Lives saved and hospitalizations averted by COVID-19 vaccination in New York City: a modeling study

We extended our previously established age-stratified agent-based model of COVID-19,

incorporating Alpha (B.1.1.7), Gamma (P.1), Delta (B.1.617.2) and Iota (B.1.526) variants as identified in NYC, along with the original Wuhan-1 strain. The model implemented the natural history of disease (Appendix, Figure S1) with epidemiological classes for susceptible; latently infected (not yet infectious); asymptomatic (and infectious); pre-symptomatic (and infectious); symptomatic (and infectious) with either mild or severe illness; recovered; and dead. Disease-specific parameters were sampled from their associated distributions (Appendix, Table S1). The model population was stratified into five age groups resembling the demography of New York City. Contact patterns between and within age groups were derived from empirical data from the pre-pandemic and pandemic eras.

Risk of infection for susceptible individuals depended on their interaction with infectious individuals in the pre-symptomatic, symptomatic, or asymptomatic stages of infection. The transmission probability was calibrated by fitting the model to the reported incidence per 100,000 population in NYC from September 1, 2020 to July 15, 2021, accounting for the chronological introduction of the Iota, Alpha, Gamma, and Delta variants, as well as the daily vaccine doses administered to each age class in NYC. In this process, the transmission probability of the original Wuhan-I strain of SARS-CoV-2 was calibrated to an effective reproductive number of 1.17 as reported in early October 2020,

accounting for the effect of non-pharmaceutical interventions (NPIs) and 50% lower rate of contacts within and between age-groups (compared to pre-pandemic normal behavior). After calibration of the transmission probability for the original strain, we adjusted the age-specific contact rates throughout the simulations to minimize the difference between the mean cumulative incidence predicted by the model using Monte-Carlo replications and the cumulative reported cases on each day along with simulation timelines. We also adjusted the transmissibility of different variants relative to the original strain after their introduction. The Iota variant emerged in New York State in early September with an estimated 35% higher transmissibility compared with the original Wuhan-I strain.

We then introduced the Alpha variant on January 3, 2021, with a 50% higher transmissibility compared to the original strain.

The Delta variant was introduced on February 20, 2021, as reported by the NYC Department of Health and Mental Hygiene, with an elevated transmissibility of 30% compared with the Alpha variant.

Finally, we inserted the Gamma variant on March 20, 2021, with 60% higher transmissibility compared to the original strain.

We simulated COVID-19 outbreak in NYC from October 1, 2020 to July 15, 2021 assuming 20% pre-existing immunity due to infections prior to October 2020.

Asymptomatic, mild symptomatic, and severe symptomatic individuals were considered to be 26%, 44%, and 89% infectious relative to those in the pre-symptomatic stage.

We assumed that these relative infectivities remained the same for all variants in the model.

Recent studies suggest that antibodies from prior infection with the original strain may have reduced neutralizing activity against Gamma and Delta.[

We assumed that these variants evade naturally acquired immunity by an average of 21% (95% CI: 11%-36%).

This evasion rate was implemented as a 79% reduction in the risk of acquiring infection by the Gamma or Delta variant for individuals who had previously recovered from the original strain, Iota, or Alpha. We further assumed that recovery from infection with the Gamma or Delta variant provides protection against all variants in the model, preventing reinfection for at least one year.

We assumed that asymptomatic and mild symptomatic cases recover from infection without hospitalization. A proportion of those with severe disease were hospitalized within 2-5 days of symptom onset

and were therefore removed from the transmission chain. Symptomatic cases who were not hospitalized self-isolated within 24 hours of symptom onset, and reduced their number of daily contacts by an additional 72% (Appendix, Table S2). Intensive care unit (ICU) and non-ICU hospitalization rates were parameterized (Appendix, Table S3) based on clinical and epidemiological data stratified by age and comorbidities in the US.

Infection with Alpha variant was associated with 64% higher risk of death, while Iota, Gamma, Delta variants are associated with the same case fatality risk as the original Wuhan-I variant.

Starting on December 14, 2020, vaccination was implemented as a two-dose program with the reported daily number of vaccines distributed as first and second doses in NYC.

For the Pfizer-BioNTech and Moderna vaccines, intervals of 21 and 28 days between the first and second doses were implemented, respectively.

We used published estimates to parameterize vaccine efficacies following each dose of vaccines against infection, symptomatic disease, and severe disease caused by the original strain (Appendix, Table S4).[]

These efficacies were implemented as a reduction in the probability of acquiring infection, probability of developing symptomatic disease if infection occurred, and probability of severe illness if symptomatic disease developed. Until April 2, 2021, age-specific contact rates of all individuals were based on adjusted rates of the fitting process. From April 3, 2021, based on the guidelines by the US Centers for Disease Control and Prevention (CDC), the daily contact rates for fully vaccinated individuals were set to age-specific pre-pandemic patterns 14 days after the second dose of vaccine.

We implemented the model in Julia, and calibrated with case incidence data to simulate a counterfactual scenario without vaccination. We ran 500 independent simulations to obtain the daily incidence, hospitalizations, and deaths with their associated ranges, and used a bias-corrected and accelerated bootstrap method to derive the estimates of averted outcomes and their 95% credible intervals resulting from vaccination in NYC. The computational codes for reproducibility are available at https://github.com/thomasvilches/multiple_strains/tree/NYC.

Dr. Schneider is an employee of the funding organization and participated in the formulation of research questions, interpretation of results, and review of the manuscript, but not the conduct, data collection, or analysis. The Commonwealth Fund did not play any role in the decision to submit the manuscript for publication.