Strategic Choice

The concept of relationships built on strategic choice suggests that actors build relationships because they benefit directly by the tie. Theories of self-interest fall in this category and frame network behavior as organizations (actors) seeking interorganizational relationships when such network links offer a direct benefit to the organization. An economist might specify this dynamic in terms of transaction cost economics, which considers or calculates the costs associated with the creation and maintenance of any specific relationship as weighed against the benefits derived from the association. Thus, organizations seek to build relationships with those that allow them to operate more efficiently, extend their scope, increase their resources, or facilitate the long-term accomplishment of their goals.

Resource dependence theory has a significant following in the health care organization literature (Pfeffer and Salancik, 2003). Within this theory, organizations are seen as requiring resources from other organizations in the external environment, and this dynamic creates power asymmetries and interdependencies (Cook, 1977; Provan, Beyer, and Kruytbosch, 1980). This perspective places interorganizational relationships in the context of how the need for and availability of external resources influence organizational behavior. Thus, relationships are framed around gaining strategic advantage by securing access to limited resources.

In both of these cases, studies using a strategic choice perspective focus on the exchange of resources as the basis for relationships. Discussions of competitive advantage, as presented in the strategy literature, are tested against the structure of relationships. A hypothesis then may examine whether organizations that import resources (in-degree) are more successful than those that export those resources (out-degree).

Natural Selection

The term natural selection suggests the work of Darwin, which is often summarized through the phrase “survival of the fittest.” However, the theory might be better expressed as “the fittest survive.” This latter conceptualization highlights that choices influence environmental fit, and relationships may provide an adaptation advantage in that context. Organizational ecology (or population ecology) theory depicts relationships in terms of those that provide the organization the greatest chance of surviving (Baum, 1999). In these cases, organizations build relationships to improve their ability to survive environmental dynamics. Such approaches highlight variation, selection, and retention in the context of competition for scarce resources. A central assumption here is that organizations create network links to increase either their individual or collective fitness in an effort to increase the likelihood of “selection” in the Darwinian sense. An extension of this framework has been to view networks within a developmental framework. This latter work has focused on the life cycle stages of networks and the importance of building legitimacy for long-term survival (Human and Provan, 2000).

Collective Action

Theories of collective action focus on situations in which organizations create network links with each other to maximize their conjoint ability to seek or provide resources from third parties. For instance, organizations might share information to facilitate the mobilization of a community to deal with natural disasters like Hurricane Katrina. Collective action builds on the axiom that networks create a whole that is greater than the sum of its parts. For instance, in addressing complex health and human service problems, the wide sharing of knowledge and information is critical. The development of a knowledge network involves a broad set of actors functioning through a multiparty mechanism to develop a shared knowledge base that provides the ground for collaborative action (Clarke and Roome, 1999; Cummings, 2004; Tung, 2002). These interorganizational “knowledge networks” enable organizations to achieve strategic objectives by leveraging the information available from partners to a specific domain (Applegate, McFarlan, and McKenney, 1999; Premkumar and Ramamurthy, 1995).

Collective action builds on public goods theory (Olson, 1965) with the idea that individuals and organizations are motivated to join and work in networks to reap the benefits of working together. These benefits are presumably ones that could not be obtained when acting through motives of self-interest or social exchange even within a network context. Theories of collective action are useful in a broad way for explaining why organizations might form and sustain a network, and research has explored why particular network structures are formed and which structures might be most effective under particular conditions (Marwell and Oliver, 1993).

System-Structural

The fourth approach takes the concept of organization-environment and turns it on its head. Where strategic choice, collective action, and, to a lesser extent, natural selection perspectives are based on how an agent might react to an external stimulus, system-structural approaches consider relationship building as an environmental or external dynamic. For instance, in traditional schools, classrooms provide a macrolevel dynamic, which increases the likelihood that an individual might build a relationship. Individuals who share no classes in common are less likely to form relationships than those who do. Basic to understanding this perspective are four of its more common theoretical constructs: proximity, homophily, social contagion, and social persuasion.

Heterophily

In one study, heterophily—the tendency to interact with people having different backgrounds—was found to be prevalent within the broadly defined tobacco-harm-reduction network (Provan, Clark, and Huerta, 2008). Such findings suggest that researchers are generally more likely to interact with those in different disciplines than to collaborate with others within their discipline. Figure 10.1 illustrates the transdisciplinary network and shows the nature of the relationships. Most notably at the macrolevel, one can see the relative grouping of psychologists (squares) and chemists (triangles pointing up) among each other rather than with dissimilar others. In contrast, those in medicine are similarly connected, while at the same time, those connections are also heterophilous (Provan et al., 2008).

