Modeling the Distribution and Impact of Immunity to COVID-19 in Populations

Abstract

The SARS-CoV-2 pandemic has highlighted the need for scientifically-informed public health policies. Mathematical modeling is an important tool to help evaluate possible interventions and better understand disease dynamics. In this dissertation, I apply modeling approaches to evaluate vaccine allocation policies, and extend key takeaways from this analysis to suggest how future vaccine allocation modeling should take equity into account. I also use modeling to estimate the level of protective immunity conveyed by prior SARS-CoV-2 infection. In Chapter 1, I first model efficiency and fairness in a vaccine prioritization strategy. At the time of initial COVID-19 vaccination, when the number of vaccine doses available was limited, decisions had to be made on the order in which different groups of people would be eligible for vaccination. After health-care workers and people older than 75 years were made eligible, this case study compares next vaccinating either people 65-74 years old, or other front-line workers. As the burden of COVID-19 infections and death is characterized by stark disparities in age and race/ethnicity, I estimate both the number of lives saved and the number of years of life saved under each of the policies, overall and in every race/ethnicity group, in the United States and every state. Next, in Chapter 2, I expand on this work by conducting a methodological study of COVID-19 vaccine allocation modeling papers, specifically looking for publications that considered equity. I provide examples of how modeling can be useful to answer equity questions, and highlight some of the findings from the publications that did so. I also highlight eight key considerations that are important to take into account when including equity in future vaccine allocation models. Finally, in Chapter 3, I estimate how infection-derived immunity protects against the Omicron variant among unvaccinated individuals using a unique dataset from the Hutterite cohort study. I use a chain-binomial model combining serological and virological surveillance data to model how prior infection affects susceptibility to infection and infectiousness when infected. I also investigate the possible mechanism of protection and find that a leaky model is the best fit, meaning that immune protection from prior infection is dependent on the force of infection.