Keywords

Introduction

A large body of work has investigated organizational change processes, and scholars have come to agree that new work practices become established in organizations through either macro- or micro- level processes (Reay et al., 2006). Macro-level  mechanisms contend that the uptake of new practices is due to external forces and leads to organizational development over a long period. Here, scholarship has been criticized for ascribing a disproportionate emphasis at this level of analysis and ignores the purposive actors who are responsible for changes in the organizations in which they are embedded (Harmon et al., 2019). Further, a macro-level emphasis prevents developing theories that would support deeper insights into the role that individuals and collective action play in the uptake of novel practices. As such, the dominance of this perspective has sparked interest in gaining a deeper understanding of the micro-level influences responsible for the acceptance and spread of new ideas (Reay et al., 2006). At this level, the uptake of new practices involves workplace relationships such as the interactions and negotiations of various actors (Bridwell-Mitchell, 2016). In practical terms, actors use workplace relations to make modifications to their practice (Gray et al., 2015) or search for and implement “new ways of carrying out specific activities” (Smets et al., 2012, p. 894) in their efforts to complete jobs or accomplish a common goal.

Specifically, in the initial stages, when a new practice is shared and enacted, actors acknowledge that they are “doing things in new ways” (Reay et al., 2006) and evaluate the suitability of a new practice in addressing a problem with which they are faced (Bridwell-Mitchell, 2016; Dorado, 2005). New ways of working may require actors to move away from the influence of the macro-level by persuading others to adopt and engage in new practices that may diverge from the established norms in their organizational environment (Battilana & Casciaro, 2012). However, due to the interplay of macro-level influences and the micro-level interactions, a puzzle emerges that raises a cause for concern for the acceptance and spread of new practices. If actors are embedded in workplace relations and are subject to “processes that structure their cognitions, define their interests and produce their identities, how are they able to envision new practices and then subsequently get others to adopt them?” (Garud et al., 2007, p. 961). We know that embedded actors develop dense networks of relations and affiliate with organizational norms (Granovetter, 1985) and that the adoption of new practices acquires a sense of inevitability (Garud et al., 2007). However, we know very little about how embeddedness in workplace relations supports the adoption and spread of new practices.

In particular, one specific avenue for research to grapple with this puzzle is to understand the nature of workplace relations, particularly the interactions and behavior that are likely to influence the acceptance or rejection of existing practices, and effects on the spread of new practices within an organizational setting (Tolbert & Zucker, 2019). While organizational research has been enriched with insights from studies on workplace relations and organizational change, there is much to learn about the potential mechanisms of workplace relations that could explain how they influence the uptake of new practices. This is a concern for the process of structural embeddedness of individuals (Granovetter, 1985) in workplace social networks, and how this embeddedness may support or constrain the opportunities for actors and organizations to bring about organizational change (McGrath & Krackhardt, 2003). However, we have little understanding of how actors use their embeddedness in workplace relations in their attempts to influence changes within their organizational environment when new practices are introduced. In particular, the key concept of structural embeddedness is greatly overlooked for its potential to explain such an influence (McGrath & Krackhardt, 2003). Therefore, we explore how structural embeddedness serves as a foundation to support micro-level change. Our research attempts to shed light on the features of structural embeddedness among actors who seek to introduce and integrate new practices within their organization that sits between the macro-level and micro-level influences (Kilduff & Tsai, 2003), i.e., the way in which actors affect organizational outcomes and how macro features affect actors. In this study, we pose the following research question: how do actors’ embeddedness influence the acceptance and spread of new practices in organizations?

In the next section, we introduce our notion of structural embeddedness and a social network perspective as our starting point to elaborate on the role of workplace relations and organizational change, and then develop hypotheses on the relationship between actors’ relations and the uptake of new practices. To conduct our inquiry, our empirical sites selected are NHS hospital trusts (in the UK) that are partnered with the Virginia Mason Institute (VMI) to introduce and integrate novel Lean methods within these organizations, where new practices are expected to enhance the quality and provision of patient care, generate operational efficiencies and to create a culture of continuous improvement. We then describe our social network approach. As such, for the study, we collected social network data from a range of health professionals with different statuses and operating at different levels within the organization. We test our hypotheses on structural embeddedness generated from this data. The data is analyzed with Exponential Random Graph Models (ERGMs), an emerging social network analysis method, to examine joint effects between actor roles and relationships to understand embeddedness and interactions among organizational actors. In the final section, we discuss our contributions to the literature and practice.

