The potential health and economic value of SARS-CoV-2 vaccination alongside physical distancing in the UK: a transmission model-based future scenario analysis and economic evaluation

Summary Background In response to the COVID-19 pandemic, the UK first adopted physical distancing measures in March, 2020. Vaccines against SARS-CoV-2 became available in December, 2020. We explored the health and economic value of introducing SARS-CoV-2 immunisation alongside physical distancing in the UK to gain insights about possible future scenarios in a post-vaccination era. Methods We used an age-structured dynamic transmission and economic model to explore different scenarios of UK mass immunisation programmes over 10 years. We compared vaccinating 75% of individuals aged 15 years or older (and annually revaccinating 50% of individuals aged 15–64 years and 75% of individuals aged 65 years or older) to no vaccination. We assumed either 50% vaccine efficacy against disease and 45-week protection (worst-case scenario) or 95% vaccine efficacy against infection and 3-year protection (best-case scenario). Natural immunity was assumed to wane within 45 weeks. We also explored the additional impact of physical distancing on vaccination by assuming either an initial lockdown followed by voluntary physical distancing, or an initial lockdown followed by increased physical distancing mandated above a certain threshold of incident daily infections. We considered benefits in terms of quality-adjusted life-years (QALYs) and costs, both to the health-care payer and the national economy. We discounted future costs and QALYs at 3·5% annually and assumed a monetary value per QALY of £20 000 and a conservative long-run cost per vaccine dose of £15. We explored and varied these parameters in sensitivity analyses. We expressed the health and economic benefits of each scenario with the net monetary value: QALYs × (monetary value per QALY) – costs. Findings Without the initial lockdown, vaccination, and increased physical distancing, we estimated 148·0 million (95% uncertainty interval 48·5–198·8) COVID-19 cases and 3·1 million (0·84–4·5) deaths would occur in the UK over 10 years. In the best-case scenario, vaccination minimises community transmission without future periods of increased physical distancing, whereas SARS-CoV-2 becomes endemic with biannual epidemics in the worst-case scenario. Ongoing transmission is also expected in intermediate scenarios with vaccine efficacy similar to published clinical trial data. From a health-care perspective, introducing vaccination leads to incremental net monetary values ranging from £12·0 billion to £334·7 billion in the best-case scenario and from –£1·1 billion to £56·9 billion in the worst-case scenario. Incremental net monetary values of increased physical distancing might be negative from a societal perspective if national economy losses are persistent and large. Interpretation Our model findings highlight the substantial health and economic value of introducing SARS-CoV-2 vaccination. Smaller outbreaks could continue even with vaccines, but population-wide implementation of increased physical distancing might no longer be justifiable. Our study provides early insights about possible future post-vaccination scenarios from an economic and epidemiological perspective. Funding National Institute for Health Research, European Commission, Bill & Melinda Gates Foundation.


Supplementary Material Physical distancing scenarios
For scenarios with an initial lockdown, the incidence threshold for the initial lockdown and physical distancing was based on aligning the results for the estimated numbers of COVID-19 related hospitalisations and deaths to the observed numbers in the UK. As of 15 July 2020, there were 130,472 recorded admissions and 41,035 deaths due to COVID-19 in the UK. 1 The model most closely resembled these values when going into lockdown once the incidence reached 30 cases per 100,000 population and using the values shown in Supplementary Table 1, with an estimated 136,500 admissions and 37,128 deaths by 15 July 2020.
There are many unknowns surrounding the characteristics of the vaccine candidates, uncertain aspects of the longer-term COVID-19 epidemiology, what measures will be put in place in future, and the COVID-19-specific impact on costs and QALYs. However, in this economic evaluation a higher value was placed on approaching the absolute numbers as observed historically rather than closely resembling the observed disease dynamics, which are difficult to predict in future, too. Although the modelled scenarios of physical distancing may not perfectly predict future disease dynamics for the next decade, physical distancing itself was not the focus of this study evaluating the impact of SARS-CoV-2 vaccination. However, ignoring the wider economic impact of physical distancing risks distorting conclusions as an indefinite lockdown may indeed help reduce the health burden to a minimum and at minimal healthcare costs, but at substantial harm to the wider economy and society.
