key: cord-0841266-ay2isagn
authors: Zhang, Kevin; Vilches, Thomas N.; Tariq, Mehreen; Galvani, Alison P.; Moghadas, Seyed M.
title: The impact of mask-wearing and shelter-in-place on COVID-19 outbreaks in the United States
date: 2020-10-09
journal: Int J Infect Dis
DOI: 10.1016/j.ijid.2020.10.002
sha: b9186f78a8e6c0ce758ecc7a5570fd610073be0d
doc_id: 841266
cord_uid: ay2isagn

Objectives A hasty reopening has led to a resurgence of COVID-19 in the United States (US). We sought to quantify the impact of several public health measures, including non-medical mask-wearing, shelter-in-place, and detection of silent infections to help inform COVID-19 mitigation strategies. Methods We expanded a previously established agent-based disease transmission model and parameterized it with estimates of COVID-19 characteristics and US population demographics. We implemented non-medical mask-wearing, shelter-in-place, and case isolation as control measures and quantified their impact on reducing the attack rate and adverse clinical outcomes. Results We found that non-medical mask-wearing by 75% of the population reduced infections, hospitalizations, and deaths by 37.7% (IQR: 36.1% - 39.4%), 44.2% (IQR: 42.9% - 45.8%), and 47.2% (IQR: 45.5% - 48.7%), respectively, in the absence of a shelter-in-place strategy. Sheltering individuals aged 50 to 64 was the most efficient strategy, decreasing attack rate, hospitalizations, and deaths by over 82% when combined with mask-wearing. Under simulated scenarios, outbreak control can be achieved, bringing the attack rate to below 1%, if at least 33% of silent pre-symptomatic and asymptomatic infections are identified and isolated. Conclusions Mask-wearing, even with the use of non-medical masks with only 20% efficacy in preventing disease transmission, has a substantial impact on outbreak control. Shelter-in-place strategies remain an important public health intervention, amid ongoing outbreaks.

The novel coronavirus disease 2019 has marked the most devastating pandemic in modern times with significant morbidity and mortality (World Health Organization 2020) . The United States (US) has recorded more cases and deaths than any other countries with over one-fifth of global mortality as of October 1, 2020 (Johns Hopkins University; Weinberger et al. 2020) . Clinical and epidemiological studies suggest that comorbid individuals and persons older than 50 years of age have been disproportionately affected by COVID-19, in terms of both disease severity and adverse clinical outcomes (Centers for Disease Control and Prevention 2020a; Garg 2020a) . In addition to the catastrophic toll on human health, the restrictive public health measures implemented to combat the spread of COVID-19 have led to widespread disruption of education, societal functions, and the economy (Dorn et al.; Fernandes 2020; Palumbo and Brown 2020) . Rapid easing of these measures to alleviate social and economic woes in the US during the summer led to a resurgence of cases (Johns Hopkins University), forcing many states such as Texas, California, and Oregon to restructure their reopening strategies (CNN 2020a) .

A capacity to identify over one-third of silent pre-symptomatic and asymptomatic infections, in addition to the immediate isolation of symptomatic cases, is required to control the current resurgence of COVID-19 before a safe and effective vaccine becomes available (Moghadas et al. 2020a) . Given that this is unlikely to be achieved with the existing testing capacity and sample-to-result timeline in the US, additional public health measures, such as shelter-in-place strategies for specific population segments and mask-wearing, may be needed to change the trajectory of current outbreaks. Quantifying the effect of these measures can help to inform public health mitigation strategies to curb ongoing outbreaks.

We sought to project the impact on adverse clinical outcomes that could be achieved by maskwearing and implementing shelter-in-place strategies for various population segments, while considering age, and comorbidities associated with COVID-19 (Centers for Disease Control and Prevention 2020b; Garg 2020b; Stokes et al. 2020) . To this end, we expanded a previously established agent-based simulation model (Moghadas et al. 2020a; Shoukat et al. 2020) , and simulated various outbreak scenarios while assuming 5% population immunity for our base case based on recent seroprevalence studies (SeroTracker; Bobrovitz et al. 2020) . We show that attack rate, hospitalizations, and deaths can be reduced by at least five-fold if shelter-in-place of comorbid individuals or persons aged 50-64 years is combined with mask-wearing by those not sheltered-in-place. We found that outbreak control can be achieved, bringing the attack rate to below 1%, if at least 33% of silent pre-symptomatic and asymptomatic infections are identified and isolated.