Specialization in Knowledge

In principle, high specialization (differentiation in the language of organizational analysts) promotes expertise and therefore quality of care, but in reality it hinders integration and continuity of care (Nicolini et al., 2007). A half-century ago, only eighteen medical specialty boards and a handful of subspecialties existed in the United States. By 2011, there were 158 specialties and subspecialties (Detsky, Gauthier, and Fuchs, 2012). Similarly, in the United Kingdom, there were nearly 100 approved medical specialties and subspecialties in 2011 (General Medical Council, 2011). In Italy, there were “only” 55 types of medical specialty schools as of 2012 according to the Ministry of University website (http://scuole-specializzazione.miur.it/) and countless subspecialties. Medical specialties can be classified as surgical versus internal medicine; age range of patients (children or elderly); organ based or technique based; and diagnostic or therapeutic. Some specialties may be multidisciplinary, such as occupational medicine, preventive medicine, disaster medicine, and gender medicine.

In other words, physicians know more and more about less and less. More important, as classic organization theory argues (Lawrence and Lorsch, 1967), as we move to differentiation through the development of specialties and subspecialties, we can expect a concomitant need for integration. Understanding the social networks that individuals construct in health care will allow interested champions to identify and analyze who knows whom or who knows who knows what, which makes the theoretical and methodological knowledge of social networks particularly useful for knowledge management (Cross et al., 2001; Dandi and Sammarra, 2009).

Complex Demand

With the aging of populations and widespread chronic diseases and comorbidities, there is today a greater need than ever before for integration of care professionals and nonprofessionals. The increasing demand for long-term care services by the elderly adds urgency to the need for better coordination between health and social care. Constraints on government and public health spending promote the creation of hub-and-spoke networks of care where economies of scale and territorial coverage are balanced. The increasing demand for quality care requires better integration between primary and acute care providers, as well as the development of information systems for the integration of patient data produced across care levels and specialties.

Shrinking Professional Autonomy

Health care organizations are made of professionals who standardize their behavior and knowledge according to their professional rules of conduct and training. As professionals, they are autonomous in clinical decisions related to specific cases. However, the diffusion of litigation, insurance costs, defensive medicine practices, the consumer movement, the increased importance of midlevel professionals, and the spread of EBM today challenge this autonomy. More and more, professionals must rely on standard procedures and interprofessional collaboration rather than their own skills.

Nature of Medical Knowledge and Difficult Knowledge Update

Heathfield and Louw (1999) have estimated that medical knowledge increases fourfold during the professional lifetime of a doctor. Doctors and health care organizations should therefore update their competencies and ensure that innovations such as new drugs, technologies, and clinical guidelines are adopted, but the superabundance of medical knowledge makes it impossible for any doctor to be thoroughly aware of all that is known about medical progress. Furthermore, even with a clear understanding of the evidence, only 10 to 20 percent of the cases that a typical doctor encounters can be handled using guidelines alone (Heathfield and Louw, 1999). For all the other cases, doctors must rely not on scientific knowledge but on their practical experience, peer collaboration, informal communication, teamwork, and the integration of different capabilities. The influence of network structure on health care professionals' clinical guideline is one of many examples in the literature that have relied on the theory and methodological approach in SNA that we explore later in the chapter.

The increasing awareness of the need to integrate people, competencies, and tasks in health care service delivery favors a network-based approach to organizational analysis and design. Organizational behavior and performance, in terms of efficiency and quality of care, are explained not only through the analysis of the characteristics of the organizations but also through the analysis of the role of their relations and coordination. Thus, SNA is an excellent approach for providing the tools and the theoretical framework that assist health care managers and policymakers in analyzing, measuring, and designing health care networks of organizations and individuals.

Evidence-Based Medicine Implementation

Quality of care is increasingly associated with the implementation of EBM and clinical guidelines in clinical practices. But a new effort must be put in place in order to link research to practice and to standardize clinical pathways across organizational and disciplinary boundaries. This is because of the so-called know-do gap, the inability to apply research results to specific patients. In other words, the incapacity of bringing the latest research results to the patient bed makes ineffective much of the effort of producing new knowledge. It has been estimated that the time for research to make its way into practice is on the order of fifteen to seventeen years (Balas and Boren, 2000). This gap may account in part for why, according to the World Health Organization, 7 million children die every year from preventable causes (Bryce et al., 2005).