Supplementary Note: The relative scaling of contacts during the initial lockdown was chosen to achieve a 75% reduction of Rt as observed historically in the UK in 2020, 3 while closely matching observed admissions in the UK in mid-July 2020. Future summer and winter holidays were chosen in line with the dates of school holidays in England in 2020/2021.

Burden estimation
The first two COVID-19 vaccines to have dossiers submitted to authorise supply in the UK market use an mRNA platform (Pfizer, BioNTech) and an adenovirus vector (AstraZeneca, Oxford University), respectively. 4,5 The trial data for the Pfizer-BioNTech vaccine demonstrated vaccine efficacy in adults aged 16+ years of 95.0% (95% credible interval of 90.3% to 97.6%), and in individuals aged 65+ years the vaccine efficacy was 94.7% (two-sided 95% confidence interval of 66.7% to 99.9%). 6-8 Of note, the dossier submitted to the FDA also reported a vaccine efficacy of 52.4% (29.5% to 68.4%) after the first dose and before the second dose, indicating some protection against disease after one dose. 7,8 High vaccine efficacy has also been reported for other vaccines, 5,9 with the AstraZeneza-Oxford University vaccine showing a vaccine efficacy in adults aged 18+ years of up to 90.0% (67.4% to 97.0%) in participants who received a low dose followed by a standard dose of the vaccine. 10 Preliminary results of the trial also reported protection against asymptomatic infection with a vaccine efficacy of 58.9% (95%-CI 1.0% to 82.9%). 10 A third vaccine using an mRNA platform has been authorised on 08 January 2021 (Moderna), which has lower cold chain requirements than the Pfizer-BioNTech vaccine at an overall vaccine efficacy in adults aged 18+ years of 94.1% (95% credible interval of 89.3% to 96.8%). 11 Based on these initial trial data we thus evaluated a vaccine that can be given to all individuals aged 15+ years, which is reflective of (1) the current guidance of vaccination prioritisation for the rollout in the UK that targets vaccinating clinically extremely vulnerable individuals from 16+ years; 4 (2) the paediatric vaccine trials that are planned and are likely to get a push given that younger ages are becoming increasingly linked to transmission; and (3) the timeframe of our analysis spanning the next 10 years.
Note that the transmission model is stratified into 16 age bands of 5 years each (starting from 0-4 years and ending at 75+ years), and for technical reasons thus the strategies started at 15+ years. However, we have provided additional sensitivity analyses that explored alternative vaccination strategies starting vaccinating from 20+ years or indeed 10+ years (should the future licensure permit it); Supplementary Figure 6. Furthermore, it is important to note that while the AstraZeneca vaccine trials found similar efficacy when participants who were seropositive were included, most of the clinical trials published so far have been conducted in COVID-naïve individuals, and the real-world vaccine effectiveness will need to be established later. However, the true vaccine effectiveness will fall within the extreme range of bestcase to worst-case scenarios we have considered in this study.
Of note, while any infection event (i.e., involving either symptomatic or asymptomatic disease progression) can be an important driver of transmission between individuals, the primary endpoint of many clinical trials for vaccines has been "symptomatic infection" instead of "asymptomatic infection".