We extended our agent-based transmission model (Moghadas et al. 2020a; Shoukat et al. 2020) to simulate post-reopening COVID-19 outbreak scenarios. The natural history of COVID-19 was implemented by the inclusion of individual classes with the epidemiological statuses as susceptible; latently infected (not yet infectious); asymptomatic; pre-symptomatic; symptomatic with either mild or severe illness; recovered; and dead (Appendix: Figure S1 ). The model J o u r n a l P r e -p r o o f population was stratified into five age groups, 0-4, 5-19, 20-49, 50-64, and ≥65 years, based on US demographics (U.S. Census Bureau QuickFacts: United States 2020) . We parameterized the model (Table 1) , and determined the proportion of individuals in each age group who had one or more comorbidities associated with COVID-19 complications based on age-specific means derived from the 2017 Behavioural Risk Factor Surveillance System (Appendix : Table  S1 ) (Divo et al. 2014; Adams et al. 2020) . In the absence of any social distancing measures, the daily number of interactions within and between different age groups were sampled from negative-binomial distributions, and implemented based on an empirically determined contact network (Mossong et al. 2008) . When shelter-in-place as a social distancing measure was integrated, the network of interactions changed to an age-dependent contact matrix derived from a representative sample population during COVID-19 lockdown (Appendix: Table S2 ) (CMMID COVID-19 working group et al. 2020).

Disease transmission occurred probabilistically when susceptible individuals interacted with infectious individuals in asymptomatic, pre-symptomatic, or symptomatic stages of the disease. For each newly infected individual, we sampled an incubation period from a LogNormal distribution with an average of 5.2 days (Lauer et al. 2020; Li et al. 2020a) . A proportion of infected individuals develop symptoms after a highly infectious pre-symptomatic stage as part of their incubation period (He et al. 2020) . The pre-symptomatic period was sampled from a Gamma distribution with a mean of 2.3 days (He et al. 2020) . The infectious period following symptom onset was also sampled from a Gamma distribution with a mean of 3.2 days (Li et al. 2020b ). We used age-dependent estimates to determine the probability of developing mild, severe, or critical illness during the symptomatic infection period (Moghadas et al. 2020b; Shoukat et al. 2020) . Infected individuals who were not pre-symptomatic after the latent period became asymptomatic until recovery, with an infectious period that was sampled from a Gamma distribution with a mean of 5 days (Gatto et al. 2020; Li et al. 2020b) . Recovery from infection was assumed to provide adequate immunity for the duration of the outbreak, preventing reinfection.

Infectivity of individuals in asymptomatic, mild symptomatic, and severe symptomatic stages was parameterized relative to the infectivity during the pre-symptomatic stage. This parameterization was based on the proportion of secondary cases resulting from disease transmission during each stage of infection (Ferretti et al. 2020; Moghadas et al. 2020a ). Since the pre-symptomatic stage accounts for the highest proportion of secondary infections (Ferretti et al. 2020; Moghadas et al. 2020a ), we determined the relative infectivity of asymptomatic, mild symptomatic, and severe symptomatic stages to be 11%, 44%, and 89%, respectively (Moghadas et al. 2020a ).

We assumed that mild cases recover without the need for hospitalization. The probability of hospitalization and intensive care unit (ICU) admissions for severe and critical cases were J o u r n a l P r e -p r o o f informed by estimates from COVID-19 outbreaks in the US, with further classification of individuals with and without comorbidities (Centers for Disease Control and Prevention 2020b; Garg 2020b). Hospitalized patients were isolated and did not contribute to further transmission in the population. The average time from symptom onset to hospital admission was sampled in the range of 2 to 5 days (Moghadas et al. 2020b; Shoukat et al. 2020) . The length of non-ICU and ICU stays for hospitalized patients were sampled from Gamma distributions with means of 11.5, and 14.4 days, respectively (Sanche et al. 2020; Yang et al. 2020) . We assumed that severe symptomatic cases who were not hospitalized self-isolated immediately upon symptom onset, limiting their daily contacts to a maximum of 3 until recovery.

