key: cord-0814326-kuyzraqm
authors: Yang, J.; Marziano, V.; Deng, X.; Guzzetta, G.; Zhang, J.; Trentini, F.; Cai, J.; Poletti, P.; Zheng, W.; Wang, W.; Wu, Q.; Zhao, Z.; Dong, K.; Zhong, G.; Viboud, C.; Merler, S.; Ajelli, M.; Yu, H.
title: Can a COVID-19 vaccination program guarantee the return to a pre-pandemic lifestyle?
date: 2021-02-05
journal: medRxiv : the preprint server for health sciences
DOI: 10.1101/2021.02.03.21251108
sha: 960e698b2764e0b1d4c539e9d39c7195068cebe7
doc_id: 814326
cord_uid: kuyzraqm

COVID-19 vaccination has been initiated in several countries to control SARS-CoV-2 transmission. Whether and when non-pharmaceutical interventions (NPIs) can be lifted as vaccination builds up remains key questions. To address them, we built a data-driven SARS-CoV-2 transmission model for China. We estimated that, to prevent local outbreaks to escalate to major widespread epidemics, stringent NPIs need to remain in place at least one year after the start of vaccination. Should NPIs be capable to keep the reproduction number (Rt) around 1.3, vaccination could reduce up to 99% of COVID-19 burden and bring Rt below the epidemic threshold in 9 months. Maintaining strict NPIs throughout 2021 is of paramount importance to reduce COVID-19 burden while vaccines are distributed, especially in large populations with little natural immunity.

The novel coronavirus disease 2019 (COVID-19) pandemic is far from over with cases still surging in many countries across the globe (1). In 2020, epidemic suppression and/or mitigation have relied on non-pharmaceutical interventions (NPIs), including social distancing, school closure, masking, and case isolation. Although effective and widely adopted to limit SARS-CoV-2 transmission and reduce COVID-19 burden, these interventions entail enormous economic costs and negatively affect quality of life (2) . Additionally, in many countries, relaxation of NPIs has led to a resurgence of the epidemic as no location has reached herd immunity thus far (3) -even in Manaus, Brazil where it is estimated that over >70% of the population has been naturally infected, the epidemic is seemingly not over (4) .

Effective vaccines against COVID-19 remain the only foreseeable means of both containing the infection and returning to pre-pandemic social and economic activity patterns. Globally, several vaccines have been licensed, and vaccination programs have been initiated in several countries including China (5). However, in the near future, the projected global production and delivery capacities are likely to be inadequate to provide COVID-19 vaccines to all individuals who are still susceptible to SARS-CoV-2 infection (3) . The effectiveness of COVID-19 vaccination campaigns will depend on several factors, including vaccine supply, willingness to receive the vaccine, and strategies for vaccine allocation and deployment (6) . In particular, estimating whether and when NPIs can be lifted while vaccination campaigns are ongoing is a top priority for policy making. Moreover, optimal strategies for vaccine allocation in a shifting landscape of infections are urgently needed as well.

In this study, we aim to address these questions by using China as a case study. To do so, we build an age-structured stochastic model to simulate SARS-CoV-2 transmission in mainland China, based on a susceptible-infectious-removed (SIR) scheme (Fig. S1 ). We account for heterogeneous mixing patterns by age (7) and progressive vaccine deployment among different population segments based on a broadly accepted priority scheme (essential workers, older adults and individuals with underlying conditions, etc.). Further, we overlay a disease burden model on the transmission model to estimate the number of symptomatic cases, hospitalizations, ICU admissions, and deaths under different vaccination scenarios and based on empirical data (8) (9) (10) (11) (12) (13) . The resulting model is informed by data on COVID-19 natural history, age-mixing patterns specific to China in the pandemic period, and the size of the different vaccination targets in the Chinese population (e.g., individuals with pre-existing conditions). We also leverage data on the Chinese healthcare system to estimate vaccine administration capacity. A summary of model parameters and data sources is presented in Table 1 . Model details are described in Supplementary Materials 1.