Perspectives on Embeddedness

To explore the adoption of new practices in organizations, this study draws upon two streams of scholarship: embeddedness and a social network perspective. Embeddedness provides a key link between actors’ relationships and micro-level change (Tolbert & Zucker, 2019). If we assume that actor embeddedness and interactions are related to the likelihood of micro-level change (Garud et al., 2007), a counter-argument can be presented which explains that due to the interplay of actor embeddedness and organizational structure, actors can change the organizations in which they are a part of (Bridwell-Mitchell, 2016). Dacin et al. (1999) explain that embeddedness can allow individuals to shape their context in ways that allow actions to occur by serving as “means of stratification by opening windows of opportunity for some while erecting barriers” for others (p. 335). This perspective widens the scope and contrasts the long-standing argument that individuals’ actions are constrained by their embeddedness and existing organizational arrangements (Battilana, 2006).

Social network perspectives have a long tradition in examining embeddedness and how social structures, groupings, and positions among actors influence both opportunities and constraints for action (Burt, 2001). A core assumption for this perspective is that patterns of relationships matter because actors take on identities and give meaning to social action through their relatedness to others (Kilduff & Brass, 2010). A social network notion does not give prominence to actors’ inherent characteristics or attributes, but rather the emphasis is placed on the relationships among a group of actors. Based on this understanding, social relationships and interactions have been comprehensively examined in the past, and it is widely accepted that the informal networks of actors are built from the day-to-day interactions and communication within an organizational setting (Tasselli, 2015). These social structures both shape and explain organizational outcomes and processes. Thus, extant research emphasizes that social structures emerge from patterns of social interactions and relationships, where different structures and ties have distinct functions that play a significant role in the achievement of organizational outcomes (Ibarra et al., 2005).

Regarding embeddedness, at one extreme, there are over-socialized accounts of actors who lack agency, are constrained by their environments, and do not envision and enact change (Felin et al., 2012). At the other extreme, there are heroic accounts of actors who single-handedly envision and enact change, despite their embeddedness and the surplus of forces responsible for an organization’s persistence and stability (Powell, 2019). These contrasting perspectives suggest that traditional views regarding the embeddedness of actors are exaggerated and conflated and that by distinguishing a form of embeddedness, we can argue that it does not just constrain agency or action, but it also serves as a fabric to support change (Reay et al., 2006). Likewise, our understanding of the micro-level forces of change is incomplete since “heroic actors and cultural dopes are a poor representation of the gamut of human behaviour” (Powell & Colyvas, 2008, p. 277). These over-socialized and heroic narratives ignore the complex social processes at play and the diverse interests and perspectives within organizations. Therefore, new work is needed that strikes a balance between these two extremes and accounts for a more inclusive dialogue regarding actors, embeddedness, and practices within organizations. When we focus on embeddedness and draw on social network concepts, we see that structural embeddedness gives attention to the types of relationships and structures that influence social action and organizational outcomes (Dacin et al., 1999). As such, structural embeddedness focuses on examining the social patterns and structures that emerge within the network to understand the specific social processes present in a network (Moran, 2005). It, therefore, becomes a fundamental source and necessary precondition for the adoption of new practices.

Structural embeddedness is a conceptualization of social structure based on actor ties and direct relationships (Dacin et al., 1999). It refers to the extent of connectivity, the likelihood of interactions among actors within a social space, and the extent to which individuals are anchored in closely knit social communities (Goldberg et al., 2016). Nahapiet and Ghoshal (1998) extend this definition to incorporate the configuration of interpersonal linkages between people or groups, including the presence or absence of network ties between actors and other structural features such as cohesion. Social networks and structures are considered to be built from local patterns and configurations of relationships that arise from social processes among actors (Lusher et al., 2013b). As such, structural embeddedness from a micro-perspective has centered on two broad structural forms, closure and bridging, and these have varying consequences for adopting practices among actors (Reagans & McEvily, 2008). Closure is associated with cohesive groups of closely connected actors, and bridging is associated with spanning the gap between disconnected actors; however, it is unclear what role these structures play in the adoption of new practices. We also know that closure mediates concerns among actors when they are faced with conditions of ambiguity and uncertainty (Möllering, 2014). Research points to the importance of closure as it builds interpersonal trust, which garners support for undertakings that are new to the organization (Zhelyazkov, 2018). When closure is high, actors are less likely to question behaviors, increasing the likelihood of sharing and engaging in new practices (Bridwell-Mitchell, 2015). Cohesive workplace relations also influence collective interpretations more deeply than individual actors (Beckert, 2010).

