key: cord-1025361-b7tdd8uy
authors: Reperant, Leslie A; DME Osterhaus, Albert
title: COVID-19 vaccination and critical care capacity: perilous months ahead
date: 2021-03-10
journal: Vaccine
DOI: 10.1016/j.vaccine.2021.03.035
sha: a2d17445826447e696ed8bb62bfa18f56e1c99d9
doc_id: 1025361
cord_uid: b7tdd8uy

nan

widespread implementation and relaxation can be particularly damaging to society. COVID-19 vaccination programs were initiated by the end of 2020 or beginning of 2021 in most countries, but the risk of renewed widespread lockdowns remains. Using available data on daily COVID-19 incidence and growth rate in eleven European countries that reinstated a national lockdown in fall 2020, we show a high risk of short-term overwhelming of critical care capacity, as levels of population immunity remain below targeted herd immunity. To avoid the repeated widespread implementation and relaxation of lockdowns, we advocate that controlled conditions brought by current national lockdowns should be leveraged to eliminate SARS-CoV-2 circulation in the community.

Early mathematical models rapidly demonstrated the need for swift suppression of viral spread to curtail the burden on hospital care, largely due to precipitously overtaken regular and intensive care services. [3] [4] [5] When overwhelmed, healthcare systems face unsurmountable challenges for the provision of standards of care for COVID-19 and non-COVID-19 inpatients alike. Eventually, the unavoidable need for inpatient triage can dramatically increase the mortality burden and overall emerging pandemic impact. Improved standards of care and treatment options for COVID-19 patients have contributed to decreased case fatality rates in hospitals. 6 Likewise, nonpharmaceutical interventions have contributed to reduced transmission, 7 as measured by the effective reproduction number (R eff ), i.e., the average number of new cases caused by one infected individual. However, the pandemic potential of SARS-CoV-2 has intrinsically not diminished and continues to threaten society, including through the emergence of more transmissible strains, like B.1.1.7 8 and others, with hospitals in the front line and economies suffering dramatically. Before envisaged COVID-19 vaccination coverage achieves sufficient levels of herd immunity, critical care capacity remains at risk of becoming overwhelmed.

Considering the maximum number of daily cases of COVID-19 infections occurring during the week prior to the reinstatement of a national lockdown, as a proxy to define the incidence threshold to engage into a national lockdown, one can estimate the time needed for critical care capacity to risk becoming overwhelmed, as a function of a baseline daily incidence, R eff and corresponding doubling time (Appendix). Such approximations are justifiable when population immunity levels remain well below targeted herd immunity, as currently observed in most countries. 9, 10 While the time to the incidence threshold tends to infinity when R eff = 1.0, it rapidly decreases with small R eff increments and reaches zero when the baseline incidence equals the incidence threshold (Fig. 1a ). On average, the time to the incidence threshold reached 6 months with R eff values of 1.10 (1.07 -1.13) and 3 months with R eff values of 1.22 (1.15 -1.28) for the eleven European countries studied (Fig. 1b) .

In those European countries that had lifted their first lockdown in May, R eff averaged 1.1 (1.05 -1.12) from May to the end of October (Fig. A1 ). At R eff = 1.1, it would take an average of 6.4 (5.2 -7.7) months to reach the incidence thresholds from average daily incidence observed in May (Table 1 ). In most European countries that reimposed a lockdown in fall 2020, R eff remained above 1.0 since July with an average of 1.2 (1.17 -1.23) until the end of October. At R eff = 1.2, it would take an average of 3.1 (2.8 -3.5) months for these countries to reach the incidence thresholds from average daily incidence observed in July ( Fig. 1c and Table 1 ). Both estimates fit well with the actual timing of the reinstatement of national lockdowns in the respective countries, end of October or early November 2020.

The proposed calculations have the advantage to estimate the time to risking overwhelming critical care capacity based on R eff , independently of the intrinsic transmissibility of circulating viral strains. Time-varying R eff estimated over the course of an epidemic accounts for the impact of public health control measures, the build-up of immunity in the population, or both. Given k the proportion of immune individuals and g the relative reduction in transmission rates due to non-pharmaceutical interventions, R eff can be calculated as R eff = (1 -k)(1 -g)R 0 . 10 The resulting relationship between k and g to maintain R eff at particular values is shown in Fig. 1d .