There is no evidence that any COVID-19 vaccine blocks transmission in vaccinees who are infected. However, they may contribute to decreasing transmission by preventing infection in the first place and/or decreasing the severity (and hence transmissibility) of infections. This is an important semantic distinction helping to clarify the working of the COVID-19 vaccines and the focus of the clinical trial data published so far.  5,10 In this optimistic scenario, we assume that disease averted also eliminates the possibility of infectiousness. There is some evidence of reduced viral load and protection against infection from animal studies of COVID-19 vaccines. 12, 13 Note that we assumed 0% protection against disease for the remaining 5%, but if the vaccines are able to prevent disease then the benefits of the best-case scenario are expected to be even slightly higher than we estimated. Vaccine effectiveness in base case analysis: worst-case scenario 0% against infection 50% additional protection against disease Based on (i) the minimum requirements for vaccine licensure as formulated by the FDA; 14 , (ii) the reported vaccine efficacy of 52.4% after the first dose and before the second dose for BNT162b2; 8 , (iii) and the values usually observed for seasonal influenza vaccines. 15 Note that we assumed 0% net protection against infection. This assumes that all disease cases prevented will be converted to asymptomatic infection with half the transmissibility. This is not reflective of pre-clinical trial data or of the AZD1222 trial results that included asymptomatic infection, 10 but in the absence of longer and more robust clinical trials this assumption is more conservative in reflecting a worst-case scenario. Vaccine effectiveness in sensitivity analysis: AZD1222 (AstraZeneca) vaccine best-case scenario 70% against infection 70% additional protection against disease (91% overall disease protection)

Supplementary
Based on the upper-bound values of the reported overall vaccine efficacy for AZD1222, 10 and in line with values that may be observed for seasonal influenza vaccines. 15 In the phase 3 trial for AZD1222, 62.1% -90% efficacy against symptomatic COVID-19 was observed (with the range depending on the dose regimen received), and 3.8% -58.9% efficacy against asymptomatic infection. Even the lower efficacy against asymptomatic infection (3.8%, and not significant) is likely to be consistent with some protection against infection, since the vaccine will have converted some episodes of symptomatic disease to asymptomatic infection, and hence must additionally prevent asymptomatic infection in order to avoid negative efficacy against asymptomatic infection. Efficacy of 70% against infection and an additional 70% against disease is consistent with reported efficacy of 91% against symptomatic disease, 56% against asymptomatic disease, and asymptomatic infection accounting for 85% of all infections. a Vaccine effectiveness in sensitivity analysis: AZD1222 (AstraZeneca) vaccine worst-case scenario 13% against infection 56% additional protection against disease (62% overall disease protection) See above; based on the lower-bound values of the reported overall vaccine efficacy for AZD1222, 10 . 15 Efficacy of 13% against infection and an additional 56% against disease is consistent with reported efficacy of 62% against symptomatic disease, 4% against asymptomatic disease, and asymptomatic infection accounting for 85% of all infections.
Vaccine effectiveness in sensitivity analysis: values of BNT162b2 (Pfizer-BioNTech) vaccine 50% against infection 90% additional protection against disease (95% overall disease protection) Based on the reported overall vaccine efficacy for BNT162b2 vaccine. 6-8 Efficacy of 50% against infection and an additional 90% against disease is consistent with reported efficacy of 95% against symptomatic disease, 36% against asymptomatic disease, and asymptomatic infection accounting for 85% of all infections.
a: In our model we assume total efficacy against disease; if VE against disease is additive to VE against infection then VE should be lower. To illustrate, suppose asymptomatic infections represent a proportion A of all infections in clinical trials of a vaccine. Further, let V i and V d be vaccine efficacy against infection and disease, respectively. If someone who has been successfully immunised receives an infectious dose of SARS-CoV-2 (which would normally cause disease in the person), then there are a number of possibilities: 1. The infection and disease could be both completely prevented: this occurs with probability V i . 2. The disease could be prevented, but the person still becomes asymptomatically infected: this occurs with probability (1-V i ) * V d . 3. The vaccine could fail to prevent disease: this occurs with probability (1-V i ) * (1-V d ) Suppose a clinical trial reports the following efficacy figures: -Efficacy against symptomatic COVID-19 = T s -Efficacy against asymptomatic SARS-CoV-2 infection = T a Then we have T s = V i + (1-V i ) * V d and T a = A * V i -(1-A) * (1-V i ) * V d . If A = 85%, V i = 70% and V d = 70%, then we get T s = 91% and T a = 56%.