We considered non-medical cloth masks (hereinafter referred to masks) as an intervention measure for the general population, with compliance rates of 0%, 25%, 50%, and 75% for individuals aged 2 and older based on Centers for Disease Control and Prevention guidelines (CDC 2020) . This range of compliance is based on recent polling results suggesting that, despite mandatory mask-wearing in some states, the rate of compliance remains below 75% (CNN 2020b). We chose a conservative non-medical mask efficacy, eM = 20%, within the estimated range (Davies et al. 2013; MacIntyre et al. 2015; Konda et al. 2020; Mondal et al. 2020) for reducing disease transmission during interactions between susceptible and infected individuals. The probability of disease transmission was then reduced by a factor of (1-eM) or (1-eM) 2 depending on whether only one or both interacting individuals wore masks, respectively. We also implemented shelter-in-place strategies based on age and comorbidities by considering eight scenarios as described in Table 2 . Under any of these scenarios, the daily interactions of individuals who were sheltered-in were parameterized from an age-dependent contact matrix (Appendix: Table S2 ). The total number of contacts was sampled from a negative binomial distribution, with parameters derived from a recent study of contact patterns during COVID-19 lockdown (CMMID COVID-19 working group et al. 2020). When both shelter-in-place and maskwearing interventions were applied, only individuals who were not sheltered-in wore masks during daily interactions. In our model, the baseline scenario corresponds to the simulation of outbreaks without these interventions. For all other scenarios, we measured the effect of shelter-in-place and mask-wearing on reducing attack rate, hospitalizations, and deaths throughout the outbreak. In order to compare scenarios for shelter-in-place, we estimated the strategy efficiency (Se) by Se=La /Ns, where La is the cumulative number of the averted outcome of interest (i.e., infections, hospitalizations, or deaths) compared to no shelter-in-place intervention, and Ns is the number of individuals being sheltered-in. Using Se, we determined the most efficient strategy for shelter-in-place among simulated scenarios.

We calibrated the model to a baseline transmission probability per contact to obtain a reproduction number R0=2.5 (defined as the average number of secondary cases generated by a primary case), as estimated for initial COVID-19 outbreaks (Li et al. 2020a; Wu et al. 2020) . This reproduction number corresponds to an attack rate of 60% in an entirely susceptible population. Recent seroprevalence studies suggest that initial outbreaks have generated about 3.35% population level of immunity at the global scale (SeroTracker). For the US population, this level of immunity varies (95%CI: 3.59% -9.36%) (SeroTracker), and we considered scenarios with 5%. To account for the age distribution of pre-existing population immunity, we first simulated the model in an entirely susceptible population, replicating the scenario for initial outbreaks. We then used infection rates in different age groups and initialized our model with a population that includes immune individuals according to the age-specific distribution of preexisting immunity (Appendix:, Table S3 ). For the evaluation of intervention measures, we seeded simulations with one initial infection in the latent stage of the disease in a population of 10,000 individuals. We averaged the results over 1000 independent Monte-Carlo realizations in each scenario. The model was coded in Julia language and is available at https://github.com/thomasvilches/covid-shelterin.

We evaluated the impact of mask-wearing in the absence of a shelter-in-place strategy, assuming 5% population immunity. As the proportion of the population wearing masks increases from 0% to 75%, the peak incidence is delayed and its magnitude decreases ( Figure 1A ). For instance, with no mask-wearing (0% compliance) and no shelter-in intervention, a mean peak incidence of 146 per 10,000 population is observed 58 days into the outbreak. However, with 75% mask-wearing compliance, the peak incidence is reduced to 55 per 10,000 population (62.3% reduction) and occurs with a 3-week delay on day 81 of the outbreak. For a maskwearing compliance of 25%, we projected a median reduction of 9.6% (IQR: 7.3% -11.7%) for attack rate, 11.9% (IQR: 9.8% -14.3%) for hospitalizations, and 14.0% (IQR: 11.9% -16.1%) for deaths, compared to no mask-wearing ( Figure 1B ). As we increased mask-wearing compliance to 75%, the median reduction of attack rate, hospitalizations, and deaths was substantially higher at 37.7% (IQR: 36.1% -39.4%), 44.2% (IQR: 42.9% -45.8%), and 47.2% (IQR: 45.5% -48.7%), respectively ( Figure 1B ).

We also estimated the reduction of secondary infections that can be achieved by mask-wearing. Compared to the no intervention scenario (corresponding to model calibration with R0=2.5), mask-wearing compliance of 25%, 50%, and 75% reduced the reproduction number to 2.30, 2.04, and 1.79, respectively.