This has important implications. The time estimates to reach the incidence threshold apply similarly to levels of population immunity conferred by previous infection or by vaccination, in the absence of control measures, when these remain below targeted herd immunity. Assuming levels of immunity to SARS-CoV-2 from previous infection and vaccination at 30% in the population, R eff would be reduced to 1.75 (based on a conservative R 0 of 2.5), risking overwhelming critical care capacity typically in less than one month, in the absence of other control measures. With 30% of the population immune, control measures would need to reduce transmission rates by 31% to maintain R eff at 1.2 and by 43% to maintain R eff at 1.0 (Fig. 1d ).

This would correspond to changes in R eff (∆R eff ) of -0.75 to -0.55. The thorough analyses of Haug et al. 7 have shown that i) no single control measures have such an impact on R eff and ii) that the most effective combined measures to significantly reduce R eff can be particularly intrusive in restricting physical contact and movement. In other words, as population immunity levels remain low, substantial limitations on gatherings and mobility will need to be maintained upon SARS-CoV-2 circulation to prevent overwhelming of critical care capacity.

As immunity builds up in the population, the main assumption supporting the proposed approximations eventually will be falsified. However, as seen above, reaching the incidence threshold risking the overwhelming of critical care capacity inevitably calls for drastic restrictive measures, such as lockdowns, to curb the disease spread, resulting in cyclic slow-down of active virus circulation as widespread lockdowns are implemented and relaxed. This results in a stepwise population immunity build-up during each cycle. Using the same approximations as previously, the cumulative number of infections at the time of reaching the incidence threshold can be estimated based on R eff (Appendix; Fig. A2 ). The higher R eff , the shorter the time to the incidence threshold and the lower the cumulative number of infections upon reinstatement of drastic control measures. This results in a particularly slow build-up of immunity in the population, e.g. of an average of about 5% when R eff = 1.1, and likely contributed to the seemingly low COVID-19 seroprevalence in most countries after the first wave of the pandemic. 9,10 Consequently, conditions for the proposed approximations are likely to be maintained over several successive lockdown implementation and relaxation cycles and thus several months, in the absence of vaccination and continued circulation of SARS-CoV-2.

The implementation of safe and effective vaccination programs is fraught with uncertainty and challenges, especially amidst the current climate of public hesitancy, misinformation and distrust towards vaccines. 11 Although analyses of recently developed COVID-19 vaccines suggest overall high efficacy, for example shortly after booster vaccinations, 12,13 the level of efficacy and duration of protection in different age and risk groups and against transmission, including upon infection with newly emerging strains, remain to be determined. Duration of protection may be uncertain for months or years. 11 In addition, extensive research will be needed to address the differential risk factors contributing to infection, morbidity and mortality, the infectious role of younger age-groups and of individuals with asymptomatic or mild infection, the impact of 'super-spreaders' and individuals with prior exposure to the virus on the spread of SARS-CoV-2 and its emerging variants, as well as the effect of vaccination on these issues. In parallel, improved surveillance at the human-animal interface will be key to inform about the risk posed by putative animal reservoirs of SARS-CoV-2, such as wildlife and farmed mink. Vaccination programs with currently licensed COVID-19 vaccines further face both logistics and strategic challenges to optimize their impact on controlling the pandemic, requiring complex calculations and well-defined sets of assumptions. 11 By the end of February 2021, COVID-19 vaccination coverage had attained levels below 20% in most countries and below 10% in many countries, including across Europe, and therefore may yet only demonstrate a modest impact on virus circulation.

Targeted vaccination of groups at risk of developing severe disease typically has been envisaged first by most governments, as these individuals make the majority of COVID-19 patients requiring intensive care. Targeted vaccination of healthcare and other essential workers is typically also a priority. Based on available data for the French population (Appendix), populations at risk and essential workers, who would receive vaccination first, may represent about 40% of the population. A vaccination coverage of 90% in these priority groups would remain below targeted herd immunity levels, even with current levels of population immunity.

