Behaviour, booster vaccines and waning immunity: modelling the medium-term dynamics of SARS-CoV-2 transmission in England in the Omicron era

Abstract

England has experienced a heavy burden of COVID-19, with multiple waves of SARS-CoV-2 transmission since early 2020 and high infection levels following the emergence and spread of Omicron variants since late 2021. In response to rising Omicron cases, booster vaccinations were accelerated and offered to all adults in England. Using a model fitted to more than 2 years of epidemiological data, we project potential dynamics of SARS-CoV-2 infections, hospital admissions and deaths in England to December 2022. We consider key uncertainties including future behavioural change and waning immunity, and assess the effectiveness of booster vaccinations in mitigating SARS-CoV-2 disease burden between October 2021 and December 2022. If no new variants emerge, SARS-CoV-2 transmission is expected to decline, with low levels remaining in the coming months. The extent to which projected SARS-CoV-2 transmission resurges later in 2022 depends largely on assumptions around waning immunity and to some extent, behaviour and seasonality.

Figure 1.. Compartmental model diagram.

A three-variant deterministic dynamic compartmental model with vaccination describes SARS-CoV-2 transmission in England. We model seven NHS England regions separately, with each divided into 16 five-year age groups: 0–4 years up to 70–74 years, and 75 years and older. The model incorporates COVID-19 vaccination with two vaccine products (corresponding to the viral-vector (Va) and mRNA-based (Vb) vaccines in use in England), each with first- (Va1, Vb1) and second-dose (Va2, Vb2) protection, and each with lower levels of protection for individuals who received their primary vaccinations but no booster vaccine and have waned (Va2w, Vb2w). All vaccinated individuals have increased protection against different SARS-CoV-2 outcomes compared to susceptible individuals, according to the vaccine product administered and their vaccine dose/waned status (Table S2). Vaccinated individuals transition from the susceptible (S) compartment into first-dose vaccinated compartments (Va1, Vb1) depending on which vaccine product was received. Following an assumed first-dose duration (Table S5B), individuals move into second-dose compartments (Va2, Vb2). Following an assumed second-dose duration (Table S5B), individuals either receive a booster vaccine or transition into waned states (Va2w/Vb2w). We assume that individuals receiving primary courses of a viral-vector vaccine (Va2) or an mRNA vaccine (Vb2) both move to the Vb2 compartment following their booster vaccination, with their second-dose duration beginning again from zero (Table S5B). This assumptions reflects the fact that all booster vaccinations in England are either the Pfizer/BioNTech or Moderna mRNA vaccines, and evidence finding higher immunogenicity for individuals receiving Pfizer/BioNTech following Oxford-AstraZeneca, compared with individuals receiving both Oxford-AstraZeneca vaccine doses. We model an additional temporary increase in vaccine protection for individuals receiving their first booster vaccine in late 2021/early 2022, which lasts for 180 days. We utilise three separate SARS-CoV-2 variants in the model to capture the introduction and spread of the Alpha, Delta and Omicron variants of concern in England. The model assumes a traditional infection process: upon being infected, individuals leave the susceptible (S), vaccinated (V) or recovered (R) states and move through exposed (E), infectious (I) and recovered (R) states. The latent (L) state is used in addition to the exposed (E) state for breakthrough infections following vaccination and for re- and cross-infections, to achieve additional vaccine protection against disease (Tables S2, S3). When individuals are infectious (I), they either progress through a subclinical (Is) pathway or a clinical pathway with pre-clinical (Ip) and clinical (Ic) states. Once individuals have been infected and recover (R), we allow for loss of immunity (where individuals return to a susceptible (S) state) and re- and cross-infections (where individuals with immunity become infected, see Table S3). On the left hand side (yellow shaded background labelled I), solid black arrows represent primary vaccinations, solid coloured arrows represent booster vaccinations, and dotted black arrows represent loss of immunity. Coloured dash-dotted arrows denote susceptible and vaccinated individuals becoming infected and moving into the SARS-CoV-2 infection process (grey boxes on red shaded background labelled II). Here, solid black arrows denote individuals moving through the infection process and recovering (R) (purple boxes on purple shaded background labelled III). Recovered individuals can lose their immunity (dotted black arrows) and return to the susceptible disease state (S) (Table S4) or be re- or cross-infected (dashed grey arrows) with other SARS-CoV-2 variants (Table S3).

Figure 2.. Comparison between aggregated model fits and epidemiological data from England between March 2020 and May 2022.