Several SNA-based studies have addressed the problem of the adoption of EBM by clinical practices. An effect of the high specialization of health professionals, the overwhelming volume of new clinical research, and the tacit nature of much medical knowledge is the know-do gap. Instead of accessing explicit formal sources such as guidelines, clinicians rely on their expertise, also called mindlines, which are defined as “internalised, collectively reinforced, tacit guidelines in the head” (Gabbay and Le May, 2004, p. 1014). In this context, the term collectively reinforced refers to the manner in which clinicians develop these mindlines through “their interactions with each other and with opinion leaders, patients, and pharmaceutical representatives and by other sources of largely tacit knowledge that built on their early training and their own and their colleagues' experience” (Gabbay and Le May, 2004, p. 1014). In fact, through a process of enrichment, control, and refinement resulting from peer interaction, the individual mindline becomes a collective mindline, which is more easily implemented than guidelines.

SNA-based studies have confirmed the importance of social relationships with colleagues in the adoption of clinical guidelines (Keating et al., 2007; Salvatore, 2006). Colleagues help make sense of scientific findings that are ambiguous or not immediately applicable to the case under scrutiny (Dopson et al., 2002). A recent empirical project examining the impact of professional networks on the use of EBM guidelines among pediatricians confirmed this hypothesis (Dandi et al., 2011; Mascia, Dandi, and Di Vincenzo, 2012). By analyzing advice relationships among 120 pediatricians in seven local health authorities in Italy, these authors found that the network has a pivotal role in EBM implementation. In these cases, the authors found that networks of family pediatricians characterized by a high proportion of highly connected players is associated with a higher self-reported frequency of use of clinical guidelines. This means that the most respected pediatricians (those who are more frequently cited by the colleagues as sources of advice) are those using guidelines more often. Also, a high centrality in the network of advice relations between family pediatricians and hospital pediatricians is also associated with a higher use of EBM. In other words, pediatricians who ask more hospital colleagues for advice are more exposed to hospital EBM pathways and therefore are the most keen to use clinical guidelines.

SNA: Primarily Domain or Issue Focused

First, network research requires the explicit definition of a domain of interest. Underlying all network analysis is a theoretical framework that is used to suggest how and why actors engage with each other. That framework consists of questions about collaboration, collective action, cooperation, interaction, relationships, and trust. Also, organizations build relationships to match the dynamics of the environment in which they are embedded. Brandenburger and Nalebuff (1996) note that organizations build competitive and cooperative relationships in the same problem domain. Therefore, it is important to clearly define the scope of the issue that is to be studied. We provide two examples. The first, the Italian National Transplant Network, is an interorganizational multilevel formal network. The second explores the informal network engagement created by efforts to deal with the obesity epidemic in the United States.

The National Transplant Network is responsible for coordinating organ procurement in Italy. The network is organized in four levels. First, the National Transplant Center governs the network at a national level by fixing criteria and procedures for organ sharing, training professionals, laying down national guidelines, and controlling waiting lists through the Transplant Information System. Second, three interregional centers (NITp, AIRT, and OCST) coordinate interregional organ exchanges. Third, twenty-one regional centers govern procurement and transplants at the fourth and local level, where coordinators organize the procurement and transplant processes at hospitals. The governance of the network is therefore a complex task because there is a trade-off between the need to standardize procedures at the national level and the need to protect autonomy at a local level (Dandi et al., 2013). Also, the two main processes—procurement and transplant—involve different professionals with different competencies, aims, time spans, and rewards. Given these aims, the boundary of the network includes all local hospital coordination centers (about 280).

Weight management programs are also multileveled interventions. While many programs are aimed at the individual (such as offering nutrition and lifestyle advice), success at the population level is more likely if the individual efforts coincide with interventions at the organizational level (such as healthy food options in the workplace and at home) and at the environmental level (such as community approaches to food). A comprehensive approach to obesity then requires collaboration among a mix of individuals and organizations with varied interests, talents and skills, knowledge, and resources. Furthermore, because of the complex and multidimensional factors that influence healthy living, no one approach is likely to solve the problem either. Rather, among researchers, there is a concomitant need for a high degree of collaboration that explores the complexity of the genetic, psychosocial, and environmental aspects of the problem. In this case, the issue of obesity frames the inquiry. This is important because some organizations within the network build relationships for reasons other than weight management. The failure to do so explicitly would result in a study without boundaries and, therefore, without testable hypotheses.

