Estimation of the Timing and Intensity of Reemergence of Respiratory Syncytial Virus Following the COVID-19 Pandemic in the US

Key Points Question What are the factors associated with the timing and intensity of reemergent respiratory syncytial virus (RSV) epidemics following the COVID-19 pandemic? Findings In this simulation modeling study of a simulated population of 19.45 million people, virus introduction from external sources was associated with the spring and summer epidemics in 2021. Reemergent RSV epidemics in 2021 and 2022 were projected to be more intense and to affect patients in a broader age range than in typical RSV seasons. Meaning These findings suggest that the timing and intensity of reemergent RSV epidemics might be different from the usual RSV season, depending on the duration of mitigation measures and the extent of virus introduction from other regions.


eAppendix. Transmission dynamic models
Mathematical models were used to reproduce the annual RSV epidemics before the COVID-19 pandemic based on the inpatient data of New York (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014) and California (2003California ( -2011. Parameters to produce biennial RSV epidemics and year-round RSV activity were taken from models fit to similar datasets from Colorado  and Florida (1989Florida ( -2009, respectively. This model assumes infants are born with transplacentally-acquired antibodies against RSV infections from their mothers (M). As transplacentally-acquired protective antibodies wane, infants become susceptible to infection (S 0 ). Following each infection (I i ), individuals gain partial immunity that lowers both their susceptibility to subsequent infections and the duration and infectiousness of subsequent infections (see eFigure 1). The force of infection for a specific age group , ( ), for time t is defined as: Seasonality in the force of infection is represented by (1 + 1 cos( 2 − 12 )), where 1 is the amplitude of seasonality and ϕ is the seasonal offset. The chance of susceptible individuals in age group being infected is influenced by their contacts with infectious individuals in the entire population. , is the transmission rate from age group k to age group . The proportion of infected individuals and their relative infectiousness at time t is denoted by ( 1, ( ) + 1 2, ( ) + 2 3, ( ) + 2 4, ( ))⁄ ( ), where 1, is the number of infectious individuals of age k during their first infection; 2, , 3, and 4, are the number of infectious individuals who have been infected two, three and four or more times, respectively; 1 and 2 denote the relative infectiousness of the second and subsequent infections; and N k is the total population of age k.
The transmission parameter , can be further decomposed into the age-specific contact probability between age group and per unit time ( , ) and the probability of transmission given contact between an infectious and a susceptible individual (q). Age-specific mixing patterns were obtained from several previous studies, including detailed contact patterns for infants under 1 year of age and location-specific contact patterns. [1][2][3] Age was stratified into thirteen groups: infants younger than 3 months, 3-5 months, 6-8 months, 9-11 months, 1 year, 2 years, 3 years, 4 years, 5-9 years, 10-19 years, 20-39 years, 40-59 years, and ≥60 years.
The disease transmission process is linked to observation-level information. The probabilities of developing lower respiratory tract disease and being hospitalized upon RSV infection are informed by cohort studies conducted in the US and Kenya. 1,[4][5][6][7][8][9][10][11][12][13][14] The number of lower respiratory tract infections (LRI) due to RSV is given by: where ( ) is the force of infection for a specific age group at time t (as defined above). 0, is the number of fully susceptible individuals of age a; 1, , 2, and 3, are the number of susceptible individuals who have been infected once, twice and more times, respectively. 1 , 2 and 3 denote the relative risk of infection following the first, second, and more infections. ℎ 1, , ℎ 2, and ℎ 3, are the proportion of the first, second, and more infections that are hospitalized.
The average age of hospitalization among children under 5 in month t is given by: 15 where the weight is the midpoint of age group .
Several model parameters were fixed based on data from previous cohort and modeling studies. 1,[4][5][6][7][8][9][10][11][12][13][14] We used Bayesian inference to estimate the average duration of transplacentally-acquired immunity, age-specific probability of hospitalization in the 40-59 year and >60 year age groups, the transmissibility coefficient, and seasonal parameters by fitting the model to the hospitalization data from New York and California. 16,17 We identified the best-fit parameter sets by maximum a posteriori estimation. 18 The likelihood was calculated by assuming the observed number of hospitalizations in the entire population was Poisson-distributed with a mean equal to the model-predicted number of hospitalization, and that the observed age distribution was multinomialdistributed with probabilities equal to the model-predicted distribution of RSV hospitalizations in each age group.
To validate our model predictions, we fitted the transmission model to the inpatient data for California from 2003 to 2011; we then compared the model predictions with data on the percent of clinical specimens positive for RSV from a separate sentinel surveillance database from 2012 to 2018. We rescaled the percent positive data by calculating a scaling factor based on overlaying the surveillance data and inpatient data from 2009 to 2011 (see eFigure 3).
We initialized the transmission models with 1 infectious individual in each age group (except for infants under 6 months) in July 1981 and used a burn-in period of 24 years and 22 years in New York and California, respectively. We also performed a sensitivity analysis around what re-emergence might look like in a state with a biennial pattern of epidemics, using parameters fitted to earlier data from Colorado as an example and assuming a linearly declining birth rate (from 17 to 10 births per 1,000 people per year). We used the same number of infectious individuals to initialize transmission model, and a burn-in period of 40 or 41 years starting from 1971 or 1970 to allow for greater incidence in even or odd years.      The numbers on the top show the percentage difference between the expected incidence and the counterfactual incidence in each age group.