However, findings on the extent of influence of closure on the uptake and spread of new practices are equivocal. For instance, Battilana & Casciaro (2012) examined the link between structural embeddedness and varying degrees of organizational change. They found that low levels of closure in a healthcare professional’s network facilitated the initiation of changes that diverged from their organizational status quo but hindered their adoption. In another study, Birdwell-Mitchell (2016) found that patterns of interactions in large, cohesive, homogeneous groups supported the process of micro-level change compared to several small groups as professionals could converge on widely shared understandings of appropriate practice (Birdwell-Mitchell, 2016, p. 183). It was observed that the large cohesive groups fostered strong social pressures that encouraged conformity regarding new practices and reinforced feedback about how to address practice dilemmas. The smaller fragmented groups, which were considerably diverse, produced weak social pressures that did not support the spread of new practices within the various groups. These differences highlight that the uptake of new practices depended on the extent of socialization and patterns of interaction, either cohesive or diverse that allowed actors to develop more consistent and shared understandings of new practices’ technical requirements (Birdwell-Mitchell, 2016, p. 174). This work revealed that, in some contexts, professional networks are organized to allow the uptake of new ideas and practices, whereas, in others, the patterns of interactions result in “social disorganization”. Therefore, new ideas and practices “do not take root or spread” (Birdwell-Mitchell, 2016, p. 162) in low cohesive groups. We, therefore, suggest the following hypothesis.

H1

Professionals in high cohesive networks will take up new ideas or practices more so than those in low cohesive networks.

Bridging has been a major theme within social network research in recent years (Parkhe et al., 2006) building on Burt’s “structural holes” theory (1992), which focuses on benefits outside the group structure rather than within (Burt, 1992, 2004). Burt describes a structural hole as the “separation … or a relationship of non-redundancy” between two actors that enable them to “provide network benefits that are … additive rather than overlapping” (Burt, 1992, p. 18). While disconnection between actors and groups provides an essential requirement for the existence of a structural hole, some explanations are underpinned by the opportunity, ability, or motivation to bridge information or advice between disconnected actors and groups (Kleinbaum et al., 2015). This bridging behavior is represented by an open triad, where two disconnected actors are connected through their relationship with a third actor. When bridging a hole between disconnected actors, the focal actor is bestowed with distinct positional advantages between groups, as they bridge between actors or groups (Quintane & Carnabuci, 2016). Unlike closure, much more emphasis is placed on the individual, and value comes from the position and ability to efficiently span the structural hole and exploit the opportunities it creates. From this perspective, benefits, such as the uptake of new practices, accrue for both the actor and the network, as structures rich in holes capture diversity and novelty by accessing the proficiencies of actors who are disconnected from each other and who have different perspectives, skills, and expertise (Zaheer & Soda, 2009). To the extent of professional networks, we hypothesize that:

H2

In professional networks, actors will link to bridging actors and will take up new ideas or practices more so than those who do not link to bridging actors.

The outcomes are understood to be determined by the focal actor’s ability to identify opportunities, bridge the gap, and create value for themselves and disconnected groups. However, these opportunities to control the access to resources and control benefits between disconnected groups are risky, as this position is associated with the ambiguity surrounding coordination and tensions due to conflicting norms, practices, and perspectives between the groups (Obstfeld, 2005). In considering bridging in networks as a mechanism to support the uptake of practices among actors, we see another level of interaction and engagement among actors to share and adopt new practices within an organizational context.

Professional groups tend to display varying patterns of interactions that support and inhibit change initiatives. For example, West et al. (1999) found that nursing networks are centralized, allowing them to gather and disseminate information more effectively. In contrast, clinical directors and doctors have more hierarchical networks that are more densely connected than nursing networks, which allows them to be “more potent instruments for changing, or resisting changes, in clinical behaviour” (p. 633). Therefore, professional roles have a significant relationship with social structure and influence varying organizational outcomes and healthcare change initiatives (West et al., 1999, p. 633). Thus, we need to account for the role attributes of the actors. The dominant hypothesis in social network research is that actors with similar roles will be more likely to form network ties than actors with different roles. Referred to as homophily, it is consistently identified as a vital determinant of network structure (McPherson et al., 2001). The theory predicts that similarities in the attributes of actors lead to similar organizational preferences and predispose actors to cooperate (Cole & Teboul, 2004). Professional roles are important in shaping patterns of interactions within a network (Evans & Scarbrough, 2014; Fitzgerald & Harvey, 2015). However, it is unclear how the presence of both actors with similar role attributes and dissimilar characteristics will influence the context that shapes the uptake of new practices. We might predict that the uptake of new practices will be enhanced by homophilic ties. Therefore,

H3

Actors who share similar roles (homophily) are more likely to create links across structural holes, and this will lead to the uptake of new practices.

Study Context

This study is aligned with a broader research initiative that seeks to evaluate the impact of the Virginia Mason Production System (VMPS) in five NHS hospital trusts. In 2015, a five-year partnership between the NHS Improvement (NHSi), the Virginia Mason Institute (VMI), and the five hospital trusts was established to develop a culture of continuous improvement, to enhance the organizational culture and quality and efficiency of healthcare services within each trust (Health Foundation, 2018). The Virginia Mason Institute (VMI) specializes in a healthcare management system that promotes lean principles, developed from Toyota’s Production System (VMI, 2019). The VMPS is a lean management method that aims to change practices through small, incremental changes within the work environment that are consistently applied and sustained over long periods. This initiative intended to provide healthcare professionals with new knowledge, tools and approaches to healthcare provision, which essentially challenged the existing norms, practices, and approaches to daily work within their respective organizations.

