Determining the best mathematical model for implementation of non-pharmaceutical interventions

At the onset of the SARS-CoV-2 pandemic in early 2020, only non-pharmaceutical interventions (NPIs) were available to stem the spread of the infection. Much of the early interventions in the US were applied at a state level, with varying levels of strictness and compliance. While NPIs clearly slowed the rate of transmission, it is not clear how these changes are best incorporated into epidemiological models. In order to characterize the effects of early preventative measures, we use a Susceptible-Exposed-Infected-Recovered (SEIR) model and cumulative case counts from US states to analyze the effect of lockdown measures. We test four transition models to simulate the change in transmission rate: instantaneous, linear, exponential, and logarithmic. We find that of the four models examined here, the exponential transition best represents the change in the transmission rate due to implementation of NPIs in the most states, followed by the logistic transition model. The instantaneous and linear models generally lead to poor fits and are the best transition models for the fewest states.