Modeling for COVID-19 college reopening decisions: Cornell, a case study

Abstract

We consider epidemiological modeling for the design of COVID-19 interventions in university populations, which have seen significant outbreaks during the pandemic. A central challenge is sensitivity of predictions to input parameters coupled with uncertainty about these parameters. Nearly 2 y into the pandemic, parameter uncertainty remains because of changes in vaccination efficacy, viral variants, and mask mandates, and because universities’ unique characteristics hinder translation from the general population: a high fraction of young people, who have higher rates of asymptomatic infection and social contact, as well as an enhanced ability to implement behavioral and testing interventions. We describe an epidemiological model that formed the basis for Cornell University’s decision to reopen for in-person instruction in fall 2020 and supported the design of an asymptomatic screening program instituted concurrently to prevent viral spread. We demonstrate how the structure of these decisions allowed risk to be minimized despite parameter uncertainty leading to an inability to make accurate point estimates and how this generalizes to other university settings. We find that once-per-week asymptomatic screening of vaccinated undergraduate students provides substantial value against the Delta variant, even if all students are vaccinated, and that more targeted testing of the most social vaccinated students provides further value.

SARS-CoV-2 Transmission and Impacts of Unvaccinated-Only Screening in Populations of Mixed Vaccination Status

“…Second, compartmental SEIR models such as ours assume uniform infectiousness in the I compartment, contrasting empirical observations [34] and more sophisticated models [7, 32]. While our model’s latent and infectious periods are well aligned with other SEIR models [9, 35–37], they nevertheless lead to unrealistically long generation times. Decreasing these periods proportionally to achieve the same reproductive number while aligning more closely with generation time estimates [38] would change the time-scale across all simulations, but would not impact the cumulative metrics or dynamics discussed in our key results.…”

Impact of the representation of contact data on the evaluation of interventions in infectious diseases simulations

“…This effect can be assessed using any data representation. Moreover, even in the presence of vaccination, all data representations agree on the ranking of protocols according to their efficacy or cost, and on the fact that NPIs remain very valuable additional tools at intermediate vaccination coverages [20, 43].…”

“…different mixing patterns and/or schedules can be performed from limited information such as the existence of classes in a school or of departments in offices. A population can also be separated in groups according to an expected diversity of behaviours, as for instance in [43] that singles out the group of “more social” students in a US campus as the ones belonging to fraternities and show that targeted testing of this category can be an efficient strategy.…”

“…Although data with a limited resolution were shown to be insufficient to inform interventions at individual scale [40], they are still useful to inform strategies at intermediate scales [41, 42, 14, 15, 43]. In practice however, a general issue faced by models concerns the comparison of strategies or control measures, in terms of both costs and benefits.…”

“…In practice however, a general issue faced by models concerns the comparison of strategies or control measures, in terms of both costs and benefits. In the case of COVID-19 for instance, the computational models mentioned above have considered a wide variety of measures (contact tracing, regular testing, telework, class or school closures), with each study using specific empirical or synthetic data and a specific representation of contacts [17, 20, 44, 45, 22, 19, 46, 47, 48, 21, 43]. However, just as the data representation can affect the identification of risk groups [37], it might also impact the ranking of different strategies.…”