This study focuses on two NHS Trusts (NHS-A and NHS-E) and the Trust Guiding Team (TGT), Kaizen Promotion Office (KPO), Lean for Leaders (L4L), and rapid process improvement workshops (RPIW) levels of the partnership within each organization to narrow this context’s scope due to this partnership’s complex nature. In sum, the Transformation Guiding Board (TGB) is an inter-organizational group comprised of five chief executives and senior members from the Virginia Mason Institute (VMI) and NHS Improvement (NHSI). The Trust Guiding Team (TGT) is an organizational group comprised of the chief executives of each trust, the Kaizen Promotion Office Lead (KPO Lead), and other senior trust members and directors. The TGT strategically align improvement plans and formal training that are translated to the Kaizen Promotion Office (KPO), which are the implementation teams that oversee the improvement work’s execution. The KPO has many duties, one of which involves training organizational leaders in the Lean methodology. These Lean for Leaders (L4Ls) are expected to apply and share the new methods, knowledge, and practices, with the expectation of creating a culture of learning and change that leads to efficiencies and savings within their work environments. The rapid process improvement workshops (RPIWs) refer to the specific improvement initiatives identified from the value stream mapping process, where specific interventions are suggested and initiated within a five-day period and then tracked over time, for example, ninety (90) days. The L4Ls and RPIWs collectively mobilize teams to eliminate delays, redundancies, errors and waste within the healthcare delivery process by applying the Virginia Mason Production System (VMPS).

Prior to this investigation, varying contextual influences are observed. First, there are differences in organizational performance, based on Care Quality Commission (CQC) ratings, and differences in improvement progress, based on the number of the value stream and RPIWs conducted between 2015 and 2018. NHS-A, is an “Outstanding” rated organization, therefore delivering a high level of patient care. NHS-A has average improvement progress, as they conducted a similar number of values streams (6, 5), RPIWs (23, 21), and training a similar number of L4Ls (218, 205). NHS-E, however, has a lower organizational performance rating, “Inadequate” but above-average progress, as it conducted the highest number of value streams (8) and RPIWs (38) among the networks. One explanation for this dynamic is that since NHS-E was the lowest-performing organization among the networks, it conducted more improvement initiatives to achieve an acceptable performance level. Therefore, this setting is a suitable empirical context to examine professionals’ structural embeddedness to introduce and encourage the uptake of new practices in organizations.

Method and Data

This research employs a standardized cross-sectional network analytic methodology to systematically examine the nature and structure of social relationships from local and global perspectives (Wasserman & Faust, 1994). This design is standard as it determines the prevalence of an issue, attitude, or behavior, provides a snapshot of a subject at a specific point in time (Kumar, 2011), and allows multiple concepts to be simultaneously measured to examine the underlying relationships and relevant patterns of association between them (Easterby-Smith et al., 2018). The network boundary was specified to include members from the four organizational levels of the partnership TGT, KPO, L4Ls, and RPIWs. The research adopted an ego-network approach to data collection, which identifies actors within the network boundary and directly asked by an interview or survey with whom they interact regarding advice around practice (Ibarra, 1992). For example, the respondent was asked, “Who do you go to for advice about work matters relating to new VMPS?” The sample includes clinical and non-clinical healthcare professionals at all levels of the partnership in each organization, including senior executives, clinical managers, consultants, matrons, nurses, pharmacists, radiologists, dieticians, physiotherapists, and non-clinical management professionals.

Data Collection

The study used two self-reporting instruments to collect ego-network social network data. The first approach was to gather social network data from the KPO & TGT via a paper-based socio-metric survey, and the second approach was administered via a web-based survey that captured L4L ego-networks and attributes data. Both survey instruments were designed using the method discussed by White et al. (2014, 2016), where a 5-point Likert scale is used to measure the subjective evaluations of relationships among actors. This instrument’s use in previous studies contributes to the items’ validity and robustness and has been used as the standard approach for socio-metric survey data collection (Ibarra & Andrews, 1993; Morrison, 2002). The surveys asked persons to list the name, role, and organization of at least five persons who are in some way involved in their improvement work regularly. Next, a network of relationships is constructed based on their responses. The survey instruments allowed respondents to list and rate the relationships with whom they share knowledge about improvement work.