Reduction of virus spread by the vaccinated at-risk population may furthermore be expected to be less than that achieved by vaccinating a younger and otherwise healthy part of the population.

Yet, since the vaccinated at-risk population would be protected against severe disease upon high vaccine efficacy, such approach may nonetheless reduce the risk of overwhelming critical care capacity.

Using the proposed calculations, we found that targeted vaccination programs of at-risk groups and essential workers, assuming 90% vaccination coverage and 90% effectiveness of COVID-19 vaccines against severe disease in these priority groups (Appendix), resulted in only a 1.2-month delay in reaching the incidence threshold when R eff = 1.1 and a 0.6-month delay when R eff = 1.2, starting from July 2020 baseline incidence ( Fig. 1c and Table 1 ). Here also, at the time of reaching the incidence thresholds, the cumulative number of infections and vaccinated individuals remained below the population immunity levels necessary to bring R eff at these respective values (Fig. A3 ).

It is important to note that the time to reach the incidence threshold strongly depends on the baseline incidence. In July 2020, most European countries had relatively low daily incidence of COVID-19. Higher baseline incidence may occur upon lifting ongoing lockdowns (Appendix).

This would result in shorter times to reach the incidence threshold at corresponding R eff , as illustrated in Fig. 1c (and Table 1 ). The initial deployment of vaccination against COVID-19 therefore may not prevent the overwhelming of critical care capacity in the coming months, if R eff remains close to, yet above, the critical value of 1.0 upon lifting restriction measures.

The proposed calculations are based on simplified assumptions of the transmission dynamics of SARS-CoV-2 and have limitations. As seen above, the approximations can only be justified when levels of immunity in the population remain below targeted herd immunity and thus cannot apply when the build-up of immunity through infection or vaccination approaches or reaches such levels. We nonetheless argue that the proposed calculations serve their purpose to demonstrate the risk of overwhelming critical care capacity in the coming months while vaccination programs become initiated, in view of enduring uncertainties on achievable vaccination coverage and timelines. The calculations are dependent on relatively uncertain measures of daily incidence and time-varying effective reproduction numbers, based on incomplete data of varying accuracy. However, sensitivity analyses demonstrate the robustness of the calculations to variations in the estimated parameters (Table A1 and Fig. A4 ). The proposed calculations further assume that critical care capacity and the proportion of hospitalized patients needing intensive care remain unchanged. Critical care capacity has increased in most countries since the start of the pandemic and improved treatment options may reduce the proportion of inpatients requiring intensive care in the near future. The impact of such changes is nevertheless expected to be limited due to the different orders of magnitude between these changes and the flow of inpatients during current and potential upcoming waves.

National lockdowns are the only option left to prevent overwhelming healthcare systems in countries where the spread of COVID-19 tends to fall out of control. The WHO urges governments to avoid the use of national or widespread lockdowns as the main control strategy against COVID-19, due to the disruptive socio-economic consequences of their repeated implementation and relaxation. 2 Current mitigation strategies-aimed at controlling but not eliminating SARS-CoV-2 circulation-as adopted by most northern hemisphere countries have shown limitations in maintaining R eff close to 1.0. Based on the work of Haug et al. 7 , similarly restrictive measures on contact and mobility as applied during lockdowns may be necessary to maintain SARS-CoV-2 R eff at sufficiently low levels to prevent repeated short-term overwhelming of critical care capacity. Continuing rigorous non-pharmaceutical interventions will be necessary before vaccination programs for the population at large achieve sufficient levels of herd immunity.

Enforcing restrictive measures over a long period tend to lead to public fatigue and waning public compliance, further complicating effective control of virus circulation levels. Mitigation strategies thus will likely be ineffective as well as unsustainable and socio-economically costly to prevent the repeated overwhelming of critical care capacity in the months to come. This calls for the implementation of control measures aiming at bringing and maintaining R eff below 1.0 towards the elimination of SARS-CoV-2 community transmission, followed by prevention of COVID-19 importation and prompt stamping out of emerging clusters of infection. Such an approach has been applied successfully in countries of the eastern and south-eastern hemisphere.

