Date of Award
2024
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Computer Science
First Advisor
Laurent Hébert-Dufresne
Abstract
Within the study of spreading phenomena, one goal is to identify methods of preventing the spread of epidemics. Networks are particularly interesting under consideration of this goal. There are common aspects of the heterogeneity of contacts which allow information about the contact structure to be leveraged for powerful interventions with minimal system changes. However, identifying the best interventions is a challenging task. The range of options is often constrained by external factors which make network information difficult to obtain, or which restrict or increase external costs of potential system changes from logistical constraints, making the relative utility unclear.Because of the complex nature of networked systems, small differences between the input expectations of an intervention heuristic and the constraints of an actual system can lead to large differences in the utilities of these heuristics. The present work offers strategies for identifying interventions against epidemics on networks under a variety of logistical constraints. These strategies address the above issues, either by identify- ing those strategies which are robust to differences in system constraints, or proposing new strategies which are viable in highly constrained systems. We focus first on exploring how robust commonly proposed network immunization strategies are under conditions of nodes missing at random — identifying the impact of missing data on these strategies’ efficacy and identifying the scenarios under which more robust strategies surpass more popular ones in efficacy. Next, we explore this concept further by developing a new strategy for the extreme case of beginning an network intervention with no node data by adapting one of the more robust strategies from chapter two. We explore some interesting results from this strategy using master equation and simulation methods, identifying the conditions where it is efficacious. Finally, we zoom out from the specific application of immunization strategies on abstract networks, to the more general question of efficacy across a range of possible interventions applied to the specific case of the spread of the Omicron variant of COVID-19 in the United States. We demonstrate that network approaches can be used as part of a larger, more specified agent-based modeling approach, allowing network-leveraging interventions to be simulated, compared to, and combined with more common types of intervention. We demonstrate that the combination of multiple weaker interventions rivals more intense but costly singular strategies.These studies collectively demonstrate the power and versatility of network-based approaches in developing pragmatic interventions against epidemics, even under significant constraints. By bridging theoretical network science with practical challenges, this work provides valuable insights for both researchers and policymakers, offering a framework for developing robust, adaptable strategies to combat the spread of epidemics.
Language
en
Number of Pages
189 p.
Recommended Citation
Rosenblatt, Samuel Frederick, "Pragmatic interventions against epidemics on networks" (2024). Graduate College Dissertations and Theses. 1956.
https://scholarworks.uvm.edu/graddis/1956