that should, in principle, be supportive of organizational reforms to rationalize care pathways, improve patient outcomes and reduce inconsistency of provision. In practice, though, and reflecting the findings of Addicott et al. (2007), the network in Nottley was dysfunctional in this regard, with a focus on performance management and efficiencies that squeezed out any non-mandated initiatives. In Derton, the cancer network was more supportive of the pilot. In part, this seemed due to the more productive relationships already present, but it was notable also that the issue addressed by the pilot was already high on the political agenda-not least, due to financial concerns. This confirms the important point made by Van de Ven and Rogers (1988), that we should not overemphasize the role of organizational resistance in impeding innovation diffusion. Rather, it may be more instructive to view organizational resistance as the aggregate resistance of members, or the inappropriateness of the innovation to the core concerns of the organization, as seemed to be the case in Nottley ( Van de Ven and Rogers 1988). In Derton, in contrast, the pilot was aligned with the managerial agenda of the wider network, indicating that the susceptibility of networks to top-down pressures (Keast and Brown 2006;Addicott et al. 2007) is not necessarily an obstacle to collaboration within them (cf. Provan and Milward 1995;Moynihan and Ingraham 2003;Olsen 2006)-provided those concerned know which -mild pressures‖ to exert, as the lead here put it. Rather, there was complementarity between the collegiate, ‗quiet' leadership embodied by the pilot team and the slightly more directive leadership deployed by the network manager, who recognized the synergy between the aims of the network and those of the pilot. Here we echo the findings of various authors about the diversity of skills needed by leaders in networked settings (Riccucci 1995;Huxham and Vangen 2000;Vangen and Huxham 2003), the importance of influential champions to innovation diffusion (Rogers 2003), and indeed the need for leaders with different skills and organizational positions in relation to different leadership tasks (Shortell et al. 2000;Crosby and Bryson 2005). Furthermore, we highlight that publicservice networks in particular, with diffuse objectives and loci of power, may require this combination of leadership styles, and all the more so if they are professional bureaucracies with the extant complexities of power and accountability that these entail. Especially important in this context was the ability of the pilot team to make their service relevant to the concerns of others, increasing the likelihood of acceptance (Harrison 1998;Ferlie et al. 2005;Fitzgerald and Dopson 2005) by creating, as Crosby and Bryson (2005, 115) have it, -communal stories that help diverse stakeholder groups develop a sense of what they have in common with each other and what they This is a pre-copy-editing, author-produced PDF of an article accepted for publication in Journal of Public Administration Research & Theory following peer review. The definitive publisher-authenticated version is available online at http://jpart.oxfordjournals.org/. © 2009 The authors might do to tackle common problems and create a better future.‖ This was a crucial way in which heterophilous links, in Rogers' (2003) terms, were created to ensure that the Kenilworth model was adopted and accepted by diverse stakeholders. Evident in Derton was the way in which the structure of the network facilitated effective leadership through formal and informal forums which provided pilot members with the opportunity to foster a collective vision with other actors, mirroring closely the role for visionary leadership articulated by Crosby and Bryson (2005). Through the work of the pilot team in engaging with others through these forums, and the directive leadership of the network manager, they were able to make effective use of the three ‗leadership media' identified by Huxham and Vangen (2000): the structure of the network provided a normative tool that could be used to bring ‗deviant' practitioners into line; the processes of the network offered the opportunity for the development of a shared vision; and, crucially, through other participants in the network, the vision was spread further and became established across diverse stakeholders. Central to this, then, was not just the quiet, distributed leadership within the pilot, but also the quiet, dispersed leadership (Buchanan et al. 2007b;cf. Hartley 2005) of stakeholders in the wider network. The former was to some extent under the control of the project, whereas the latter was more diffuse, but nevertheless important to widespread uptake. To this extent, then, the agency of the leaders relied on the structures (and processes and participants) of the network. This finding also reflects the intimations of Alexander et al. (2001) on the importance of what they call ‗collateral leadership' among both staff and community groups in the community-care networks they study, as a complement to-not a substitute for-more vision-based leadership. Our findings here converge with and develop existing literature in Europe and North America on distributed leadership in health-care settings (Denis et al, 2000;Huxham and Vangen 2000;Buchanan et al. 2007a), local government (Hartley and Allison 2000) and education (Leithwood et al. 1999;Spillane et al. 2001;Gronn 2002). These studies illustrate the importance of subtle, distributed forms of leadership in settings characterized by the ambiguity of objectives and power relationships that typifies public-service organizations, particularly professional bureaucracies. Our study shows how, with the added complication of the inter-organizational network, multiple forms of distributed and dispersed leadership are required to engage and inculcate diverse, powerful stakeholders, given the powerlessness to achieve change of any single actor or group of actors. This combination of leadership styles-and, perhaps more importantly, leaders-seems from our study crucial to achieve transformative change in such a context.
