Understanding Network Failure and Success in Veteran Care Networks

Gibson, Zachary Michael

doi:https://doi.org/10.21985/n2-xycj-ra77

Work

Understanding Network Failure and Success in Veteran Care Networks

Public

Download PDF

Download All Files (.zip)

This dissertation describes three studies which develop a deeper understanding of both failure and performance in human service referral networks focusing on veterans. It draws on theories across network science, public administration, organizational communication, and technology design to explore three major questions: 1) how stakeholders in veteran care networks define failure; 2) what the factors are that facilitate and hinder performance in veteran care networks; and 3) what the role of technology in veteran care networks is. This dissertation addresses these questions through a mixed-method approach. Study 1 examines the robustness of a novel mixture model technique for comparing relational structures in networks. Using leave-one-out and leave-n-out approaches, it simulates over 700 million networks to test the technique’s sensitivity to subsample size, subsample composition, and the number of simulations per subsampled network. Analysis of the 65 leave-one-out conditions finds that the technique is desirably robust against subsample composition, suggesting that which networks the subsample includes does not affect the stability of the mixture model. Analysis of the 390 leave-n-out conditions finds that the technique produces the most reliable results when simulating from all of the observed networks rather than a subset, and that reliability increases as the number of simulations increases. This study therefore identifies two best practices for the mixture model technique—1) simulate from the full sample 2) until the mixture model is reliable—and offers the Kolmogorov-Smirnov statistic as a measure of reliability for the technique. Study 2 defines and develops a theory of failure using a grounded theory approach. Critical incident interviews with 30 providers of one AmericaServes network raised four elements of the referral that providers use to evaluate its failure. Providers considered referrals as failing when they both could not provide services directly and could not connect a client to services, or when a referral resulted in a negative outcome for either the client or the provider. Interviews also highlighted six common challenges that promoted referral failure and thereby network failure. These are: 1) mismatch between supply and demand, 2) misperception of services, 3) existing workflows and institutions, 4) reinforcement of network roles, 5) transaction versus communication and coordination, and 6) lack of client cooperation. These six issues engage in a complex interplay that makes both their individual and collective resolution difficult. As well, responses to the interviews also described how the providers’ four varied uses of the network’s community referral technology contributed to these challenges. Study 3 then combines extant theories on network structure, both relational and whole-network, and network learning to explore what affects network success. By using data from AmericaServes’ referral technology, this study tests a conceptual model of network success using 8,954 service episodes (i.e., referrals) across 11 networks over the year 2019. A cross-classed hierarchical logistic regression of service provision revealed that the relational structure of the inter-service network could both promote and hinder success while experience provided the strongest positive effect. Features of the whole-network had no significant effect on networks’ performance. Overall, this dissertation provides five major contributions to research on human service referral networks, and care and service networks more broadly. First, it identifies best practices for a novel network comparison approach that can support further explorations into the linear effects of relational structure on network performance. Second, this dissertation examines network performance holistically, studying failure from a bottom-up perspective and success from a top-down perspective. Third, it gives voice to client disenfranchisement and how the problems which providers face in human service networks reinforce disenfranchisement. Fourth, it is among the first to describe how providers actually use community referral technologies in the field rather than just what features users like or dislike. Finally, this dissertation recommends future directions and best practices for scholars, designers, and practitioners.

Creator