In March 2020 the COVID-19 pandemic took hold in the United States and threatened the lives a livelihood of all Americans. At this time, the pandemic is being driven by person-to-person transmission of the coronavirus known as SARS-CoV-2. The U.S. government, both state and federal, responded to the pandemic by launching social distancing as an intervention, ordering schools and non-essential businesses to close throughout most of the country. Mathematical models are required in order to measure the transmission of COVID-19 in the U.S. This research focuses on building mathematical models that include data on social distancing to measure how effective the intervention is at slowing the disease. The researchers will also evaluate strategies for reopening schools and workplaces. Importantly, the researchers will measure the risk for a second pandemic wave by accounting for how the immune system reacts to the infection. This research is important because it will provide the U.S. government with models that it can use to choose among options for reopening cities. These models will also reveal the number of deaths averted by social distancing policies. In addition to the great societal benefit of this research, it also brings scientific advancement by demonstrating how novel datasets can be collected in real-time and models can be deployed during a public health emergency.
Transmission models will be used to explore creative ways for phased reopening of cities in order to minimize disease-induced mortality and overburdening of hospitals. The researchers will focus on the 62 counties in New York state, the current epicenter of the pandemic. The researchers will fit a city-level COVID-19 transmission model which accounts for social distancing quantified by Google traffic data, and ground-truthed by public transportation data, live-webcam streams, and Google trends indicating individuals are staying home. The researchers will explore transmission dynamics and hospitalization trajectories under different scenarios of adaptive immunity (e.g., long-lived sterilizing immunity, waning immunity, and immunity that reduces symptoms in subsequent infections). Models will be parameterized using data on testing, COVID-19 clinical cases, hospitalizations, and mortality via Maximum Likelihood by Iterated Particle Filtering (MIF). Modeling will be done in real time, requiring the statistical inference pipeline to be sufficiently nimble to account for the rapidly changing epidemiological situation. The researchers are interfacing with policy makers as part of CDC working groups on modeling COVID-19 and will actively share data through the NIH MIDAS coordination network.