Four months into the pandemic, it has become clear that trends in case counts can be misleading as a metric of transmission because they depend to a great extent on testing policies. Thus, it remains challenging to accurately measure transmission rates over time, determine what proportion of the population remains susceptible, and assess the impacts of public health interventions such as social distancing on disease dynamics. Serological data, particularly when collected at high spatial and temporal resolutions, are a key resource as they directly quantify the proportion of the population that has been infected. However, despite the publication of results from multiple serosurveys in recent months, there are still important gaps in how to interpret this type of data and how to properly integrate it into transmission models. Results from recent studies have shown that antibody responses to SARS-CoV-2 decay a few months after infection and vary as a function of disease severity, and these factors further complicate inference from serological data, as these sources of variance need to be formally accounted for. Building on our extensive experience in conducting and analyzing serological data for evaluating transmission dynamics of multiple infectious diseases globally, we propose to a) develop cross-scale analytical frameworks integrating serological data to model the transmission of SARS-CoV-2 over time and space, and b) relate our generated spatio-temporal attack rates to fine-scale mobility data in order to estimate the public health impact of enacting and sustaining specific social distancing interventions.
Our proposal will leverage data generated from two ongoing efforts. In March, we initiated a continuous, scalable system for SARS-CoV-2 serosurveillance throughout the city of San Francisco that takes advantage of remnant blood from routine laboratory tests in the UCSF and Zuckerberg SF General Hospital health networks. We have compiled an incredibly detailed biorepository (>6,000 samples) by algorithmically selecting 500 spatially and age stratified blood samples each week, and plan to continue collecting samples over the next two years as the epidemic progresses. These samples are currently being tested for antibodies against several SARS-CoV-2 antigens using two different immunoassays. This unique serosurveillance effort, generating a high density of data in a focused geographic region over time, is also linked to detailed, individual metadata from electronic medical records. We have also initiated a post-infection longitudinal cohort study that is following individuals for up to two years following SARS-CoV-2 infection. Crucially, the cohort has already enrolled over 100 individuals experiencing a range of clinical severities (from asymptomatic infection to severe disease), and samples have been tested using four distinct immunoassays. This dataset will allow us to accurately model antibody kinetics, which will be fundamental to correctly interpret results from ongoing serosurveillance efforts. By integrating kinetic and serosurveillance data in transmission models, we will be able to reconstruct transmission dynamics at a granular scale and to provide actionable data in real time for public health officials. Finally, we have established a collaboration with Google Research to access weekly mobility data at the 5 km2 level. These data will allow us to both delineate how control measures shape patterns of mobility, as well as to estimate the impact of enacting and sustaining specific interventions on SARS-CoV-2 attack rates.
This proposal is well-aligned with the MIDAS call for New Initiative funding, as our project will produce novel insights into measuring SARS-CoV-2 transmission rates by incorporating individual-level immune dynamics into population-level models, allowing us to accurately assess the impact of control interventions. Upon successful completion of this work, we will have leveraged our real-time SARS-CoV-2 serosurveillance platform to assess fundamental parameters such as the kinetics of the antibody response and population seroprevalence, generated maps of transmission rates, spatial hotspots, and demographic risk groups, and clarified the impacts of control measures on disease dynamics. We will work closely with the San Francisco Department of Public Health to develop our findings into products tailored to inform local public health strategies. The methodologies we propose to develop will be amenable to integration with other data sources as they become available (e.g., results from mass testing efforts), and we envision that our novel approach could serve as a blueprint for adoption by other communities to conduct rapid, targeted surveillance of infectious diseases more broadly.