Thus the structures, processes and participants of the network were essential resources for the effectiveness of leadership. It is in this sense, perhaps, that, as O'Toole (1997, 49) has it, there is a need for -a consideration of networks as causal forces in the administrative setting.‖ Certainly, the efficacy of leadership in Derton relied in part on the opportunities and normative framework provided by the network. However, what is also clear, from the comparison with Nottley in particular, is the impotence of the network form in itself. Without the will and the agency of its actors, as expressed through leadership at various organizational levels, the network has no causal force. Formal structures, as scholars such as Lipsky (1980), Sandfort (2000) and Riccucci (2005) have found, have little determining effect on the behavior of front-line staff, especially where those staff enjoy professional autonomy. As Sandfort (2000) suggests, it is the informal structures of professionals' collective schemas of understanding that are determinant of behavior, not the reconfiguration of formal, organizational structures. As such, transformative agency is crucial, embodied in leadership of some form, even as leaders employ the structures, processes and participants of the network in achieving their ends. To this extent, the policymakers seem correct in their assessment of the need for complementarity.
However, what our findings also illustrate are the limits to the space in which this synergistic relationship between structure and agency can operate. In Nottley especially, but to some extent also in Derton, the influence of coexisting structural forms was always present. This was most obvious in relation to physicians and purchasers in primary care. The division between primary-care and hospital physicians has been noted as a significant barrier to integration elsewhere (Grumbach et al. 1999;Shortell et al. 2000). The purchaser-provider split is a ubiquitous one in networks in any field which is at least partly characterized by a contractual, market or quasi-market relationship . Though formally part of the network, it was clear in both sites that physicians and purchasers were susceptible to neither the principles of collaboration nor the performance of leadership in the way that the network's more central stakeholders were. It is evident, then, that the space created for effective leadership by the structural intervention of the network is a partial, constricted one, that is less open in the margins of the network where there are competing pressures on behavior from other organizational forms, such as hierarchical performance management and competition between providers. Crosby and Bryson (2005) recognize this when they identify the rather different leadership skills required in arenas and courts as compared to forums. Notable in relation to their example of the African-American Men Project in Minnesota was the administrative, decision-making power borne by a key leader as an elected official, enabling him to enact political leadership in the executive arenas concerned (Crosby and Bryson 2005). Nottley's lack of such political leadership saw it fail in its bid for resources in the arena of the purchasing decision-makers; Derton's project almost met a similar fate, saved only by the collective political clout of the network's powerful stakeholders. The more general, theoretical point in relation to Crosby and Bryson's (2005) categorization of effective leadership strategies is that while forums may be susceptible to the shaping influence of leaders within a network, arenas and courts frequently are not. As such they require not just a different style of leadership, but potentially also a different set of leaders altogether, with legitimacy and leverage within their domains.
Our contribution to theory may thus be summarized as follows. The need for distributed leadership, highlighted in the literature in relation to various public-service professional bureaucracies, is compounded by the inter-organizational network form. Leadership needs to be dispersed as well as distributed, to engage diverse, powerful stakeholder groups, and this implies a variety of leaders as well as leadership styles. The inter-organizational network itself is ineffective as a means of achieving change, but it does provide the space and the media for leaders to effect change. However, this space is a constricted one, and at its boundaries are other organizational forms, with their pressures on actors and corresponding norms of behavior. Consequently distributed and dispersed leadership are often not enough to achieve change, if change rests also on the decisions and behaviors of those governed by the pressures of these other organizational forms. The arenas and courts outside the network, in Crosby and Bryson's (2005) terms, may require not only a different form of leadership, but also a different tranche of leaders. This last point may relate especially to countries such as the UK and much of continental Europe, where the remnants of traditional, centralized systems coexist alongside the network form, though as we have seen, it is also relevant in relation to the coexistence of networks and markets.