When shelter-in-place was implemented, the timing of peak incidence was delayed and its magnitude was reduced for each scenario (Figure 2 ). Assuming 5% population immunity and in the absence of mask-wearing, the lowest effect on reducing attack rate (median: 13.9%; IQR: 11.6% -16.2%) was associated with sheltering individuals 65 and older (S4). S4 also led to the highest and earliest peak of incidence (Figure 2A ). Sheltering children aged 5-19 years (S1) was slightly more effective in reducing the attack rate (median: 19.7%; IQR: 17.5% -21.4%) while delaying and lowering the peak incidence; however, S1 underperformed in terms of reducing hospitalizations and deaths compared to S4 (Figure 3 -A1, B1, C1) . We observed the greatest effects with strategies S6 and S8 (Figure 3 -A1, B1, C1) where comorbid individuals, in combination with those aged 50-64 (S6) or 50 and older (S8) were sheltered-in-place, on reducing attack rate (over 80%), hospitalizations (over 85%), and deaths (over 87%). We also found that, despite a significantly lower number of individuals (i.e., 18.9% of the population) being sheltered-in with strategy S3 (i.e., those aged 50-64 years), S3 outperformed strategy S7 in which over 40% of the population (i.e., all individuals with comorbidities or aged 65 and older) are sheltered-in-place (Figure 3 ).

When mask-wearing was implemented in combination with shelter-in-place interventions, the peak incidence was further delayed (Figure 2) , and the performance of all strategies improved proportionally in reducing attack rate, hospitalizations, and deaths ( Figure 3) . We found that when mask-wearing compliance was 75%, sheltering individuals aged 50-64 years (S3; 18.9% of the population) reduced attack rate, hospitalizations, and deaths with comparable rates to those obtained when all comorbid individuals or those aged 50 and older are sheltered-in-place (S8; 48.9% of the population) in the absence of mask-wearing ( Figure 3 ). This suggests that mask-wearing reduces the burden of disease and improves the performance of shelter-in strategies without increasing the number of individuals being sheltered-in-place (Appendix :  Tables S4-S6 ). With 75% mask-wearing, the median reduction in attack rate for S3 was projected at 82.5% (IQR: 81.6% -83.5%) and the median reduction of hospitalizations and deaths exceeded 86% and 87%, respectively.

The number of individuals affected by each shelter-in-place strategy is variable, due to the population distribution of the US, and the various combinations of age groups and comorbidities modelled in each scenario. Based on the number of averted infections, hospitalizations, and deaths per person sheltered-in-place, we found S3, where all individuals between 50 and 64 years of age are sheltered-in, to be the most efficient strategy (Figure 4 ). In the absence of mask-wearing, S3 resulted in the highest number of averted infections (median: 1.69; IQR: 1.65 -1.75), hospitalizations (median: 0.085; IQR: 0.083 -0.087), and deaths (median: 0.0091; IQR: 0.0088 -0.0093) per person sheltered-in, as seen in Figure 4 . This is in contrast to strategies S6 and S8, which provided the highest reduction in adverse clinical outcomes, as seen in Figure 3 , but with relatively low efficiency due to a high proportion of the population being sheltered-inplace. Although strategy S1 (children aged 5 to 19) calls for sheltering-in the same proportion of the population as strategy S3, S1 represents the least efficient strategy in terms of averting hospitalizations (0.021, IQR: 0.019 -0.024) and deaths (0.0024, IQR: 0.0022 -0.0027) per person sheltered-in. This is largely attributed to a high proportion of daily contacts occurring within their own age group (5-19 years), as well as better clinical outcomes when children are infected with COVID-19, compared to older age groups (Shekerdemian et al. 2020) .

Shelter-in-place and mask-wearing strategies can substantially reduce the magnitude of outbreaks and adverse clinical outcomes. However, since we excluded healthy individuals between 20 and 49 years of age from our shelter-in-place strategies, effective control of COVID-19 outbreaks will need to include the rapid identification of silent pre-symptomatic and asymptomatic infections (Moghadas et al. 2020a ). Therefore, we ran simulations to identify the level of non-symptomatic case detection needed with testing to bring the attack rate below 1% of the population, when a highly efficient strategy of sheltering individuals aged 50-64 years (S3) is combined with mask-wearing by those not sheltered-in-place. Our results show that if testing capacity and contact tracing allows for the identification of 33% of silent infections, outbreak control can be achieved with the implementation of S3 and 75% mask-wearing ( Figure 5D ). When mask-wearing compliance is ≤50%, attack rates in most strategies remain above 2% even with a 33% detection rate of silent infections ( Figure 5A , B, C).

