Structural Uniqueness is a newly developed metric capable of computationally assessing nodes within a social network. The task of selecting a subgroup which provides broad network coverage in one step is relevant within epidemiology, marketing, and in organizational networks where training opportunities may be limited. In these situations diffusion through networks provides a potential multiplier effect.
When selecting a subgroup, members with high in-degree alone may not provide broad coverage as these nodes may redundantly connect to the same network members. Dyadic comparison is possible with traditional Social Network Analysis methods such as structural equivalence, but may become difficult to employ when the subgroup size exceeds two members. Redundancy metrics developed by Ronald S. Burt allows quantitative exploration of overlapping ego networks, but does not consider directionality and fails to present how one node shares its audiences outside its ego network.
In contrast to these methods, Structural Uniqueness provides the advantage of being a node metric rather than a dyadic comparison, and is suitable for directed network data. The computationally difficult task of Dominating Set identification may be approximated or matched in quality by selecting nodes with sufficient in-degree and high Structural Uniqueness.
Research Setting: This metric was developed and employed as part of the Social Epidemic of Safety research project at a large DC area hospital system. This project uses a Social Network Analysis approach to identify collaborative patient care networks in hospital intensive care units.
Data from the SNA survey is aggregated with node attributes to identify 15 potential focus group members. This set of 15 is further refined into a set of 7. Focus group members are given the opportunity to voice concerns regarding patient safety and engage in the participatory development of a small-scale intervention to improve the built environment of the ICU, to refine the collaborative process among health care team members, or identify practices that may reduce avoidable errors.
Research Background: Iatrogenic events have been estimated to be the 3rd leading cause of death in America, a reality that far exceeds the estimate of 100,000 deaths per year from a decade ago. An inestimable cost to the family members of those effected, iatrogenic events and hospital acquired infections also have heavy financial implications for hospitals and damage the morale of team members in the hospital units where they occur.
In this context, the Social Epidemic of Safety model aims to employ a method which contrasts top down directives for patient safety and has the potential to leverage quick implementation via network diffusion from a well-positioned subgroup.
Identification of a focus group with broad network coverage is a key task within this research with the goal of increasing diffusion potential. Structural Uniqueness was developed specifically for this task. The computation for this method is presented as an algorithm and as a matrix multiplication process.
Computational Overview: Structural Uniqueness operates under the assumption that survey respondents distribute attention equally across out-degree ties. This assumption is similar to the premise of initial PageRank calculations. Attention inflow calculation follows and is averaged for each node with non-zero in-degree. The metric which follows is interpreted as the average attention a potential hub receives from its in-degree network.
Nodes with higher Structural Uniqueness may reach network members with limited ego networks or lower in-degree, supporting diffusion potential to diverse members of the ICU. Network coverage performance is illustrated with a set ICU data and is compared to the coverage performance of Burt’s redundancy metrics and a popularity based selection method which focuses on in-degree alone across five social network survey questions.
Practical considerations, such as computations when using multiple social network questions, are also discussed.
Initial Focus Group Output: Given a budget and an opportunity to refine existing processes the first focus group chose to develop a set of communication tools to strengthen collaborative patient care in the ICU by better signaling what types of support could be readily deployed. This aims to remove ambiguity and reduce cognitive burden for nurses who are potentially experiencing patient acuity changes.
The third phase of research will explore the use of focus group developed tools within the test ICU. The focus group of the test ICU is directly tasked with assisting with diffusion as compared to a control group which will use traditional top-down introduction of the communication tools.
Additional Focus Group Data: The Social Epidemic of Safety model aims to facilitate other aspects of network science, such as the capacity of weak ties to provide diverse information and potentially mitigate information cocoons. Limited staffing, high patient to nurse ratios and the prevalence of nonveteran nurses in ICU staffing were also voiced as strong concerns by focus group members. These concerns illustrate the diverse information front line health care practitioners can provide those in hospital budgeting and management roles.
The pervasive national shortage of nurses may support a view among administrators that high patient to nurse ratios are unavoidable. The potential presence of groupthink among hospital administrators, partially characterized by an unwillingness to voice concerns, has the potential to be mitigated in confronting the concerns of focus group members. The diverse perspective of front line care providers may help administrators better contextualize the toll on hospital staff and the potential contribution of low staffing to the iatrogenic events which occur across the American health care system.
The data may also support and expansion of budgeting to such focus groups. Expanded budgets may allow network building as a focus group outcome, such as engaging Resource Nurses to expand patient care knowledge within the health care team.
The Social Epidemic of Safety forms a first step in developing a practitioner adaptive safety model which engages health care providers, efficiently use social network data to support diffusion, and may enable better network responsiveness. This type of participatory allocation system may also lead to double loop learning in hospitals, allowing local learning to diffuse more broadly to other organizational members with the goal of improving patient outcomes.
