Wuhan lockdown: drastic but live-saving

Coronavirus / Covid-19 cases in China. (9.04.2020) Source: Center for Systems Science and Engineering (CSSE) at JHU

Editor's Introduction

An investigation of transmission control measures during the first 50 days of the COVID-19 epidemic in China

annotated by

COVID-19 has disturbed the world unprecedentedly, taking millions of lives and breaking the families behind. China, the country that reported the first case, had taken prompt measures when the contagious disease first outbroke, suspending transport and imposing social distancing. People in this country are now on the track to normal life. Other countries that attempted to follow this protocol, however, met resistance and questions. Since more contagious variants are emerging, we need to scrutinize the necessity of these measures. Are these massive measures effective? Does the impact depend on when and how these measures are implemented?

Paper Details

Original title
An investigation of transmission control measures during the first 50 days of the COVID-19 epidemic in China
Original publication date
Issue name


Responding to an outbreak of a novel coronavirus (agent of COVID-19) in December 2019, China banned travel to and from Wuhan city on 23 January and implemented a national emergency response. We investigated the spread and control of COVID-19 using a unique data set including case reports, human movement and public health interventions. The Wuhan shutdown was associated with the delayed arrival of COVID-19 in other cities by 2.91 days (95%CI: 2.54-3.29). Cities that implemented control measures pre-emptively reported fewer cases, on average, in the first week of their outbreaks (13.0; 7.1-18.8) compared with cities that started control later (20.6; 14.5-26.8). Suspending intra-city public transport, closing entertainment venues and banning public gatherings were associated with reductions in case incidence. The national emergency response appears to have delayed the growth and limited the size of the COVID-19 epidemic in China, averting hundreds of thousands of cases by 19 February (day 50).


On 31 December 2019, less than a month before the 2020 Spring Festival holiday, including the Chinese Lunar New Year, a cluster of pneumonia cases caused by an unknown pathogen was reported in Wuhan, a city of 11 million inhabitants and the largest transport hub in Central China. A novel coronavirus (12) was identified as the etiological agent (34) and human-to-human transmission of the virus (SARS-CoV-2) has been since confirmed (56). Further spatial spread of this disease was of great concern in view of the upcoming Spring Festival (“chunyun”) during which there are typically three billion travel movements over the 40-day holiday period, which runs from 15 days before the Spring Festival (Chinese Lunar New Year) to 25 days afterwards (78).

As there is currently neither a vaccine nor a specific drug treatment for COVID-19, a range of public health (non-pharmaceutical) interventions has been used to control the epidemic. In an attempt to prevent further dispersal of COVID-19 from its source, all transport was prohibited in and out of Wuhan city from 10:00h on 23 January 2020, followed by the whole of Hubei Province a day later. In terms of the population covered, this appears to be the largest attempted cordon sanitaire in human history.

On 23 January, China also raised its national public health response to the highest state of emergency—Level 1 of 4 levels of severity in the Chinese Emergency System, defined as an “extremely serious incident” (9). As part of the national emergency response, and in addition to the Wuhan city travel ban, suspected and confirmed cases have been isolated, public transport by bus and subway rail suspended, schools and entertainment venues have been closed, public gatherings banned, health checks carried out on migrants (“floating population”), travel prohibited in and out of cities, and information widely disseminated. Despite all these measures, COVID-19 remains a danger in China. Control measures taken in China potentially hold lessons for other countries around the world.

Although the spatial spread of infectious diseases has been intensively studied (1015), including explicit studies of the role of human movement (1617), the effectiveness of travel restrictions and social distancing measures in preventing the spread of infection is uncertain. For COVID-19, coronavirus transmission patterns and the impact of interventions are still poorly understood (67). We therefore carried out a quantitative analysis to investigate the role of travel restrictions and transmission control measures during the first 50 days of the COVID-19 epidemic in China, from 31 December 2019 to 19 February 2020 (Fig. 1). This period encompassed the 40 days of the Spring Festival holiday, 15 days before the Chinese Lunar New Year on 25 January and 25 days afterwards. The analysis is based on a unique geocoded repository of data on COVID-19 epidemiology, human movement, and public health (non-pharmaceutical) interventions. These data include the numbers of COVID-19 cases reported each day in each city of China, information on 4.3 million human movements from Wuhan city, and data on the timing and type of transmission control measures implemented across cities of China.