Furthermore, we accept that not much is known about the duration of natural immunity to SARS-CoV-2, and there is substantial uncertainty around our base case assumption of 45 weeks. However, a shorter duration of immunity is consistent with current evidence including: (i) short-lived humoral 16 and cellmediated 17 immunological memory, (ii) studies of other coronaviruses suggesting immunity lasting around 1 year, 18 (iii) evidence of short-lived protection emerging from places like Manaus, Brazil, where ~75% of its population had been estimated to be infected but which continues to see a worsening of the local outbreak (as reported in the media in January 2021), (iv) documented reinfections in cohorts such as healthcare workers, 19 (v) evidence of immune escape from new variants which may continue to emerge in the future. All in all immunity is still poorly understood and requires further research, which is why we have considered a shorter value of 45 weeks in the base case analysis that we contrasted with a wider range of values in a sensitivity analysis.
In line with the provisional vaccination prioritisation strategy in the UK, 4,20 our analysis started vaccinating individuals equivalent to the proportion of care home residents (an assumed 291,000 individuals aged 65+ years), 21 health and social care workers (1.52 million social care workers and 1.31 million NHS workers), 22,23 individuals aged 75+ years, 70+ years and those at high-risk but under 65 years (for which we obtained age-stratified data from Clark et al., 2020), 24 65+ years, those at moderate risk under 65 years, 24 those aged 60+ years, 55+ years, 50+ years, and then everyone else remaining and aged between 15 and 49 years. In sensitivity analysis we also explored vaccination at different ages, and vaccinating uniformly across all age groups.
We have assumed an initial uptake of 75% in line with recent cross-sectional survey research on the likely uptake of a COVID-19 vaccine in a UK sample, 25 and with re-vaccination coverage of 50% in individuals aged 15-64 years and 75% in adults aged 65+ years (in line with the observed seasonal influenza uptake). In both vaccination scenarios we assumed an initial vaccination rate of 100,000 individuals a day, which was informed by the initially limited supply of 800,000 doses for the vaccination rollout on 08 December of the vaccine authorised first in the UK. 26 After the first prioritisation groups have been vaccinated, we assumed a higher vaccination rate of 200,000 individuals a day based on the total stock of doses procured in contracts of the UK government that suffices to vaccinate each resident of the UK thrice; 5 a total number of three vaccines being authorised so far in the UK in early 2021 with more expected; 5 the total number of 1.18 million GP appointments on a weekday; 27 following the recent amendment of the regulation allowing the administration of vaccines under supervision by non-registered healthcare professionals during a pandemic such as with COVID-19 (regulation 247A); 28 and the observed vaccination rate of 181,300 doses a day based on administering 9,790,576 vaccines over 54 days since start of the rollout on 31 January 2021. 1 Higher and lower vaccination rates were explored in sensitivity analysis. For re-vaccinations, we assumed a lower rate of up to 100,000 individuals daily (which is not reached in the base case with assumed coverage levels of 50% in individuals aged 15-64 years and 75% in older individuals aged 65+ years).
We considered the health burden of COVID-19 and related interventions in terms of symptomatic cases, non-fatal hospitalisations, intensive-care unit (ICU) survivors, adverse-events following immunisation (AEFI), and premature fatalities due to COVID-19. For the healthcare costs, we considered visits to general practitioners, remote helpline calls, hospitalisations (ICU and non-ICU), enhanced personal protective equipment, AEFI, vaccine administrations, and the vaccine costs. We assumed twice the QALY loss and costs for AEFI, and twice the costs of vaccine administration and vaccine dose to reflect a 2-dose vaccine. 5 Key epidemiological illness parameters were taken from CovidM (Supplementary Table 3), 29 which in turn informed these values by published estimates. For our study, we extended the model to incorporate demography in terms of births and (disease-unrelated) deaths to replenish susceptible individuals, assuming a death rate identical to the birth rate based on 713,000 live births in the UK in 2019, 30 which allows exploring a longer timeframe over ten years (2020)(2021)(2022)(2023)(2024)(2025)(2026)(2027)(2028)(2029). We also updated the proportion of inpatients who were admitted to critical care (0.17) and who died in critical care (0.32) based on a large study from the UK of more than 20,000 inpatients. 31 Similarly, we used the age distribution of hospitalisations to inform the age-dependent proportion of hospital admissions in the UK. 31 Supplementary In addition, we informed mortality in individuals aged <75 years per 5-year age-band using ensemble infection-fatality rates based on data on COVID-19 confirmed deaths from 45 countries and 22 seroprevalence studies, 33 while infection-fatality rates in individuals aged ≥75 years were based on the REACT3 study from the UK 34 (Supplementary Table 4). Our results are slightly different from earlier analyses 29 due to using different input parameters for mortality, and the model including waning and demography. Our analysis assumed that input values can be extrapolated for the whole of the UK.