We considered a baseline reactive vaccination scenario where: 1) vaccination starts 15 days after an outbreak triggered by 40 breakthrough imported COVID-19 infections;

2) vaccine efficacy (VE) against SARS-CoV-2 infections for a two-dose schedule is set at 80%; 3) vaccination coverage is capped at 70%; 4) 6 million doses are administered daily (4 per 1,000 individuals; twice the capacity estimated for the 2009 H1N1 influenza pandemic vaccination campaign ); 5) the first priority target consists of older adults and individuals with underlying conditions (descriptions in details shown in Table S1 ); 6) there is no prior population immunity from natural infection, which aligns with the situation in most of China where there has been little circulation of SARS-CoV-2 in 2020 (3); 7) we assume an initial effective reproductive number R t =2.5 homogeneous across age groups at the start of the outbreak, in the absence of NPI and vaccination; and 8) we let the model run for two years.

In the absence of NPIs, the vaccination program is too slow to lower and delay the epidemic (Fig. 1A) and does not effectively reduce COVID-19 burden. R t falls below the epidemic threshold (<1) 69 days after the epidemic start (Fig. 1B) , but this is primarily attributable to immunity gained through natural infection rather than vaccination. Indeed, in this time frame, 52.2% of population gets infected, while only 6.7% of population has been vaccinated (Fig. 1C) . The cumulative disease burden of All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (Fig. 2) compared to a situation with moderate NPI alone, and R t falls below the epidemic threshold about 9 months after the epidemic start ( Fig. 1 ). At the time that R t falls below 1, we estimate that 50.8% of the total population would have been vaccinated, while 0.8% would have been naturally infected ( Fig. 1G-I) . This scenario also suggests that NPI should be maintained for one year after the onset of vaccination.

For instance, if NPIs are relaxed 9 months into the vaccination program, allowing a 25% increase in SARS-CoV-2 transmissibility, the cumulative death toll could increase by three folds from 76,700 to 318,300. In contrast, there is a small increase in cumulative deaths to 93,500 if NPIs are relaxed one year after vaccination ( Fig. S2-S3 ). Earlier or more drastic relaxations of NPIs lead to substantial increases in deaths (Fig S2-S3) .

A combination of more stringent NPIs (i.e., capable of keeping R t =1.1) and vaccination (vax + high NPIs scenario) could suppress the epidemic, with <2,300 symptomatic cases, and <50 deaths on average. Although the majority of the reduction of COVID-19 burden is ascribable to NPIs in this case (over 85%), the deaths averted due to vaccination are about 1.2 million (Fig. 1J -L, and Fig. 2 ).

If we consider a set of mild NPIs (vax + mild NPIs scenario), even a relatively low All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted February 5, 2021. ; https://doi.org/10.1101/2021.02.03.21251108 doi: medRxiv preprint initial reproduction number under NPIs of R t =1.5 could still lead to a disastrous epidemic, with nearly two million deaths. Despite the high death toll of the resulting epidemic, NPIs and vaccination would jointly reduce around 80% of the disease burden compared to a non-NPI non-vaccination scenario (namely, 239 million symptomatic cases and 8.2 million deaths averted) ( Fig. 1D-F, and Fig. 2 ).

Should the daily vaccination rollout be limited to 1. All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted February 5, 2021. ; https://doi.org/10.1101/2021.02.03.21251108 doi: medRxiv preprint

We consider alternative vaccination scenarios that prioritize essential workers (staff in the healthcare, law enforcement, security, community services, and individuals employed in cold chain, etc.) to maintain essential services and then explore different prioritization strategies for the rest of the population. Our results suggest that the relative timing of the epidemic and of the vaccination rollout play a key role in determining the most effective strategy. In particular if we consider vaccination to start at about the same time as an outbreak (ie, two weeks after 40 cases are detectedas in the other analyses presented in the main text), there is no clear prioritization strategy that minimize deaths, as the outcome of the vaccination campaign heavily depends on the timing at which the epidemic unfolds ( Fig. 4 and Fig. S11 -S12).