Relationships as the Unit of Analysis

A second reason that SNA is more than a collection of quantitative techniques resides in a key difference between SNA and most traditional methodological approaches in social science. In most studies, the individual or organization is compared or contrasted against a panel of other individuals or organizations. Consider the case where organizations employ different strategic approaches to electronic medical records (EMR) systems. Much of the research then looks at some outcome, such as productivity, and explores systematic differences that may occur as a result of these alternative approaches. In contrast, SNA looks at the relationships that exist between organizations as the unit of analysis. One can see the difference visually.

Figure 10.2 presents an idealized case of six health care facilities and the direction of patient referrals among them. At first glance, there seems to be certain basic patterns. Hospital C, the general hospital, receives referrals from all facilities but does not refer out to all those facilities. One might record the data about the Referral status (R) between facility A and B as R_AB = 1. However, because there is no reciprocal referral relationship, the status of that relationship might be recorded as R_BA = 0. The relationship between any two actors in a network is often called a dyad. SNA then is a method for exploring the structural relationship among actors. The data of these relationships are generally recorded in matrix form, as illustrated in table 10.1.

In the Italian Transplant Network case, the nodes are procurement and transplant hospitals, and relationships are organ exchanges. Organs are procured by hospitals' local procurement coordinators through a continuous monitoring of potential donors within intensive care units. The procured organ is then assigned to a patient on a regional waiting list. If no patient can be matched in a region, the interregional information system assigns the organ to another region. Transplant hospitals in the region of the receiving patient then either receive organs through an emergency shipping system or send its surgeons to the procurement hospital in order to remove it. This procedure is highly inefficient, but it is required for some types of organs because of the different surgical capacities across hospitals. Organs may be wasted if they are removed using a different surgical approach from that of the transplant surgeon who will receive the organ. Therefore, it is key for the network to minimize the distances between procurement and transplant hospitals. In this case, SNA is being used to develop an indicator for efficient organ exchanges. The average number of kilometers per organ per transplant hospital is calculated in order to measure the efficiency of the different levels of the network: transplant hospitals, regions, and interregional centers.

SNA Requires More Than Two Actors

Third, SNA requires that the domain studied include no fewer than three individuals, groups, or organizations. In most applications of SNA where primary data collection is involved, we tend to see fewer than one hundred actors, in part because of the complexity associated with achieving an 80 percent response rate when one increases the number of actors. The complexity of responding to SNA surveys is a clear stumbling block, and there are two general approaches to the establishment of the domain for a study. The first is an a priori definition based on clear external criteria. However, analysis of secondary data, especially in cases of online social dynamics, can span thousands or hundreds of thousands of individuals.

In cases of primary data collection, where actors are well known and there is a commitment to exploring some structural dynamic, SNA can be immensely powerful as a means to illustrate social dynamics. However, where the domain is not well understood, the frame can be established using the reputational or snowball sampling approach (Knoke and Kuklinski, 1982; Scott, 2000). This approach depends on researchers identifying a core sampling pool for inclusion and then engaging in an iterative process that relies on the cumulative knowledge of network members or participants about who is involved in the network. This then determines which organizations will be asked to participate and identifies the organizations that meet specific inclusion criteria in the context of each case. Reputational sampling also identifies key individuals within each organization that are most likely to be knowledgeable about the network activities and involvement of their organization. In large and diverse service organizations, it may be necessary to survey many individuals from the same organization, and in these cases, responses can be aggregated to form a single organizational response.

Actor-Focused Metrics and Research

In organizational research, the target of actor-focused research has generally been on dyadic relations (see Gulati and Gargiulo, 1999). The dyadic approach to research tends to view the network as a collection of two-party relationships rather than as a comprehensive, multiorganizational social structure in its own right. Although a number of metrics are associated with actor-focused network analysis (Brass, 1995), five major social network measures are assigned to actors. Three of these are tied to the concept of centrality, which measures the extent to which a specific actor is positioned at the center of relationships within the network.