These countries offer relevant blueprints for the implementation of this strategy in the northern hemisphere, towards a collaborative and coordinated approach to tackle this unprecedented crisis (Appendix). Priority vaccination of frontline workers and eventually vaccination of the population at large will further strengthen the countries' purposeful response towards maintaining zero COVID-19 community transmission. 13, 14 The results presented here did not markedly differ when using the daily effective reproduction numbers provided at the open dataset of OurWorldInData.org, sourced from the following reference. 15 Open data on age-stratified distribution of confirmed cases, hospitalized cases and intensive care unit (ICU) admissions in France were obtained from the French government data server (www.data.gouv.fr). This is one the rare open datasets providing such detailed level of information. We used estimates of the proportion of COVID-19 confirmed cases and ICU admissions below and above the age of 70, based on the data over the months of September and October. High testing rates and low positivity rates were consistently observed in September and October in France, thereby supporting good robustness of the collected data. The proportion of individuals under 70 with comorbidities was obtained from the 2012 ObEpi report 16 and numbers of essential workers for priority COVID-19 vaccination were retrieved from the following reference. 17

For each country, considering the maximum number of daily cases of COVID-19 infections occurring during the week prior to the reinstatement of a national lockdown M as a proxy to define the incidence threshold I to engage into a national lockdown, one can estimate the time T t from time t needed for healthcare capacity to risk becoming overwhelmed, as a function of a baseline daily incidence b t , the effective reproduction number R t and the corresponding doubling time D t at time t. For each country:

with f, the generation time, set to 5 days 18, 19 , r M the daily growth rate and R M the daily effective reproduction number, averaged over the week prior to the reinstatement of the national lockdown, and r t the baseline daily growth rate derived from R t .

The incidence threshold risking saturating intensive care capacity I depends not only on the daily incidence of cases but also on the growth rate of the epidemic at that time. Therefore, I was estimated based on the cumulative number of cases over n = 5 days upon reaching M, as informed by average length of stay in ICUs 20 , and corrected by the cumulative number of cases with growth rate r t over n = 5 days. M and b t were corrected for under-diagnosis as: Table  A2 . Daily T t and R t are shown in Fig. A1 .

Average estimates of T 0 , the time to reach the incidence threshold at the start of the pandemic at varying R eff are shown in Table 1 , with b t set at 1.0 (b 0 ). Average estimates of T May , the time to reach the incidence threshold from baseline incidence occurring in May at varying R eff are shown in Table 1 , with b t input as the average number of daily cases in May (b May ). Note that the data for Great Britain and Ireland were not included in the estimations of T May as these two countries maintained national lockdowns beyond the month of May 2020. Average estimates of T July , the time to reach the incidence threshold from baseline incidence occurring in July at varying R eff are shown in Fig. 1c and in Table 1 , with b t input as the average number of daily cases in July (b July ). Average estimates of T Dec , the time to reach the incidence threshold from baseline incidence occurring in December at varying R eff are shown in Fig. 1c and in Table 1 , with b t input as the average number of daily cases over the week of December 4 th to December 10 th (b Dec ).

Modelled cumulative number of cases C, over the time to reach the incidence threshold T July from average baseline incidence observed in July (b July ), was estimated based on effective reproductive number R eff as follows:

Above approximations to estimate the growth of the pandemic are justifiable as the cumulative number of cases reached so far in most countries remains largely below the number of cases required for herd immunity. 8, 9 Sensitivity analyses were performed to variations in the generation time f, in the reporting rates p 0 and p x and in the number of days n used to estimate I. The analyses demonstrate good robustness of the calculations to variations in the estimated parameters (Table A1 and Fig. A4 ). The highest impact occurred for variations in COVID-19 generation time. Here, COVID-19 generation time was assumed fixed at the mean and median values of the Weibull distribution of generation times, as found by Ferreti et al. 19