Instead, if the epidemic is already underway when the vaccination campaign starts (>5,000 cases), prioritizing working-age groups minimizes the number of deaths when R t ≤ 1.3. In contrast, prioritizing older adults and individuals with underlying conditions is more effective when R t ≥ 1.5 (direct benefits are higher, Fig. 4 and Fig.   S11 -S12). Two results are independent of the adopted prioritization strategy: i) if R t ≥ 1.5, then an epidemic cannot be avoided; and ii) when R t =1.1, over 99% of deaths can be averted (Fig. S11-S12 ).

To evaluate the impact of baseline assumptions on our results, we conduct comprehensive sensitivity analyses (SE) for R t fixed to 1.3 (moderate NPIs). Provided Other factors such as vaccine coverage (SE8 and SE9), excluding detected All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted February 5, 2021. ; https://doi.org/10.1101/2021.02.03.21251108 doi: medRxiv preprint symptomatic cases from vaccination (SE10 and SE11), the time interval between two doses (SE12), and assuming an all-or-nothing vaccine (SE19), do not substantially affect estimates of deaths (Fig. 5 ) and symptomatic infections (Fig. S14-S15) . A similar trend is observed for hospitalized cases and ICU admissions (Fig. S16-S19 ).

Using a stochastic dynamic model of SARS-CoV-2 transmission in combination with a COVID-19 burden model, we estimate the impact of a COVID-19 vaccination program in the absence or presence of NPIs on SARS-CoV-2 infections, symptomatic cases, hospitalizations, ICU admissions, and deaths in China. We find than in the absence of NPIs, and independently of the vaccine prioritization strategy and capacity of the vaccination campaign, timely rollout of an effective vaccine (VE =80%) would not be enough to prevent a local outbreak to escalate to a major widespread epidemic.

Provided that NPIs are in place and capable to bring R t to 1.3, a daily vaccine rollout of 4 doses per 1,000 individuals could reduce around 99% of COVID-19 burden, and bring R t below the epidemic threshold about 9 months after the start of the vaccination campaign. A relaxation of NPIs that bring the value of R t to 1.5 could not prevent sustained epidemic growth which would cause 1.8 million deaths. A net reproduction number of 1.5 could only be sustained when accompanied by an improvement of the vaccine administration capacity up to 10 doses per 1,000 individuals per day. Relaxation of NPIs in the first 6-9 months of vaccine roll out could lead to substantial increases of COVID-19 burden.

A previous study has qualitatively discussed the segments of the population to be prioritized in a COVID-19 vaccination program in China on the basis of ethics as well as utilitarian and egalitarian principles (15) . Our study quantitatively compares the health impacts of potential alternative vaccination strategies by varying priority ranking for target populations. Should mild/no NPIs be in place (R t ≥ 1.3), we estimate that a strategy that maintains essential workers and then prioritizes older adults and individuals with underlying conditions could be most effective at preventing deaths if All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. widespread, several rounds of lockdowns have already been required, and natural immunity is building up, China has been able to suppress SARS-CoV-2 transmission for most of 2020. As a result, prior immunity is very low, thus calling for specifically tailored analysis.

We rely on the integration of more realistic data specific for China. We account for enhanced vaccine administration capacity (increase from 1.4 million doses/day in routine vaccination to 6 million doses in COVID-19 vaccination campaigns). Our analysis is informed by the ongoing COVID-19 vaccination program in Beijing (17), estimates of vaccine supply till 2021 (18) , vaccination schedule, vaccine efficacy, and key segments of the population. Our finding confirms that if NPIs can maintain transmission rates at low levels during the vaccination campaign, strategies that target indirect benefits do better. If transmission rates remain high, strategies maximizing direct benefits will perform best. Given that China is doing so well in clamping down transmission by enforcing strict NPIs, vaccinating working age adults may generally be a better option. In most other countries, however, vaccinating older adults would be expected to save more lives (16) . Moreover, our study sheds light on the extent to All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted February 5, 2021. ; https://doi.org/10.1101/2021.02.03.21251108 doi: medRxiv preprint which NPIs can be relaxed while the vaccination campaign is underway, and on the timing when it becomes safe to do so.