On a practical level, certain specific lessons for the health sector also have broader implications for the operation of public-service networks more generally. In the health services of the UK, US and other countries, -the relative power of specialists has diminished as primary care physicians have assumed a more central role as the initial contact point with patients and as major decision makers regarding referrals to specialists and sites of care throughout the system‖ (Shortell et al. 2000, 79). In the UK, policymakers are currently seeking to consolidate this newfound power of primary-care physicians and purchasers, so that they might become the drivers of health-service change. However, evident from both these studies was the degree to which primary-care actors were sidelined within the networks and in service reform. The marginality of this locus of power was instrumental in Nottley's failure and almost derailed Derton's project too, and in more general terms, the need for policymakers to align incentives and levers, through integration or centralization within the network, is strong (Provan and Milward 1995;Milward and Provan 1998;Provan et al. 2004;Keast and Brown 2006). Client-level networking between practitioners (Provan and Sebastian 1998) needs to be supported by organizational integration that supports practitioners' efforts and enables the establishment of change in the face of prevailing institutional pressures. Inter alia, this means that notwithstanding their contractually based relationship, organizations on both sides of purchaser-provider relationships need to be incorporated in the structures, and inculcated in the ethos, of the network.

Conclusion
Our study is based on relatively thin empirical data, and so should be interpreted cautiously, especially in terms of the additions to existing theory suggested. Many of the findings, though, confirm existing evidence; moreover the findings on the importance of distributed leadership from these two case studies are mirrored in data from our wider study of 11 sites. Our findings indicate that despite the well documented obstacles to networked governance and distributed leadership in the public services, a careful alignment of objectives with managerial agenda can bring success to reforms in service provision. A combination of clear benefits to wider stakeholders within the network, and distributed and dispersed leadership, can give rise to effective collaboration and establishment of reforms, through structural integration and the harnessing of agency. To this extent, policies invoking structure and agency were indeed synergistic. However, the relationship was a contingent one, illustrating how network-based reforms to organizational structure are both potentially powerful and simultaneously impotent: powerful in creating a space within which certain enactments of leadership might flourish and achieve service reform; impotent as instruments of that reform in themselves, as a means of changing practice in the face of the much more powerful, informal structures that are determinant of behavior (Sandfort 2000;Riccucci 2005). Furthermore the space for effective leadership created was a limited one, which became more and more constricted in the face of coexistent organizational forms and modes of governance. Networks compound the need for distributed leadership in public-service contexts since they introduce new loci of power that might be influenced by a more dispersed form of leadership. However, the coexistence of networks alongside other organizational forms constricts the effectiveness of this mode of leadership, and it is likely that in many settings, complementary modes of leadership that conform with the requirements of these other organizational forms may be required.
Our study adds to a growing literature on the practice of leadership and the role of networks in public-service organizations, and brings these literatures together to highlight the potentials for and impediments to leadership in public-service networks. As we noted in our opening section, as policies focused on transforming the structures of public-service provision and the agency of public servants respectively, networks and leadership seem to some extent complementary. The tangible success of distributed, quiet leadership, and dispersed change agency, in one case, indicates how these two policies can indeed work synergistically. The less successful efforts of the second case, illustrate the fragility of this complementarity.
In terms of policy, our study reaffirms the importance of carefully aligned policy levers that ensure that organizations within networks have drivers which incline them towards networked collaboration rather than operation within their silos. In health, the gap between hospital and primary care would seem to be a particularly important one in terms of divergent drivers and illaligned incentives, reflecting the uneven influence of network-based organizational forms in other sectors, especially those characterized by market or quasi-market relationships. For organizations and practitioners, then, the key message is around ensuring the functionality of networks, and the importance of locating those crucial change agents whose influence might be harnessed towards service reform. Finally, we hope our study illustrates the distinctive contribution of qualitative study (cf. Bryman 2004), even as it highlights the need for further investigation of the processes by which individual agency interacts with structural reconfiguration