Vaccination with licensed COVID-19 vaccines in priority groups aim at reducing the health impact of COVID-19 on the most vulnerable and at preventing infection in essential workers. Desirable vaccination coverage targets in the general population may only be reached after one or few years 12 and thus will have limited impact on SARS-CoV-2 transmission in the short term. We assessed the potential impact of targeted vaccination of risk groups and essential workers on T v from average baseline incidence in July (b July ) at varying R eff . We assumed 90% vaccination coverage and 90% vaccine effectiveness against severe disease in these individuals. Based on French data on agestratified distributions of COVID-19 confirmed cases and ICU admissions over the months of September and October, we assumed individuals above 70 represented 49% of COVID-19 inpatients in ICUs and 7% of all cases. We assumed two thirds of ICU inpatients below the age of 70 had comorbidities. 21 In the general population, 23% of individuals below 70 had comorbidities 16 and 11% worked as essential workers. 17 This resulted in the vaccination of v 1 = 33% of the population and in a reduction of the number of ICU patients by v 2 = 68%. Under this scenario, and for each country, the corresponding maximum daily number of cases to be reached over the week prior to the instatement of national lockdowns, M v , was estimated as follows:

and I v calculated as previously using:

for:

We assumed a best-case scenario with a lockdown-induced reduction of R eff to 0.6 in all countries over 20 successful days using December 10 th values of daily incidence as starting values, and estimated for each country the resulting number of reported daily cases at the time of a hypothetical scenario of lockdown lifting (Table A2) . We used these estimates corrected by reporting rate p 0 as projected baseline incidence upon lifting lockdowns b Lift . We calculated T Lift as initially following:

Note that T Lift estimates assume R eff returns to values above 1.0 following lifting of the lockdowns.

Data availability: All data is available in the Appendix or can be sourced to the original data repositories.

Countries like South Korea, China, Taiwan, Singapore, Vietnam, Thailand, New Zealand and Australia pursue an elimination strategy to suppress COVID-19 community transmission, prevent importation of cases and rapidly extinguish emerging clusters of infections. These countries provide valuable blueprints for the implementation of a similar strategy in countries of the northern hemisphere. For example, New Zealand's COVID-19 alert system (available online at https://COVID-19.govt.nz/alert-system/) is well described and delineates the different steps required to achieve and maintain elimination of COVID-19 community transmission. Importantly, because of the open borders of the Schengen area, European countries need to adopt the same strategy in a collaborative and coordinated approach, to maximize the chances of success; this would likewise apply to the different provinces or states within a particular country, like Canada and the United States of America. Key aspects of effective elimination plans include the use of different alert levels that may be applied at town, city, territorial, local, regional or national scale as a function of the circulation of the virus. Intensive testing and contact tracing should be in place before lifting national lockdowns, to identify most source and downstream infections while contact and movement restrictions remain in place. Sentinel systems for other respiratory infections, such as influenza, RSV and hMPV infections, could provide robust additional insights into the risk of continued circulation of SARS-CoV-2, as contact and movement restriction measures will likewise limit their circulation. 222 Maintenance of contact and movement restrictions through short targeted local or regional lockdowns is essential where community transmission and active clusters of COVID-19 infections continue to occur, so that testing, contact tracing, and managed quarantine or isolation measures can be effectively deployed. In particular, travel from and to regions with active community transmission must be heavily restricted. This is key to preventing importation and exportation of COVID-19. Intensive testing and contact tracing will allow different geographical units to transitioning into lower alert levels at different pace, with eventual easing of travel restrictions to and from areas within the same or lower alert level. Limitations on gatherings and events must remain in place as long as there is a risk of community transmission. Common practices apply at all alert levels, including upon elimination of community transmission, such as staying at home while sick, good respiratory and hand hygiene, keeping track of one's movements and contacts, potentially with the help of digital applications, physical distancing and possibly the wearing of face coverings in public transport and crowded indoor spaces with public. Upon positive testing and identification of close contacts, self-isolation and self-quarantine is enforced at the highest alert level. However, as soon as manageable or upon reaching a lower alert level, case isolation and quarantine should be organized in managed facilities. This not only ensures better compliance but also prevention of ongoing transmission within households. Arrivals from countries with active community transmission at countries' or regions' external borders must be controlled with systematic health screening and testing programs, and mandatory quarantine in managed facilities. Similar border restriction measures should also remain in place at regions' internal borders, like that of the Schengen area in Europe, with countries not pursuing elimination of COVID-19 community transmission. Such an approach would eventually lead to the creation of COVID-19-free travel zones. In the northern hemisphere, the achievement of zero COVID-19 community transmission should be aimed at for this coming spring (2021) at the latest, to maximize its chance of success with the return of more favourable weather. Such a collaborative and coordinated strategy would further allow the re-deployment of resources, including COVID-19 vaccines, to less favoured countries and regions needing them most. Figure A2 . Estimation of the proportion of the population infected with SARS-CoV-2 at the time of reaching the incidence threshold from average baseline incidence observed in July.

