The Systematic Bias of Random Graphs in Modeling Disease Spread Dynamics
Donald Curtis, Philip Polgreen, Sriram Pemmaraju and Alberto Segre
Introduction
The success of contact network epidemiology for studying disease spread in a population depends on models that can produce typical instances real-life contact networks. The Erdos-Renyi (ER) graph is a mathematically interesting model that generates a random graph for a given number of nodes and edges but it has been shown to be a poor model for many real-world graphs. We consider two enhancements to ER graphs, the Configuration (CONFIG) graph and its extension that takes degree assortativity into account (CON-ASS) and analyze their ability to mimic disease-spread on a contact network of healthcare workers (HCWs) at the University of Iowa Hospitals and Clinics (UIHC).