The web-based survey was administered via Qualtrics to the Lean for Leaders (L4Ls) level of the network with the dual purpose of capturing their attitudes and perceptions regarding the Lean Methodology and their knowledge-sharing networks, as they were the actors trained to use the Lean Methodology in their daily work. In this data collection mode, the L4L respondents were actively engaged in applying the Lean Methodology in their daily work and environment, whereas persons at the TGT and KPO levels have a more strategic and administrative role in guiding improvement work. In terms of the population, the L4L represents healthcare professionals from various levels and backgrounds within each organization, including non-clinical and non-managerial roles. Finally, the web-based socio-metric survey gathered 54 responses from NHS-E and 80 from NHS-A. In the paper-based survey, a total of 16 were collected from NHS-E, and 14 were collected from NHS-A. Together, this uncovered 387 ties in NHS-A, and 279 ties in NHS-E.

Exponential Random Graph Models (ERGM)

ERGMs are used to analyze interdependencies of relationships within social networks and test the hypotheses (Lusher et al., 2013a). The use of ERGMs allows researchers to explicitly model the observed organization’s networks against theoretically informed and supported network configurations to estimate their effects. ERGMs are tie-based models that can examine multiple hypotheses regarding network-generating processes while simultaneously making no assumptions about independence among the dyads (Lusher et al., 2013a). Estimations and simulations of the observed networks are conducted using well-established statistical approaches such as Maximum-likelihood estimation, undertaken through Markov chain simulation-based approaches (MCMCMLE). In this approach, parameter estimates are determined based on differences between observed data and simulated distributions until parameter estimates achieve convergence (Robins, 2011). A more in-depth discussion of ERGMs, the model specification, the selection of structural parameters, and the model estimation is provided elsewhere (Robins et al., 2007).

Estimation Procedure

After the models are specified and the configurations are determined, the model is estimated in the PNet software package, simulating and conducting the stochastic analysis of social networks (Wang et al., 2009). It is an iterative process where configurations are included and excluded from the model until model convergence is achieved. A large positive parameter indicates that a hypothesized configuration appears with greater frequency in the observed network than expected by a random graph, given the presence of configurations related to other effects in the data and a negative parameter indicates that the configuration occurs less frequently than it would by chance (Robins et al., 2009). Since this estimation technique produces approximated estimates, model assessment and fit are based on predetermined criteria and heuristics, such as the parameter estimates, standard errors, convergence statistics, and goodness-of-fit ratios that compare the observed values with the fitted values (Hunter & Handcock, 2006).

Dependent Variable

Our dependent variable is the extent of advice exchange between participants. For each dyad in the network relations were derived by asking in the survey whether actor j advised actor i with regards VMPS. For each dyad, we asked each respondent to rate the strength of the relationship for each dyad (according to a 1–5 scale). Since our analysis need binary data, the scale scores were dichotomised as follows: if i reported a response about j which was greater than or equal to 4, this was coded as 1; otherwise, it was coded as 0. As a strategy, we tested several dichotomisation criteria (Conaldi & Lomi, 2013) and did not find any difference from the corresponding degree distributions of the original valued network.

Network Effects Variables

In terms of our hypotheses on closure, traditionally, scholars of SNA have employed aggregate level parameters to examine the level of interaction between individuals. Thus, we include density, spread, and reciprocity in our model (Hansen, 1999; Reagans & McEvily, 2003). First, the basic connectivity tendencies in a network (see Fig. 1) were captured by the density or arc parameter. We also specified Reciprocity (Reciprocity) parameter which estimates the actors’ tendency to engage in reciprocated relations with their connected actors. Next, four star-based parameters are selected to model cohesion via centralization and degree distribution effects within the network. The Two Out Star (2-out-star) and Three Out Star (3-out-star) model outgoing ties to two and three actors respectively and captures the tendency of an actor to collaborate with and seek advice or information regarding improvement from two or three persons (Fig. 1). The popularity Spread (AinS) parameter estimates prestige and influence within the network, and a negative or small estimate indicates that most actors have similar popularity levels. Activity Spread (AoutS) parameter estimates outgoing contact and interaction with other actors and indicates the extent to which an actor may seek out information or advice from connected actors. In this case, a negative activity spread parameter indicates that most actors have similar levels of activity, and the network is not centralized around a few key actors.

Fig. 1
figure 1

Configurations & parameters for ERGMs

Full size image

We also specified four parameters selected to examine the bridging hypotheses representing the possibility of structural holes co-existing in the network (Fig. 1). The Simple Connectivity (Path2) measures the extent to which actors who send ties also receive them and equates to an actor’s likelihood to broker information or advice with another actor. The One-In-Alternating Out Star (1inAout-star) measures the extent to which a connected actor sends ties to multiple other actors, which equates to the likelihood of an actor disseminating information or advice across a range of contacts and indicates the sharing of information and advice within the network. The Alternating-in-One-Out Star (Ain1out-star) measures the extent to which an actor who receives ties from multiple actors to be connected to at least one other actor, whereas the Alternating-in-Alternating Out Star (AinAout-star) measures the extent to which an actor who receives ties from multiple actors to be connected to multiple other actors. These parameters also have hierarchical connotations and influences. For example, the One-In-Alternating Out Star (1inAout-star) configuration can indicate a traditional bridging relationship and formal and informal superior and subordinate relationships, where one connected actor can efficiently communicate and distribute information across their network of contacts. This social process is expected in a healthcare setting to evidence communication and interaction between organizational leaders and their collaborative contacts.