Averages and standard deviations are provided over the 11 European countries. Figure A3 . Proportion of the cumulative number of infections and vaccinated individuals in priority groups at the time of reaching the incidence thresholds from average baseline incidence observed in July. Black line and markers indicate average (and standard deviation); dashed line marks the levels of population immunity necessary to bring the effective reproductive number at the respective values, based on a conservative R 0 of 2.5. Figure A4 . Sensitivity analyses. (a) Average daily effective reproduction number (R t ; and standard deviation) of the 11 European countries for monthly bins of daily T t values with reference reporting rates (p 0 = 0.5, p 1 = 0.05, p 2 = 0.1, p 3 = 0.2, p 4 = 0.3, p 5 = 0.4; black circles) and with reporting rates set at 1.0 (grey diamonds; p 0 = p 1 = p 2 = p 3 = p 4 = p 5 = 1.0). (b) Average daily R t (and standard deviation) for monthly bins of daily T t values with reference number of days to estimate the incidence threshold, n = 5 (black circles), and with n = 7 (grey diamonds) and n = 3 (white triangles). (c) Average daily R t (and standard deviation) for monthly bins of daily T t values

World Health Organization. Coronavirus disease (COVID-2019) situation reports. World Health Organisation (2020)

World Health Organization. Coronavirus disease (COVID-19): Herd immunity, lockdowns and COVID-19

Impact of non-pharmaceutical interventions (NPIs) to reduce COVID-19 mortality and healthcare demand

Projecting the transmission dynamics of SARS-CoV-2 through the postpandemic period

How will country-based mitigation measures influence the course of the COVID-19 epidemic?

Why do COVID death rates seem to be falling?

Ranking the effectiveness of worldwide COVID-19 government interventions

Transmission of SARS-CoV-2 Lineage B.1.1.7 in England: insights from linking epidemiological and genetic data. medRxiv

SARS-CoV-2 seroprevalence worldwide: a systematic review and meta-analysis

COVID-19 herd immunity: where are we?

Challenges in creating herd immunity to SARS-CoV-2 infection by mass vaccination

Covid-19: Moderna vaccine is nearly 95% effective, trial involving high risk and elderly people shows

Safety and Efficacy of the BNT162b2 mRNA Covid-19 Vaccine

A New Framework and Software to Estimate Time-Varying Reproduction Numbers During Epidemics

Different Epidemic Curves for Severe Acute Respiratory Syndrome Reveal Similar Impacts of Control Measures

Tracking R of COVID-19: A New Real-Time Estimation Using the Kalman Filter

Comité scientifique COVID-19 & Comité Vaccin COVID-19. Vaccins contre le SARS-CoV-2; 9

Estimating the generation interval for coronavirus disease (COVID-19) based on symptom onset data

Quantifying SARS-CoV-2 transmission suggests epidemic control with digital contact tracing

Time between Symptom Onset, Hospitalisation and Recovery or Death: Statistical Analysis of Belgian COVID-19 Patients

Baseline Characteristics and Outcomes of 1591 Patients Infected With SARS-CoV-2 Admitted to ICUs of the Lombardy Region

The impact of COVID-19 public health measures on detections of influenza and respiratory syncytial virus in children during the 2020 Australian winter

 Tables  Table A1 . Sensitivity of the calculations of the time to incidence threshold from average baseline incidence in July