Person:

Kissler, Stephen

Importance: New York City is the epicenter of the SARS-CoV-2 pandemic in the United States. Mortality and hospitalizations have differed substantially between different neighborhoods. Mitigation efforts in the coming months will require knowing the extent of geographic variation in SARS-CoV-2 prevalence and understanding the drivers of these differences.

Objective: To estimate the prevalence of SARS-CoV-2 infection by New York City borough between March 22nd and May 3rd, 2020, and to associate variation in prevalence with antecedent reductions in mobility, defined as aggregated daily physical movements into and out of each borough.

Design: Observational study of universal SARS-CoV-2 test results obtained from women hospitalized for delivery.

Setting: Four New York-Presbyterian hospital campuses and two Mount Sinai hospital campuses in New York City.

Participants: 1,746 women with New York City ZIP codes hospitalized for delivery.

Exposures: Infection with SARS-CoV-2.

Main outcomes: Population prevalence of SARS-CoV-2 by borough and correlation with the reduction in daily commuting-style movements into and out of each borough.

Results: The estimated population prevalence of SARS-CoV-2 ranged from 11.3% (95% credible interval 8.9%, 13.9%) in Manhattan to 26.0% (95% credible interval 15.3%, 38.9%) in South Queens, with an estimated city-wide prevalence of 15.6% (95% credible interval 13.9%, 17.4%). The peak city-wide prevalence was during the week of March 30th, though temporal trends in prevalence varied substantially between boroughs. Population revalence was lowest in boroughs with the greatest reductions in morning commutes out of and evening commutes into the borough (Pearson R = –0.88, 95% credible interval –0.52, –0.99).

Conclusions and relevance: Reductions in between-borough mobility predict geographic differences in the prevalence of SARS-CoV-2 infection in New York City. Large parts of the city may remain at risk for substantial SARS-CoV-2 outbreaks. Widespread testing should be conducted to identify geographic disparities in prevalence and assess the risk of future outbreaks.

Establishing how many people have already been infected by SARS-CoV-2 is an urgent priority for controlling the COVID-19 pandemic. Patchy virological testing has hampered interpretation of confirmed case counts, and unknown rates of asymptomatic and mild infections make it challenging to develop evidence-based public health policies. Serological tests that identify past infection can be used to estimate cumulative incidence, but the relative accuracy and robustness of various sampling strategies has been unclear. Here, we used a flexible framework that integrates uncertainty from test characteristics, sample size, and heterogeneity in seroprevalence across tested subpopulations to compare estimates from sampling schemes. Using the same framework and making the assumption that serological positivity indicates immune protection, we propagated these estimates and uncertainty through dynamical models to assess the uncertainty in the epidemiological parameters needed to evaluate public health interventions. We examined the relative accuracy of convenience samples versus structured surveys to estimate population seroprevalence, and found that sampling schemes informed by demographics and contact networks outperform uniform sampling. The framework can be adapted to optimize the design of serological surveys given particular test characteristics and capacity, population demography, sampling strategy, and modeling approach, and can be tailored to support decision-making around introducing or removing interventions.