QALY loss input data
Estimates of health-related quality of life associated with having (long-term) COVID-19 that used the preferred instrument in England (the EQ-5D) are still scarce currently. Consequently, the QALY losses were largely informed from previously published values of other respiratory infections (Supplementary  Table 5), which may underestimate the health gain from preventing persistent symptoms of COVID-19 lasting several months. 35 Once more information about COVID-19 become available these estimates could be updated later, or indeed further explored more conceptually.
QALYs lost per symptomatic case were based on ILI for 2009 H1N1 pandemic influenza in the UK, 36 while QALY loss per non-fatal hospitalisation were based on participants discharged from a large University hospital in the UK (assuming an impact for the hospital stay and maximum duration postdischarge of 71 days). 37 The resulting QALY loss per hospitalisation due to COVID-19 was estimated to be similar to seasonal influenza (0.0201 vs 0.018, respectively). 38 QALYs lost per non-fatal ICU stay were based on decrements in quality of life of two studies in ICU survivors from the UK. 39,40 The difference in utility over one year was 0.10 in the study of Griffiths et al. QALYs lost from post-acute (long) COVID symptoms were based on the relative ratio of disability weights for post-acute consequences to moderate community cases (0.219/0.051=4.29), 41 which we multiplied with the assumed QALYs lost per symptomatic case (4.29*0.008=0.034 QALYs), 36 and assuming a proportion of 10% of cases experiencing long COVID symptoms. 42 This is a conservative estimate on the impact of the quality-of-life based on the assumed QALYs lost per symptomatic case, and more research is needed on the long-term impact of long COVID. A higher QALY value would result in more favourable results by preventing larger QALY losses.
QALYs lost from adverse events following immunisation (AEFI) were assumed with 1 QALD at a chance of 10%, which is roughly following another study on influenza vaccination. 43 We did not consider longer-term or serious AEFI that may occur but are unknown yet.
QALYs lost per death were based on the most recent life expectancy in the UK as 3-year average over 2017-2019, 44 and adjusted for age-and sex-specific QALY population norms based on the EQ-5D-3L for the UK. 45 We also adjusted for the higher prevalence of comorbidities in individuals most likely to die from COVID-19 if infected, 46 by reducing the QALY norms by an assumed 10% and accounting for an assumed 25% increased risk of non-COVID-19 mortality in these individuals.
To explore parameter uncertainty, we ran the epidemiological model deterministically with R0 values of 2.7 (the base case), 1.6, and 3.9. 29 The economic model obtained 1,000 iterations using Monte Carlo sampling of the input costs and QALYs to obtain a probability distribution of outcomes. We used beta distributions for the utilities and a normal distribution for the estimated QALYs lost due to premature mortality. We explored the uncertainty using values that were ±25% of the mean value (±10% for the QALYs lost due to premature mortality) to make as few assumptions as possible on the data and variance. For the costs per AEFI we followed the assumption made by others to use the costs of 1 GP visit per AE. 48,55 again assuming a chance of 10%. 43 No costs for longer-term/serious AEFI were included.