Our study shows that in the absence of NPIs to slow down the epidemic, an outbreak triggered by 40 breakthrough COVID-19 cases in China could grow up rapidly to an epidemic with half of all populations infected within less than 3 months, even if a vaccination program could administer 6 million doses each day. Provided that R t is kept at 1.3 by NPIs, a vaccination program with ≥ 6 million daily doses administered could have substantial benefits, lowering the cumulative death toll of a COVID-19 epidemic over a 2 year period below that of an annual seasonal influenza epidemic (88,100) (14) . This highlights that, although in the long-term vaccination can ultimately lead to the suppression of COVID-19, it is necessary to maintain the NPIs currently in place, such as social distancing, contact tracing and quarantine measures, for at least the duration of the vaccine rollout in 2021.

As highlighted in vaccination studies in the UK and Australia (19) (20) (21) , in the race between the vaccination campaign to build population herd-immunity and the progress of the epidemic, the speed of vaccine deployment is critical. In the routine National Immunization Program, an average of 1.4 million doses are administered in China per day (22, 23) , while during the 2009 influenza pandemic a maximum of 3 million daily doses were administered (24) . Considering that the willingness to be vaccinated against COVID-19 is higher than that for the 2009 influenza pandemic (25) , and that the vaccine distribution capacity is likely to be improved as well (e.g., (17)), we consider the capacity of COVID-19 vaccination services could be scaled up to 6 million doses administered per day in the baseline analysis. Several manufacturers state that a total of 2.1 billion doses of COVID-19 vaccine could be produced in 2021, equivalent to about 6 million doses per day, which could be enough to cover 75% of the Chinese population (18) . Even if these candidate vaccines could be licensed and manufactured smoothly, it would take about one year to vaccinate 70% of the general All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted February 5, 2021. ; https://doi.org/10.1101/2021.02.03.21251108 doi: medRxiv preprint population, based on an optimistic vaccine delivery rate that is two-folds the maximum rate at which H1N1pdm vaccines were delivered in 2009 (24) .

In addition, limited vaccine production capacity, particularly at the initial stage, could slow the speed of vaccine rollout. In December 2020, Chinese media reported that the government planned to administer 100 million doses for emergency use by (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted February 5, 2021. ; https://doi.org/10.1101/2021.02.03.21251108 doi: medRxiv preprint susceptible individuals should be explored through further cost-effectiveness studies.

Our study has a number of limitations. First, we integrated the impact of NPIs through a simple reduction in the value of R t at the beginning of the outbreak, homogeneously across age groups. However, our analysis does not suggest which combination of NPIs should be adopted to lower R t to a certain level, and how this would affect transmission rates in different age groups. Li, et al, estimated that individual NPIs, including school closure, workplace closure, and public events bans, were associated with reductions in R t of 13-24% on day 28 after their introduction (28) . Further studies are needed to pinpoint the specific NPIs to be adopted in parallel with the vaccination campaign. Second, our study uses a static allocation strategy, which means a constant coverage is assumed for all subgroups, and vaccination starts from one group and remains in the group until target coverage achieved, and then moves to the next group in sequence. Such allocation process may not generate the maximum health benefit for a vaccination program (29) . Further studies could be designed to identify the optimal dynamic allocation strategy (e.g., different coverage required in subgroups, where vaccination could start from one group, move to the next and then return) to minimize COVID-19 burden, especially in the context of limited doses.

Third, in China, vaccines have not been licensed for older adults and children, so we assume a 50% lower or equivalent VE for them compared to other adults. Although we show that variations in these rates do not substantially affect the overall effect of the vaccination campaign, further data on age-specific vaccine efficacy could help refine priority groups. Fourth, we assumed that immunity after natural infections lasts more than the time horizon considered (two years). If this is not the case, waning of immunity would inflate the rate of susceptible individuals and thus require booster vaccinations. This could become an issue with the emergence of immune-escape variants, as reported in South Africa (30) . Given limited information at this stage, we did not consider this scenario in our analyses, but this is an important area of future research.