Actor Effects Variables

Next, organizational roles were identified as relevant to the empirical context, focusing on the clinical and non-clinical nature of these roles, and the scope of these roles. Respondents were categorized based on their clinical role, where all respondents involved in direct patient care were in clinical roles, including traditional roles such as nurses, doctors, and allied healthcare professionals. In contrast, the non-clinical staff was defined as administrative and supportive roles that do not involve patients’ direct treatment or care (Seto et al., 2011). These are non-clinical management roles such as finance and human resource professionals who work in a healthcare setting. The respondents’ roles were captured from the survey, documents, and records. Using these roles, we considered actor-relation effects, i.e., the tendency for actor attributes to affect tie formation is modeled. As such, three actor-relation parameters are included in the model: Homophily ([Attr]-Interaction), Sender Effects ([Attr]-Sender), and Receiver Effects ([Attr]-Receiver). The Homophily ([Attr]-Interaction) configuration models the tendency for ties to be more or less likely between actors similar in both professional and managerial hierarchy. In this case, homophily is indicated by a positive parameter value for these effects. Next, the Sender Effects ([Attr]-Sender) and Receiver Effects ([Attr]-Receiver) configurations are included to model the likelihood of an actor attribute promoting an actor to be more active, therefore having more outgoing or incoming connections because of a specific actor attribute. By considering these effects, both network dependencies and actor attributes are examined rather than overestimating the role of either effect in the network (Lusher et al., 2013b, pp. 26–28).

Results

Descriptive Results

From Table 1, NHS-A has the largest network size with 247 actors, 389 ties, and a density of 0.006 (SD = 0.08). NHS-E has 279 ties, and a density of 0.008. From Table 2, NHS-E has an outdegree centralization statistics higher than NHS-A, indicating higher activity and information-seeking behaviors. However, the indegree centralization statistics for NHS-A is larger than for NHS-E. First, this result suggests higher levels of incoming ties in these networks than outgoing ties, which suggests that actors in this network tend to be receivers of information and advice. The non-clinical healthcare professionals have higher group outdegree and indegree centralization, suggesting higher popularity and information-seeking behaviors in this group. However, this finding is unsurprising as most TGT and KPO actors who have strategic roles regarding improvement work are included in the non-clinical healthcare professionals’ category. Additionally, there are more actors in the non-clinical healthcare professionals’ group than in the other categories due to the composition of those professional roles. Therefore, this finding may be aligned with the categories’ size and composition rather than other underlying relational dynamics.

Table 1 Descriptive statistics
Full size table
Table 2 Centralization measure
Full size table

Exponential Random Graph Models (EGRMs) Results

The ERGM results are presented in Table 3. The table contains the estimate of the parameters and the associated p-value. For each organizational network, there are strong negative arc parameters for the density effect (NHS-A [estimate =  −7.210, SE = 0.584, p ≤ 0.05] and NHS-E [estimate = −7.826, SE =  0.316, p ≤ 0.05]). This parameter measures the baseline propensity for a tie to be formed, and the significant negative parameters indicate that ties are rare and occur at random.

Table 3 ERGM results
Full size table

Closure Effects

The two-out-star (2-out-star) parameter effect is not observed in either NHS-A or NHS-E networks, indicating that actors in these networks do not commonly interact with only two actors. However, the three-out-star (3-out-star) parameter is small, negative, and significant for both networks. NHS-A (estimate = −0.316, SE = 0.043, p ≤ 0.05) and NHS-E (estimate = −0.089, SE = 0.02, p ≤ 0.05) are large but less than 1. This result indicates a tendency for actors to interact with three collaborators. Popularity Spread (AinS) estimate is significant and negative in NHS-A (estimate = −1.493, SE = 0.236, p ≤ 0.05). The results indicate a decentralized approach to improvement-related collaboration and that most actors have similar popularity levels. This parameter is not present in NHS-E, suggesting that unusual levels of in-degree centrality do not characterize this network. The activity spread (AoutS) estimate is positive and significant in both NHS-A (estimate = 2.831, SE = 0.398, p ≤ 0.05) and NHS-E (estimate = 1.614, SE = 0.267, p ≤ 0.05). The large, positive estimates suggest cohesion, where few actors are particularly active regarding outgoing contact and interaction with many collaborators. Practically, these actors either seek or provide improvement work information from many healthcare professionals. Compared to popularity spread, this result’s magnitude is much larger and more dominant within the networks. This result also indicates that activity is centralized around a few key actors in NHS-A and NHS-E. Therefore, we find support for hypothesis H1 in NHS-A and to some extent in NHS-E. In NHS-A, many actors seek advice regarding new practices and improvements from many of their colleagues and these relationships are not concentrated around a few key actors at the network level. NHS-A is a high cohesive network, and we expect the uptake of new ideas or practices more so than in NHS-E, which is relatively lower in cohesion.