In the absence of a COVID-19 vaccine, mitigation measures to curb initial outbreaks have included strict social distancing and movement restrictions (Flaxman et al. 2020; Khosrawipour et al. 2020; Lau et al. 2020; The Lancet 2020) . The resulting societal and economic repercussions have led to a hasty reopening, causing a resurgence of cases in many states. Curtailing these outbreaks, however, is unlikely in the presence of insufficient testing, inadequate social distancing, and the ongoing debate (Feng et al. 2020) over the use of masks in public. We aimed to quantify the effect of mask-wearing and shelter-in-place to identify the optimal strategies for effective control of ongoing and future COVID-19 outbreaks. A strategic approach to lifting restrictive public health measures will help to facilitate a safe economic recovery.

Our results show that the greatest reduction of attack rate, hospitalizations, and deaths is achieved when nearly half of the population is sheltered-in-place ( Figure 3) . However, a similar impact can be realized by sheltering a substantially smaller proportion (~19%) of the population (S3: individuals aged 50-65) when 75% of those who are not sheltered-in-place wear masks. However, when considering a similar proportion of the population in S4, the strategy efficiency drops substantially, despite having a high proportion of comorbid individuals, because persons aged 65 and older have the lowest number of daily contacts of any age group. On a population level, our results show that school closures (sheltering children 5-19 years of age) is comparatively less effective in terms of reducing hospitalizations and mortality. This is predominantly explained by empirical observations that over 60% of daily contacts of school children occur among their own age group (Mossong et al. 2008 ) (Appendix: Table S2 ), and that they are more likely to exhibit milder COVID-19 outcomes with lower hospitalization rates (Shekerdemian et al. 2020) .

The use of face coverings has been recommended by the Centers for Disease Control and Prevention to help reduce the spread of COVID-19 (CDC 2020) . However, there has been widespread debate on the effectiveness of mask-wearing, despite preliminary evidence suggesting that it can help reduce transmission (MacIntyre et al. 2015; CDC 2020; Konda et al. 2020; Mondal et al. 2020) . Our results demonstrate the benefits of mask-wearing on reducing the spread of infection and adverse clinical outcomes, particularly when combined with shelterin-place strategies for vulnerable populations. The assumed non-medical mask efficacy in our J o u r n a l P r e -p r o o f analysis (20%) is likely to be conservative given the range of estimates (20%-80%) used in previous studies Ngonghala et al. 2020) ; however, factors such as materials used to make cloth masks (Konda et al. 2020) , imperfect use (Mondal et al. 2020) , and behaviour of the mask-wearer could reduce their effectiveness (Stutt et al. 2020) . If nonmedical masks are more effective than assumed in this study, our results would be conservative, and a greater impact on reducing disease burden would be expected.

Our study has important implications for COVID-19 mitigation strategies. First, a strategic and coordinated response is necessary to suppress the ongoing resurgence of cases in the US. Second, mask-wearing and shelter-in-place continue to be important measures to reduce disease burden, and enhancing the capacity for testing and contact tracing remains a critical pathway towards curbing the trajectory of developing outbreaks. Given that the majority of COVID-19 transmission is attributable to shedding from pre-symptomatic and asymptomatic individuals (Moghadas et al. 2020a) , outbreak control cannot be achieved without the detection and isolation of at least one-third of non-symptomatic cases. In this context, mask-wearing can help to reduce the risk of silent transmission. Finally, prior to vaccine availability, a judicious implementation of shelter-in-place strategies could have a large impact on the control of ongoing and future outbreaks, while minimizing socioeconomic repercussions in the coming months. Figure 1. (A) Projected incidence of COVID-19 infections per 10,000 population at different levels of mask-wearing compliance and with 5% level of pre-existing immunity. (B) Reduction of attack rate, hospitalizations, and deaths at different levels of mask-wearing compliance, compared to no mask-wearing. Simulations correspond to mask-wearing scenarios in the absence of shelter-in-place strategies.