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted February 5, 2021. ; https://doi.org/10.1101/2021.02.03.21251108 doi: medRxiv preprint

Our study proposes a general framework to evaluate the impact of COVID-19 vaccination programs in the absence/presence of NPIs and to explore priority target populations to minimize multiple disease outcomes. The proposed modeling framework is easily adaptable to other country-specific contexts, including the susceptibility of the local population (3), local risk of transmission and implemented NPIs (31), efficacy of different vaccines (32) (33) (34) (35) , vaccine supply and capacity of immunization services (6) , and the objectives of the pandemic responses.

Vaccination alone could substantially reduce COVID-19 burden, but in the foreseeable future may not be enough to prevent local outbreaks to escalate to major widespread epidemics due to limitation in the vaccine production and supply (particularly at the initial stage of the vaccination), as well as the capacity of vaccination system. This is especially relevant in contexts where most of the population is still susceptible to SARS-CoV-2 infection, as it is the case in most of China. Maintaining NPIs such as social distancing, case isolation, and careful contact tracing, wearing masks, increased teleworking and limitation on large gatherings, is necessary to prevent the resurgence of COVID-19 epidemics until a sufficiently high vaccine coverage is reached.

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted February 5, 2021. ; https://doi.org/10.1101/2021.02.03.21251108 doi: medRxiv preprint Proportion of deaths averted compared to the reference scenario, i.e., no vaccination All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted February 5, 2021. ; https://doi.org/10.1101/2021.02.03.21251108 doi: medRxiv preprint + no NPIs with R t =2.5 at the beginning of transmission. Number denotes median, and error bars denote quantiles 0.025 and 0.975. Initial cases denote breakthrough COVID-19 cases, which initiates the epidemic. We consider the impact of uncertainty in contact patterns and relative susceptibility on prioritization, and use their mean values as well. Baseline denotes first prioritizing older adults and individuals with underlying conditions. Number in the box denotes the death toll (median), with t representing thousand and m representing million.

Minimum denotes the lowest deaths in each scenario on the basis of median value. We compare other strategies to that with minimum deaths using rank sum test. E.g., in the context of initial cases=5,000, R t =1.5 and using mean values of contact patterns and relative susceptibility, the baseline is the optimal strategy to minimize deaths. Number denotes median, and error bars denote quantiles 0.025 and 0.975.

Serological evidence of human infection with SARS-CoV-2: a systematic review and meta-analysis

Three-quarters attack rate of SARS-CoV-2 in the Brazilian Amazon during a largely unmitigated epidemic

Global, regional, and national estimates of target population sizes for covid-19 vaccination: descriptive study

Patterns of human social contact and contact with animals in Shanghai

Disease burden and clinical severity of the first pandemic wave of COVID-19 in Wuhan

Clinical characteristics and outcomes of critically ill patients with novel coronavirus infectious disease (COVID-19) in China: a retrospective multicenter study

Evolving epidemiology and transmission dynamics of coronavirus disease 2019 outside Hubei province, China: a descriptive and modelling study

Case fatality risk of the first pandemic wave of novel coronavirus disease 2019 (COVID-19) in China

Clinical Characteristics of Coronavirus Disease 2019 in China

Influenza-associated excess respiratory mortality in China, 2010-15: a population-based study

Who should be Prioritized for COVID-19 Vaccination in China? A Descriptive Study

Model-informed COVID-19 vaccine prioritization strategies by age and serostatus

The Central People's Government of the People's Republic of China

Reported coverage of vaccines in the National Immunization Program of China

Surveillance of Category Vaccines in China

The Central People's Government of the People's Republic of China

Acceptance of COVID-19 Vaccination during the COVID-19 Pandemic in China

Sinopharm

COVID vaccine excitement builds as Moderna reports third positive result

Safety and efficacy of the ChAdOx1 nCoV-19 vaccine (AZD1222) against SARS-CoV-2: an interim analysis of four randomised controlled trials in Brazil, South Africa, and the UK

Safety and Efficacy of the BNT162b2 mRNA Covid-19 Vaccine

Early Transmission Dynamics in Wuhan, China, of Novel Coronavirus-Infected Pneumonia

Association of Public Health Interventions With the Epidemiology of the COVID-19 Outbreak in Wuhan

Safety and immunogenicity of an inactivated SARS-CoV-2 vaccine, BBIBP-CorV: a randomised, double-blind, placebo-controlled

All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.