Supplementary
The costs of vaccine administration were based on the agreed service payment for COVID-19 vaccination of £12.58 per dose (and £25.16 for two doses), 56 which is 25% higher than the current service payment for influenza vaccines of £10.06. 57 Although assumed to be administered via GPs, we did not account for the costs of extraordinary GP visits (i.e., we implicitly assume the vaccine to be administered as part of another visit; this assumption may be challenged given the current advise of physical distancing, and increasing the costs per vaccine was shown to be somewhat sensitive for the worst-case vaccination scenario depending on the physical distancing scenario in sensitivity analysis). We also assumed administration costs to be the same for all vaccines, which is in line with guidance of NHS England as of mid-January 2021; 58 if one was to assume higher or lower administration costs these could also be added to the price per vaccinated individual (see Figure 3, and Supplementary Table 8).
The cost per vaccine dose were conservatively assumed to be £15 to match reports of the first authorised vaccine in the UK that the government has an agreement with for 40 million doses. We varied the vaccine costs in sensitivity analyses at £0-£50, which reflects the range of prices reported for the vaccines that the UK has signed agreements for (including the reported £6 for 2 doses of the AstraZeneza-Oxford University vaccine that the UK has an agreement with for 100 million doses, and the reported £50 for 2 doses of the Moderna vaccine that the UK has ordered 5 million doses; all of these estimates are only indicative at this time as they are based on wholesale prices negotiated in volume-supply contracts between manufacturers and governments). 5 In addition, the vaccination scenarios included the public expenditures on subsidising the development of SARS-CoV-2 vaccines with £250 million by the UK government, 59 which could be regarded as an extraordinary lump-sum ex-ante premium. We also added the costs of a public tender for £3.3 million of ultra-low temperature freezers to the best-case scenario that was informed by the first authorised vaccine. 5 Although these costs could be considered as sunk costs and excluded from the analysis, we have included these costs given the counterfactual no-vaccination scenario to enable retrospective assessment of the value of these investments. Also, the impact of these cost factors on the overall cost-effectiveness results is negligible.
Additional cost factors could have been considered, including for instance an expanded testing programme or the running costs of the temporary field hospitals (estimated with approximately £15 million for the seven NHS Nightingale hospitals in England in April 2020). 60 While cost savings may be realisable on the field hospitals, expanded testing for SARS-CoV-2 may become a fixture in the years ahead for surveillance purposes irrespective of disease activity, and may thus be regarded as a fixed sunkcost for economic analyses.

Sensitivity analyses
In a first sensitivity analysis, we explored an assumed vaccination effectiveness against disease and against infection according to the lower and upper bounds of the vaccine efficacy reported in the AstraZeneca-Oxford University phase 3 clinical trial (Supplementary Figure 1), and one according to the vaccine efficacy against disease of the Pfizer-BioNTech phase 3 clinical trial supplemented with reasonable assumptions for the vaccine effectiveness against infection (Supplementary Figure 2). Strategy B is similar to the base case analysis, while in strategy C transmission is ongoing over the ten years ( Supplementary Figure 1-2).
In a second sensitivity analysis, we explored the epidemiological impact of vaccinating the targeted population uniformly. Compared to the targeted vaccination strategy using prioritisation groups (shown in the main text; Figure 1), vaccinating uniformly leads to similar disease dynamics in terms of epidemics but higher case numbers (Supplementary Figure 3).
In a third sensitivity analysis, we explored different uniform discount rates for benefits and costs (Supplementary Figure 4).
In a fourth sensitivity analysis, we explored the change in the net monetary value of vaccination if vaccine introduction is delayed up until the start of 2022. The economic value of introducing vaccination was estimated to be lower if the vaccine is introduced in December 2020 compared to an earlier introduction (Supplementary Figure 5). The incremental net monetary value of introducing vaccination does not always decrease as the delay increases, due to the interaction with seasonal cycles in incidence, holidays, and periods of physical distancing. In particular, vaccine introduction before an epidemic peak decreases the case numbers and height of said peak and leads to higher net values of vaccination, while vaccine introduction at the height of a peak or even afterwards will lead to lower net values of vaccination due to not having been able to prevent cases and the associated disease burden. This sensitivity analysis illustrates the (counterfactual) economic costs of delaying vaccination with a safe and effective vaccine, and assuming a vaccine has been authorised for supply by competent regulatory authorities with demonstrated effectiveness and safety from completed phase 3 trials.