Mesolevel Metrics and Research

A mesolevel perspective on social networks involves a focus on groupings of dyads, for instance, triads, and their larger structures—cliques and clans. This level of analysis focuses on the existence of identifiable structures within the network. An interesting study by Ennett and Bauman (1993) explored the formation of cliques within a high school. In the study, the authors draw on B. Brown (1989) to define cliques as “interaction-based clusters of adolescents who spend more time with each other than with other adolescents and who tend to share similar attitudes and behavior” (Brown, 1989, p. 189). Thus, looking at relationships in schools, one can find structural artifacts that exist above the dyad level and would be lost by looking solely at the whole network level. Mesolevel dynamics appear when individuals create cohesive groups of relationships. Mesolevel SNA focuses on whether the presence or absence of substructure is associated with the presence of social dynamics. A study by West and colleagues (1999) found that cliques play an important role in helping to define social identity and establish a sense of belonging. In cases of multidisciplinary environments, the siloing phenomenon created by different communities of practice in health care (Glouberman and Mintzberg, 2001a 2001b) creates a significant barrier to patient care and serves as one basis for the development of barriers to relationships that manifest themselves at the mesolevel.

An example of a mesolevel substructure is a triad with a focus on relationships among groups of exactly three actors. Exploring the dynamics of interactional influence within such relationships illustrates the concept of transitivity. Transitivity pertains to relationships among three actors, which we will label A, B, and C, such that A and C are connected through B. This phenomenon, called transitive triples, is summed over all potential transitive triples and provides a mesolevel perspective on betweenness centrality.

Georg Simmel (1964) argued that dynamics within triads are not reducible to their dyadic components, and he advanced a theory based on the propensity of specific triad types to form. The sixteen forms of triads derived from his work are shown in table 10.2. SNA researchers use a common typology for the identification of specific triad connections called the M-A-N system. Let us imagine a friendship between individuals A and B and a friendship between individuals A and C. Now consider the case where A and C are in fact enemies. This dynamic is called a forbidden triad because theory suggests that the tension created by A and C in their relationship with B should result in the dissolution of either the AB or BC relationship or the reconciling of AC. According to the classic M-A-N labeling scheme, each triad is classified using a three- to four-letter code. The first three characters are numbers corresponding to the number of mutual, asymmetric, and null dyads, followed by a fourth character that further distinguishes types as either D(own), (U)p, (T)ransitive, or (C)yclic. Simmel (1964) recognized that for any network with no underlying social dynamic that influenced link creation, the distribution of each type of triad is prescribed. He argued that social dynamics pushed people to other, nonrandom arrangements. He also argued that certain forms were unstable (transient) and others stable. One can think of this in terms of the old axiom, “Any friend of yours is a friend of mine.” This work was developed into a set of theories of balance to explain these relationships. In their study of five National Tobacco Control Networks, Luke and colleagues (2010) used triad analysis to test model robustness and fit, meaning that the results were robust on varying assumption based on the empirical data gathered.

		Specialized Link Topography
Number of Ties in Triad		D(own)	U(p)	C(yclic)	T(ransitive)
0	003
1	012
2	102	021D	021U	021C
3		111D	111U	030T	030C
4	201	120D	120U	120C
5	210
6	300

At an even broader level, sociologists identified the development of structural forms that they called cliques: a set of three or more actors who share links with every other actor in the subgroup. The definition of clique was loosened to explore less tightly formed subgroups and include k-cliques (Luce, 1950), k-clubs (Alba, 1973; Mokken, 1979), and k-plexes (Seidman and Foster, 1978). K-cliques are subgroups of nodes whose shortest paths among them are all below or equal to a value “k.” K-clubs are subgroups where all the nodes are reachable within a certain diameter “k,” and K-plexes are subgroups where each node is connected to at least n-k other nodes. With the structural changes to care associated with efforts to transform care (such as patient-centered medical homes), we can expect to see the cliques that exist because professional silos alter with service delivery practices. As better data become more readily available on relationships in health care, the exploration of relationships at the mesolevel will take an increasingly prominent role.

Network-Level Metrics and Research

Network-focused metrics draw on and use many of the same ideas and measures developed by actor-centric researchers to conceptualize the totality of the network. Rather than focusing on the role or contribution of any one agent, network measures are analogous to the average in statistics—a representation of the whole through central tendency. As a result of this macrofocus, network-level metrics are useful when making comparisons among two or more networks. They are also often reported in case studies or conceptual presentations. Although some measures are averages, others are unique conceptualizations based on properties taken from graph theory, a branch of mathematics that SNA draws heavily on. Some key network-level indicators are network centrality measures, cohesiveness, reachability, symmetry, triplet census, and transitivity.