Bridging Effects

Four parameters were selected to examine bridging effects within the network, as these configurations have two levels of connectivity and an intermediary or brokering actor. Simple Connectivity (path2) has one incoming and one outgoing tie from an actor and is present in the networks with varying results. For both networks, this parameter is negative and non-significant (NHS-E [estimate = −0.012, SE = 0.040], and NHS-A [estimate = −0.176, SE = 0.104]), indicating that there is little evidence that people who send more ties also receive them.

The One-In-Alternating Out Star (1inAout-star) measures the extent to which a connected actor sends ties to multiple other actors. Practically, this equates to an actor receiving information or advice about new practices from one actor disseminating information or advice across a broad range of contacts within the network. In NHS-A (estimate = 0.782, SE = 0.315, p ≤ 0.05), the effect is positive, suggesting that actors have a higher tendency to share and disseminate information about new practices among multiple connected actors.

The Alternating-in-One-Out Star (Ain1out-star) examines the extent to which a broker receives information from multiple actors and shares information with at least one other actor. In NHS-A (estimate = −0.152, SE = 0.279) and NHS-E (estimate = −0.014, SE = 0.158), but was not significant. Similarly, the Alternating-in-Alternating Out Star (AinAout-star) this effect is observed in both networks NHS-A (estimate = −0.930, SE = 0.759) and NHS-E (estimate = −0.214, SE = 0.352), but was not significant. Overall, we found no evidence of hierarchical effects such as bottom-up interactions where improvement work information is communicated from the lower levels of the organization to actors at higher levels of the organization.

Overall, three of the four bridging parameters are present in NHS-E; however, none of these estimates is large or significant. This result suggests that although varying forms of brokering activities are present in this network, they do not occur more than expected, and the magnitude of these effects is weak compared to other parameters within the model. All the bridging parameters tested are present in NHS-A’s network; however, three are negative and not significant, which indicates that they do not occur more than expected. Thus, we find some support for hypothesis H2 for NHS-A (1inAout-star) and not for NHS-E.

Actor-Relation Effects

In our study, actor-relation effects are used to provide insights into the presence and dominance of macro effects by dissecting relationships among actors based on their professional and partnership roles. In this case, homophily models the tendency for ties to be more or less likely between actors similar in the professional and partnership hierarchy, and this was detected in our networks. In NHS-A, the parameter was not significant among nurses (estimate = 0.726, SE = 0.138) and doctors and consultants (estimate = 0.814, SE = 0.144). However, it was positive and significant among the allied healthcare professionals (estimate = 0.916, SE = 0.155, p ≤ 0.05), non-clinical management professionals (estimate = 0.760, SE = 0.155, p ≤ 0.05), and actors in leadership roles (estimate = 0.284, SE = 0.114, p ≤ 0.05). This result indicates that actors in the same professional group or organizational status tend to communicate; however, it occurred more than expected among allied healthcare, non-clinical management professionals, and those in a leadership role.

In NHS-E, homophily is observed (positive and significant) in all the professional roles except nurses (estimate = 1.384, SE = 0.207). These effects are significant among doctors and consultants (estimate = 1.721, SE = 0.4282, p ≤ 0.05), the allied healthcare professionals (estimate = 1.335, SE = 0.180, p ≤ 0.05), non-clinical management professionals (estimate = 0.647, SE = 0.129, p ≤ 0.05), but not among those in leadership roles. The findings indicate that these professionals communicate more than expected within the network. The magnitude of the nurses, doctors, and allied healthcare professionals’ groups is large indicating that interaction regarding improvement work among these professional groups is more dominant than most of the structural effects observed. TGT, KPO, and RPIW actors’ tendency to communicate with each other is also detected, as these effects are all positive and significant in NHS-E. The TGT (estimate = 2.184, SE = 0.369, p ≤ 0.05) and KPO (estimate = 1.384, SE = 0.348, p ≤ 0.05) estimates are significantly larger than most of the structural effects and are larger than most of the professional role effects, indicating that the network is dominated by TGT and KPO interactions. A surprising finding is that there are no homophily effects among the L4L participants, and this group was the main category surveyed to conduct this research. This finding suggests that L4L actors in NHS-E are unlikely to interact with each other regarding improvement work and may have more diversity in their networks. Overall, we find support for hypothesis H3 in NHS-E, but somewhat less so in NHS-A.

