Estimating Excess COVID-19 Infections with Nonparametric Self-Exciting Point Processes

TL;DR

This paper introduces a nonparametric self-exciting point process model to estimate excess COVID-19 infections caused by large events, specifically analyzing the impact of Trump’s campaign rallies on local infection rates.

Contribution

It develops an innovative nonparametric approach to quantify excess infections from events, improving causal inference in epidemiological modeling.

Findings

Estimated excess infections attributable to rallies

Identified duration of increased infection risk post-events

Demonstrated model's effectiveness in real-world data analysis

Abstract

The COVID-19 pandemic has led to a vast amount of growth for statistical models and methods which characterize features of disease outbreaks. One class of models that came to light in this regard has been the use of self-exciting point processes, wherein infections occur both "at random" and also more systematically from person-to-person transmission. Beyond the modeling of the overall COVID-19 outbreak, the pandemic has also motivated research assessing various policy decisions and event outcomes. One such area of study, addressed here, relates to the formulation of methods which measure the impact that large events or gatherings of people had in the local areas where the events were held. We formulate an alternative approach to traditional causal inference methods and then apply our method to assessing the impact that then President Donald Trump's re-election campaign rallies had on COVID-19 infections in areas where the rallies were hosted. By incorporating several adaptions to nonparametric self-exciting point process models, we estimate both the excess number of COVID-19 infections brought on by the rallies and the duration of time in which these excess infections persisted.