Black lines show reported data, with black ribbons showing 95% confidence intervals for PCR prevalence. Coloured lines and shaded areas show medians, 50% and 90% interquantile ranges from the fitted model. The original model fitting is done independently for each NHS England region (see Figs. S1A–B), with the aggregated model output for the whole of England shown here. (a) COVID-19 deaths over time, where data was provided by the UK Health Security Agency (UKHSA). (b) COVID-19 hospital admissions over time, where data was provided by NHS England. (c) COVID-19 hospital bed occupancy over time, where data was provided by NHS England. (d) COVID-19 ICU bed occupancy over time, where data was provided by NHS England. (e) COVID-19 PCR prevalence over time, where publicly-available PCR prevalence data was obtained from the Office for National Statistics’ COVID-19 Infection Survey (ONS-CIS). The data sources for COVID-19 deaths, hospital admissions, hospital and ICU bed occupancy are unpublished and not publicly available, but are closely aligned with the UK Government’s COVID-19 dashboard. ICU = intensive care unit. NHS = National Health Service.

Figure 3.. Mobility scenarios, transmission adjustments and overall transmission potential for the fitted model, shown from March 2020 to December 2022.

Top: Historic Google Community Mobility data (grey) and assumed future mobility in England for no change (pink), a 3-week return to pre-pandemic baseline levels (green), a 3-month return to pre-pandemic baseline levels (orange) and a 6-month return to pre-pandemic baseline levels (purple) scenarios used for model projections. Mobility indices are measured relative to baseline mobility levels recorded during early 2020, prior to the COVID-19 pandemic. The beginning of each lockdown and each roadmap Step is marked with a vertical dashed line and ‘L’ and ‘S’ labels, respectively. Vertical dashed lines with ‘PBA’ and ‘PBE’ labels correspond to the announcement of ‘Plan B’ measures for England on 8th December 2021 and the ending of these measures on 27th January 2022. Middle: Fitted transmission adjustments between April 2020 and May 2022 by NHS England region (coloured lines) and the average across regions (black line), example projection between May and December 2022 for East of England (blue line) and mean (black line) and interquartile range (red shaded) for projected transmission adjustments between May and December 2022 across NHS England regions. Bottom: The overall “transmission potential” captures the combined impact of mobility and transmission adjustments on the time-varying potential for effective transmission, ignoring the impact of immunity and novel variants, though including the impact of school vacation periods. NE & Y = North East & Yorkshire. NHS = National Health Service.

Figure 4.. Summary of basecase model fits and projections and key results on uncertainty, behaviour, booster vaccinations and waning immunity.

(a) The number of COVID-19 hospital admissions in England, for the basecase scenario, between March 2020 and December 2022. Black lines show reported data, provided by UKHSA. Coloured lines and shaded areas show medians, 50% and 90% interquantile ranges from the fitted model and from the model projection. (b) PCR prevalence in England, for the basecase scenario, between March 2020 and December 2022. The black ribbons show 95% confidence intervals for PCR prevalence data. Coloured lines and shaded areas show medians, 50% and 90% interquantile ranges from the fitted model and from the model projection. (c) The fitted and projected number of COVID-19 deaths in England between March and December 2022, shown for the very high waning scenario (see Table S4). The black line shows the median trajectory of COVID-19 deaths in England over time, with the shaded areas showing the 50% and 90% interquantile ranges. Individual model trajectories are plotted in coloured lines. (d) The effect of future behaviour on COVID-19 deaths and cumulative deaths (thousands) over time is shown with four scenarios for future mobility: a 3-week, a 3-month and a 6-month return to baseline levels, and a no change scenario (see Table 1). (e) The effect of booster vaccination policy on cumulative infections and deaths since October 2021 is shown with four scenarios for booster policies (Table 1). (f) The effect of waning immunity on cumulative infections and deaths between March and December 2022 is shown with three scenarios for waning (Tables 1, S4). The basecase scenarios (shown in panels a and b) and scenarios marked with an asterisk (*) are equivalent.

Figure 5.. Summary of projected cumulative numbers (log-scale) of COVID-19 deaths (thousands), hospital admissions (thousands) and infections (millions) in England between May and December 2022, across behavioural, waning, seasonality and vaccination scenarios considered.

Each box plot shows the projected median, 5th, 25th, 75th and 95th percentile values across all simulations for the relevant scenario, calculated between May and December 2022. Scenarios are coloured according to the result type (from left to right: behaviour, waning immunity, seasonality, and vaccination policies for children aged 5 years and older). A full list of scenarios and relevant modelling assumptions is given in Table 1. Scenarios marked with an asterisk (*) are equivalent and correspond to the basecase scenario. N.B. The y-axes are plotted on a log scale, and are truncated and do not extend to zero.