Projected incidence of COVID-19 with different shelter-in-place strategies, in combination with mask compliance of 0% (A); 25% (B); 50% (C); and 75% (D) among those not sheltered-in-place. The level of pre-existing immunity in the population was assumed to be 5% for all scenarios. Figure 3 . Reduction of attack rate, hospitalizations, and deaths achieved with each shelter-inplace strategy, in combination with mask compliance of 0% (A1, B1, C1); 25% (A2, B2, C2); 50% (A3, B3, C3); and 75% (A4, B4, C4). The level of pre-existing immunity in the population was assumed to be 5% for all scenarios. 

Population based estimates of comorbidities affecting risk for complications from COVID-19 in the US. medRxiv

Lessons from a rapid systematic review of early SARS-CoV-2 serosurveys

Preliminary Estimates of the Prevalence of Selected Underlying Health Conditions Among Patients with Coronavirus Disease 2019 -United States

Quantifying the impact of physical distance measures on the transmission of COVID-19 in the UK

Some US states return to previous restrictions to slow surge of coronavirus cases

The Northeast leads the country in mask-wearing

Testing the efficacy of homemade masks: would they protect in an influenza pandemic

Ageing and the epidemiology of multimorbidity

Achievement gap and coronavirus | McKinsey

To mask or not to mask: Modeling the potential for face mask use by the general public to curtail the COVID-19 pandemic. Infectious Disease Modelling

Rational use of face masks in the COVID-19 pandemic. The Lancet Respiratory Medicine

Economic Effects of Coronavirus Outbreak (COVID-19) on the World Economy

Social Science Research Network

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

Estimating the effects of non-pharmaceutical interventions on COVID-19 in Europe

Hospitalization Rates and Characteristics of Patients Hospitalized with Laboratory-Confirmed Coronavirus Disease 2019 -COVID-NET, 14 States

Hospitalization Rates and Characteristics of Patients Hospitalized with Laboratory-Confirmed Coronavirus Disease 2019 -COVID-NET, 14 States

Spread and dynamics of the COVID-19 epidemic in Italy: Effects of emergency containment measures

Temporal dynamics in viral shedding and transmissibility of COVID-19

Johns Hopkins Coronavirus Resource Center

Failure in initial stage containment of global COVID-19 epicenters

Aerosol Filtration Efficiency of Common Fabrics Used in Respiratory Cloth Masks

The positive impact of lockdown in Wuhan on containing the COVID-19 outbreak in China

The Incubation Period of Coronavirus Disease 2019 (COVID-19) From Publicly Reported Confirmed Cases: Estimation and Application

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

Substantial undocumented infection facilitates the rapid dissemination of novel coronavirus (SARS-CoV-2)

A cluster randomised trial of cloth masks compared with medical masks in healthcare workers

Estimating the asymptomatic proportion of coronavirus disease 2019 (COVID-19) cases on board the Diamond Princess cruise ship

Projecting hospital utilization during the COVID-19 outbreaks in the United States

Utility of Cloth Masks in Preventing Respiratory Infections: A Systematic Review

Social Contacts and Mixing Patterns Relevant to the Spread of Infectious Diseases

Mathematical assessment of the impact of non-pharmaceutical interventions on curtailing the 2019 novel Coronavirus

Estimation of the asymptomatic ratio of novel coronavirus infections (COVID-19)

Coronavirus: A visual guide to the economic impact. BBC News

The Novel Coronavirus, 2019-nCoV, is Highly Contagious and More Infectious Than Initially Estimated

COVID-19 Seroprevalence

Characteristics and Outcomes of Children With Coronavirus Disease 2019 (COVID-19) Infection Admitted to US and Canadian Pediatric Intensive Care Units

Projecting demand for critical care beds during COVID-19 outbreaks in Canada

Coronavirus Disease 2019 Case Surveillance -United States

A modelling framework to assess the likely effectiveness of facemasks in combination with 'lock-down' in managing the COVID-19 pandemic

India under COVID-19 lockdown

Estimation of Excess Deaths Associated With the COVID-19 Pandemic in the United States

World Health Organization. WHO Coronavirus Disease (COVID-19) Dashboard

Nowcasting and forecasting the potential domestic and international spread of the 2019-nCoV outbreak originating in Wuhan, China: a modelling study. The Lancet

Clinical course and outcomes of critically ill patients with SARS-CoV-2 pneumonia in Wuhan, China: a single-centered, retrospective, observational study. The Lancet Respiratory Medicine

Availability of data and materials: Details of the model are provided in Appendix. TheThe authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.