Discussion

This research sought to employ an embeddedness lens to investigate workplace relations among the networks of professionals involved in improvement work. The analysis and results revealed synergies among the networks and broad themes that are important in understanding the uptake of new practices among professionals. Overall, our research confirms that embeddedness exemplified by two structural forms, closure and bridging, has varying consequences for the uptake of new practices among professional actors (Burt, 2001; Reagans & McEvily, 2008). We find that closure emerging from actors who are directly connected to members of a subgroup, and with closely knit ties are more likely to endorse behaviors or actions regarding new practices, if they share joint partners (Simmel, 1955). Closure among professional actors also promotes normative justification and influences the professionalization of practice, which further encourages practitioners and organizations to adopt and implement new practices (Smets et al., 2012). Accordingly, our findings confirm the link between closure and the uptake of new practices is that closure engenders strong social pressures that foster familiarity and shared values that determine the way actors create, use and share new practices (Brown & Duguid, 2001; Tortoriello et al., 2012).

In terms of bridging, the NHS-A network is exemplary. Multiple forms of bridging are present within the network, where professionals broker information and advice to and from several sources and collaborators. Our findings suggest that multifaceted bridging relationships are a distinctive feature of this network, where professionals are engaged in various interactions to receive and share information and advice with multiple collaborators. As previously mentioned, bridging is associated with network spread, the diffusion of innovations (Ingold et al., 2021) and having multiple forms of bridging activities that are beneficial when attempting to disseminate information about new practices in the wider network. Bridging is thus facilitated as actors broker information or advice between disconnected actors and groups that emerge from closure among actors (Mehra et al., 2001). Our findings show positive homophily effects are found in the networks, suggesting actors from similar professional groups tend to form relationships. In NHS-A, relationships among allied healthcare professionals and non-clinical and management professionals have a more pronounced effect on the network. In contrast, relationships among doctors, allied healthcare professionals and non-clinical and management professionals have a more pronounced effect in NHS-E. This finding is important as it signals that NHS-E is more susceptible to macro-level (institutional) influences for the uptake of new practices. This confirms the findings by West et al. (1999) that clinicians in these settings have more influence on changing or resisting the adoption of new practices.

Our findings make further contributions to the literature. First, we observed a form of structural embeddedness combining both closure and bridging activities simultaneously, we call here collective embeddedness, in NHS-A and NHS-E. The name is derived from the high degree of structural embeddedness among the strategic change agents, which fosters collective action that supports the uptake of new practices. This collective embeddedness, combining closure and bridging activities, provides a foundation to support the adoption and spread of new practices since actors’ relationships support interactions. Closure promotes normative justification and influences the professionalization of practice and encourages practitioners to adopt and implement new practices. Similarly, a high degree of bridging facilitates sharing information and advice about new practices within and across professional groups. Our observation shows that collective embeddedness, as a high degree of structural embeddedness, is a key mechanism underpinning the uptake of new practices. In order to suggest what organizations should do, we describe two exemplars of this form of embeddedness.

The first exemplar is broker-driven; as its name suggests, brokers have more prominent relationships in the network. This was observed in NHS-A, where most brokering roles and cohesive relationships are held by clinical professionals who are actively engaging with new practices in their everyday work. In contrast, this behavior, however, was not observed in NHS-E. The practical implication is that clinicians as brokers are vital for promoting the uptake of new practices. The second exemplar of collective embeddedness is conjointly-driven, where no change-oriented group supersedes another. This behavior is again observed in NHS-A, where there is a high degree of interaction among both brokers and other change agents. In this case, homophily effects were observed non-clinical professions. The practical implication is that conjointly-driven activities among non-clinicians is vital for promoting the uptake of new practices. However, when homophily effects are high for clinical professionals, this hinders conjointly-driven activities (as seen in NHS-E). Finally, our study confirms that whether networks are characterized by broker-driven or conjointly led, the critical point is that structural embeddedness highlights the extent of workplace relations among actors, which establishes the context of and likelihood of peer learning among and socialization among professionals. Ideally, organizations and practitioners should cultivate collective embeddedness among all groups, however, their focus should be to first introduce and encourage peer learning and socialization among brokers, then within departments and environments when seeking to enhance the uptake of new practices, as a high degree of cohesive and brokering relationships must be present to increase the likelihood of practice change.

Conclusion

Overall, our study shows that the combination of macro and micro processes results in a shift, where change amplifies from the micro to macro levels through the process of enacting, sharing, supporting, and participating in the practice itself (Smets et al., 2012). When actors encounter new practices, they tend to provide information and seek advice from colleagues to understand the purpose, relevance, and suitability of new practices within their immediate work environment. This research confirms that social interactions influence the uptake of new practices through professional norms and relationships, thereby providing an alternative explanation of organizational change processes based on social phenomena and collective actors.