In a fifth sensitivity analysis, we explored the incremental impact on the net monetary value in terms of efficiency per vaccinated individual in the targeted age groups using prioritisation groups or vaccinating uniformly (Supplementary Figure 6). The efficiency of a targeted programme in terms of incremental net monetary value per vaccinated individual can be increased by moving from older age groups to younger age groups (all of which start vaccinating the oldest ages first according to the JCVI prioritisation advice). For a uniform programme, the efficiency in terms of incremental net monetary value per vaccinated individual is largest when vaccinating the oldest age groups. These findings mostly align for the worstcase and best-case scenarios, although a disease-preventing vaccine may lead to negative incremental net values compared to no-vaccination.
In a sixth sensitivity analyses, we varied the re-vaccination coverage level in ages 15-64 years, with the net monetary value increasing with higher coverage levels for the best-case vaccination scenario but may become negative for the worst-case vaccination scenario (Supplementary Figure 7).
In a seventh sensitivity analysis, we further explored a different rate of vaccination uptake in the UK after the first prioritisation groups and before annual revaccinations. The results show that the speed of the initial vaccination rate has only a minor impact on results (Supplementary Figure 8).
In an eighth sensitivity analysis, we limited the importation rate of new infections to the first 6 months, which would in theory allow elimination. Results are not different from the base case as the locally transmitted cases rapidly exceed the low number of imported cases (Supplementary Figure 9).

Supplementary Figure 1: Epidemiological impact of vaccination effectiveness (AstraZeneca-Oxford
University vaccine). Sensitivity analysis on the vaccination effectiveness according to the lower and upper bounds of the vaccine efficacy reported in the AstraZeneca-Oxford University phase 3 clinical trial. Days highlighted indicate summer and winter holidays (in light-blue), periods of physical distancing (light-red), or neither of the two (white). Note: The y-axis is truncated at 100,000 cases daily to allow meaningful visual comparisons across panels.

Supplementary Figure 2: Epidemiological impact of vaccination effectiveness (Pfizer-BioNTech vaccine)
. Sensitivity analysis on the vaccination effectiveness according to the vaccine efficacy against disease of the Pfizer-BioNTech phase 3 clinical trial supplemented with reasonable assumptions for the vaccine effectiveness against infection. Days highlighted indicate summer and winter holidays (in lightblue), periods of physical distancing (light-red), or neither of the two (white). Note: The y-axis is truncated at 100,000 cases daily to allow meaningful visual comparisons across panels. Figure 3: Vaccinating uniformly. Sensitivity analysis on the vaccination programme when vaccinating uniformly (instead of targeted vaccination using prioritisation groups as shown in the main text). Days highlighted indicate summer and winter holidays (in light-blue), periods of physical distancing (light-red), or neither of the two (white). Note: The y-axis is truncated at 100,000 cases daily to allow meaningful visual comparisons across panels. Supplementary Figure 6: Age group of vaccination and target groups. Sensitivity analysis on the age groups targeted in the vaccination programme scenarios B and C versus the no-vaccination scenario A using prioritisation groups (a-f) or vaccinating uniformly (g-l). Figure 8: Vaccination uptake rate. Sensitivity analysis on the vaccination uptake rate after the initial phase and before starting annual re-vaccinations. Figure 9: Epidemiological impact of imported cases. Sensitivity analysis on the number of imported cases each month (limited to the initial six months). Days highlighted indicate summer and winter holidays (in light-blue), periods of physical distancing (light-red), or neither of the two (white). Note: The y-axis is truncated at 100,000 cases daily to allow meaningful visual comparisons across panels.

Background and objectives
3 Provide an explicit statement of the broader context for the study.