Reconstructing the impact of COVID-19 on the immunity gap and transmission of respiratory syncytial virus in Lombardy, Italy

Abstract

Background: Respiratory syncytial virus (RSV) is a leading cause of hospitalisation and mortality in young children globally. The social distancing measures implemented against COVID-19 in Lombardy (Italy) disrupted the typically seasonal RSV circulation during 2019-2021 and caused substantially more hospitalisations during 2021-2022. The primary aim of this study is to quantify the immunity gap-defined as the increased proportion of the population naïve to RSV infection following the relaxation of COVID-19 restrictions in Lombardy, which has been hypothesised to be a potential cause of the increased RSV burden in 2021-2022.

Methods: We developed a catalytic model to reconstruct changes in the age-dependent susceptibility profile of the Lombardy population throughout the COVID-19 pandemic. The model is calibrated to routinely collected hospitalisation, syndromic, and virological surveillance data and tested for alternative assumptions on age-dependencies in the risk of RSV infection throughout the pandemic.

Findings: We estimate that the proportion of the Lombardy population naïve to RSV infection increased by 60.8% (95% CrI: 55.2-65.4%) during the COVID-19 pandemic: from 1.4% (95% CrI: 1.3-1.6%) in 2018-2019 to 2.3% (95% CrI: 2.2-2.5%) before the 2021-2022 season, corresponding to an immunity gap of 0.87% (95% CrI: 0.87-0.88%). We found evidence of heterogeneity in RSV transmission by age, suggesting that the COVID-19 restrictions had variable impact on the contact patterns and risk of RSV infection across ages.

Interpretation: We estimate a substantial increase in the population-level susceptibility to RSV in Lombardy during 2019-2021, which contributed to an increase in primary RSV infections in 2021-2022.

Funding: UK Medical Research Council (MRC), UK Foreign, Commonwealth & Development Office (FCDO), EDCTP2 programme, European Union, Wellcome Trust, Royal Society, EU-MUR PNRR INF-ACT.

Keywords: COVID-19 restrictions; Catalytic models; Immunity gap; Mathematical modelling; RSV.

Fig. 1

Weekly RSV data from 2018–2019 to 2021–2022. Panels show: (a) hospital discharges, (b) average reconstructed RSV-attributable ILIs, and (c) RSV test positivity rates. RSV, respiratory syncytial virus; ILI, influenza-like illness.

Fig. 2

Summary of model fit to data. Observed and modelled (a) number of hospital discharges associated with RSV, (b) age-distribution of hospital discharges (age 0 refers to 0–6-month-olds and age 0.5 refers to 7–12-month-olds), (c) number of RSV-attributable ILI cases, (d) age-distribution of RSV-attributable ILI cases, and (e) RSV test positivity ratio. In all panels, error bars represent the 95% exact binomial CIs (where applicable) and the 95% CrIs for the model estimates. RSV, respiratory syncytial virus; ILI, influenza-like illness; CIs, confidence intervals; CrIs, credible intervals.

Fig. 3

Model estimates. Estimated (a) FOI for the three age-groups 0–4, 5–14, and 15+ years, (b) per-capita RSV transmission rates, as estimated by dividing the FOI model estimates by the cumulative number of RSV-attributable ILI cases reported each season (y-axis on a log-scale); and probability of (c) hospitalisation, (d) a case being symptomatic and reported to surveillance, and (e) test positive ratio associated with RSV. In all panels, points represent the mean and error bars represent 95% CrIs. FOI, force of infection; RSV, respiratory syncytial virus; ILI, influenza-like illness; CrIs, credible intervals.

Fig. 4

Attack rates and estimates of the proportion of the population naïve to RSV. Estimated (a) RSV attack rates by age-group per season (bars represent the mean, and error bars the 95% CrIs) and (b) proportion of Lombardy population naïve to RSV infection at the start of each season. RSV, respiratory syncytial virus; CrIs, credible intervals.

Supplementary Figure S1

Summary of model fit to data for scenarios with MDI. Observed and estimated (a) number of hospital discharges associated with RSV, (b) age-distribution of hospital discharges (age 0 refers to 0–6-month-olds and age 0.5 refers to 7–12-month-olds), (c) number of RSV-attributable ILI cases, (d) age-distribution of RSV-attributable ILI cases, and (e) RSV test positivity ratio. In all panels, error bars represent the 95% exact binomial CIs around the data (where applicable) and the 95% CrIs for the model estimates. MDI, maternally derived immunity; RSV, respiratory syncytial virus; ILI, influenza-like illness; CIs, confidence intervals; CrIs, credible intervals.

Supplementary Figure S2

Summary of model fit to data for scenarios without MDI. Observed and estimated (a) number of hospital discharges associated with RSV, (b) age-distribution of hospital discharges (age 0 refers to 0–6-month-olds and age 0.5 refers to 7–12-month-olds), (c) number of RSV-attributable ILI cases, (d) age-distribution of RSV-attributable ILI cases, and (e) RSV test positivity ratio. In all panels, error bars represent the 95% exact binomial CIs around the data (where applicable) and the 95% CrIs for the model estimates. MDI, maternally derived immunity; RSV, respiratory syncytial virus; ILI, influenza-like illness; CIs, confidence intervals; CrIs, credible intervals.

Supplementary Figure S3

Proportion naïve to RSV over time. Estimates obtained with the baseline model (left), and the sensitivity analysis with Model D and no MDI (right). MDI, maternally derived immunity; RSV, respiratory syncytial virus.

Supplementary Figure S4

Summary of model fit to data for ahypothetical25% and 50% increase in reporting during pandemic affected seasons versus baseline. Observed and estimated (a) number of hospital discharges associated with RSV, (b) age-distribution of hospital discharges (age 0 refers to 0–6-month-olds and age 0.5 refers to 7–12-month-olds), (c) number of RSV-attributable ILI cases, (d) age-distribution of RSV-attributable ILI cases, and (e) RSV test positivity ratio. In all panels, error bars represent the 95% exact binomial CIs around the data (where applicable) and the 95% CrIs for the model estimates. RSV, respiratory syncytial virus; ILI, influenza-like illness; CIs, confidence intervals; CrIs, credible intervals.

Supplementary Figure S5

Estimated proportion naïve to RSV over time. Estimates obtained with the baseline model (left), and the two sensitivity analyses with 25% increased reporting (central) and 50% increased reporting (right). RSV, respiratory syncytial virus.

Supplementary Figure S6

Parameter estimates for the baseline and calibrated$\mathit{r}$ (reduced probability of hospitalisation for post-primary infections) models. (a) Probability of hospitalisation, (b) probability of reporting to surveillance, and (c) Test positive probability.

Supplementary Figure S7

Summary of model fit to data for calibrated$\mathit{r}$(reduced probability of hospitalisation for post-primary infections)versus baseline model. Observed and estimated (a) number of hospital discharges associated with RSV, (b) age-distribution of hospital discharges (age 0 refers to 0–6-month-olds and age 0.5 refers to 7–12-month-olds), (c) number of RSV-attributable ILI cases, (d) age-distribution of RSV-attributable ILI cases, and (e) RSV test positivity ratio. In all panels, error bars represent the 95% exact binomial CIs around the data (where applicable) and the 95% CrIs for the model estimates. RSV, respiratory syncytial virus; ILI, influenza-like illness; CIs, confidence intervals; CrIs, credible intervals.

Supplementary Figure S8

Proportion naïve to RSV over time. Estimates obtained with the baseline model (left) and the sensitivity analysis with a reduced probability of hospitalisation for post-primary infections (right). RSV, respiratory syncytial virus.