SARS-CoV-2 dynamics in New York City during March 2020-August 2023

Abstract

Background: The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has been widespread since 2020 and will likely continue to cause substantial recurring epidemics. However, understanding the underlying infection burden and dynamics, particularly since late 2021 when the Omicron variant emerged, is challenging. Here, we leverage extensive surveillance data available in New York City (NYC) and a comprehensive model-inference system to reconstruct SARS-CoV-2 dynamics therein through August 2023.

Methods: We fit a metapopulation network SEIRSV (Susceptible-Exposed-Infectious-(re)Susceptible-Vaccination) model to age- and neighborhood-specific data of COVID-19 cases, emergency department visits, and deaths in NYC from the pandemic onset in March 2020 to August 2023. We further validate the model-inference estimates using independent SARS-CoV-2 wastewater viral load data.

Results: The validated model-inference estimates indicate a very high infection burden-the number of infections (i.e., including undetected asymptomatic/mild infections) totaled twice the population size ( > 5 times documented case count) during the first 3.5 years. Estimated virus transmissibility increased around 3-fold, whereas estimated infection-fatality risk (IFR) decreased by >10-fold during this period. The detailed estimates also reveal highly complex variant dynamics and immune landscape, and higher infection risk during winter in NYC over the study period.

Conclusions: This study provides highly detailed epidemiological estimates and identifies key transmission dynamics and drivers of SARS-CoV-2 during its first 3.5 years of circulation in a large urban center (i.e., NYC). These transmission dynamics and drivers may be relevant to other populations and inform future planning to help mitigate the public health burden of SARS-CoV-2.

Fig. 1. Model fit and validation.

Upper panel shows model fit to weekly number of COVID-19 cases (a), emergency department (ED) visits (b), and deaths (c), during the week starting 03/01/20 (mm/dd/yy) to the week starting 08/27/23 (see x-axis). Blue lines show the median estimates and blue areas show 50% (darker) and 95% (lighter) credible intervals (CrIs); dots show the corresponding observations. Lower panel shows model validation using wastewater surveillance data, for the 2^nd wave (d), Delta wave (e), the Omicron period (f). Lines and shaded areas show the estimated infection prevalence (i.e., the number of all infectious individuals including those not detected as cases; median, 50% and 95% CrIs; left y-axis). Dots show measured SARS-CoV-2 concentrations in wastewater (right y-axis, in million copies per day per population) for the corresponding weeks (black dots show measurements using RT-qPCR and red dots show measurements using RT-dPCR but converted to RT-qPCR equivalents; note that the wastewater concentrations are scaled for each wave/period to facilitate comparison with model estimates; see Methods for details).

Fig. 2. Key model estimates.

a shows estimated infection rates; b shows estimated population immunity dynamics; and (c) shows estimated virus transmissibility. In (a), colored bars show estimated median weekly infection rates, for each variant (see legends). In (b), we overlay estimated population susceptibility [left y-axis; blue line = median, blue areas = 50% (darker) and 95% (lighter) CrIs], and proxies of cumulative infection (colored stacked bars from top to bottom, right y-axis; same legends as in A for different variants) and vaccine-induced immunity against infection (open bars; see Methods). In (c), we show estimated virus transmissibility [left y-axis; blue line = median, blue areas = 50% (darker) and 95% (lighter) CrIs] and infection rates [boxplot and right y-axis; middle bar = median, edges = 50% CrIs, and whiskers = 95% CrIs] for the corresponding weeks.

Fig. 3. Infection annual pattern and detection rates.

a shows estimated infection rates (light blue bars, full height; i.e., not stacked) and reported case rates (darker blue portion) by month; error bars show estimated 95% CrIs. To examine the infection annual pattern, b shows the monthly infection rates scaled to the annual maximum (here, a year starts in September, the start of fall/cold months in the Northern hemisphere, and ends in the next August, the end of winter/cold months in Southern hemisphere). March–August 2020 is not shown due to the incompleteness. c shows estimated infection-detection rate [blue line = median, blue areas = 50% (darker) and 95% (lighter) CrIs], for each week. Note the vertical dashed line indicates the week starting 11/21/21 when Omicron BA.1 was first detected in NYC, and estimates to the right of the dashed line are for Omicron (sub)variants alone.

Fig. 4. Estimated infection-fatality risk (IFR) over time.

a–e show estimates by age group and (f) shows the estimates combining all ages. Blue lines and shaded areas show the median estimates and 50% (darker blue) and 95% (lighter blue) CrIs, for each week (see date of week start in mm/dd/yy in the x-axis). For clarity, insets show estimates during the most recent months.