Fig. 1 Dates of discovery of the novel coronavirus causing COVID-19, and of the implementation of control measures in China, from 31 December 2019.

Draw a control measure timeline

This figure summarizes China’s control measures in the first 50 days since the breakout of COVID-19 in Wuhan and the timing of implementation. Outlining these measures allowed the authors to identify measures whose impacts can be studied. There must be differences in the implementation or timing of execution for a measure to be included in a correlational study.

You can gather information about your country and draw a similar timeline. Do you see some similarities or differences in the measures taken? What do you think of the effectiveness of these measures?

The effectiveness

As shown in the figure, massive control measures have been taken in China during the early stage of the breakout, and many involved restraining human mobility. On studying the effectiveness and necessity of the mobility restrictions, the present study is joined by many more. Watch this video and listen to Prof. Yang Yang explaining her findings.

We first investigated the role of the Wuhan city travel ban, comparing travel in 2020 with that in previous years and exploring how holiday travel links to the dispersal of infection across China. During Spring Festival travel in 2017 and 2018, there was an average outflow of 5.2 million people from Wuhan city during the 15 days before the Chinese Lunar New Year. In 2020, this travel was interrupted by the Wuhan city shutdown, but 4.3 million people travelled out of the city between 11 January and the implementation of the ban on 23 January (Fig. 2A) (7). In 2017 and 2018, travel out of the city during the 25 days after the Chinese Lunar New Year averaged 6.7 million people each year. In 2020, the travel ban prevented almost all of that movement and markedly reduced the number of exportations of COVID-19 from Wuhan (78).



Fig. 2 The dispersal of COVID-19 in China 15 days before and 25 days after the Spring Festival (Chinese Lunar New Year). (A) Movement outflows from Wuhan City during Spring Festival travel in 2017, 2018, and 2020. The vertical dotted line is the date of Spring Festival (Chinese Lunar New Year). (B) The number of recorded movements from Wuhan city to other provinces during the 15 days before the Spring Festival in 2020. The shading from light to dark represents the number of human movements from Wuhan to each province. The area of circles represents the cumulative number of cases reported by 30 January 2020, one week after the Wuhan travel ban on 23 January. (C) Association between the cumulative number of confirmed cases reported before 30 January and the number of movements from Wuhan to other provinces. 

Outflow from Wuhan 

Panel A summarizes the outflows from Wuhan during the Spring Festival season in two normal years (2017 and 2018) and the year when COVID-19 struck Wuhan. The patterns for 2017 and 2018 are fairly consistent. The consistency implies that the trends are representative. Having examined the representativeness, the author thus used the data from the year 2018 for their linear regression, which yielded results in Table 1 and Figure 3B.

The sharp decline in the outflow after the Wuhan lockdown, compared to the typical trends of two previous years, indicates the power of this measure in controlling human movement from Wuhan.

Human mobility and spread of COVID-19

Panel B gives us a first expression of the potential correlation between human movement from Wuhan and the growth of COVID-19 in one province. Bigger red dots are generally present in provinces in darker blue, meaning a larger number of cumulative cases by 30 January 2020 were reported in provinces with more inbound travelers from Wuhan. The linear regression presented in Panel C further shows a positive correlation between the cumulative number of COVID-19 cases and human movement from Wuhan.

Further questions

Where would the data point for Hubei province be in Panel C if Wuhan was considered? Based on your answer to the first question, why do you think Wuhan was excluded from the calculation of cumulative cases in Wuhan?

The dispersal of COVID-19 from Wuhan was rapid (Fig. 3A). A total of 262 cities reported cases within 28 days. For comparison, the 2009 influenza H1N1 pandemic took 132 days to reach the same number of cities in China (see methods in Supplementary Materials). The number of cities providing first reports of COVID-19 peaked at 59 per day on 23 January, the date of the Wuhan travel ban.



Fig. 3 Spatial dispersal of COVID-19 in China. (A) Cumulative number of cities reporting cases by 19 February 2020. Arrival days, defined as the time interval (days) from the date of the first case in the first infected city (Wuhan) to the date of the first case in each newly infected city (a total of 324 cities), to characterize the inter-city transmission rate of COVID-19. Dashed line shows the date of Wuhan travel ban (shutdown). (B) Before (blue) and after (red) the intervention by 30 January 2020, one week after Wuhan travel ban (shutdown). The blue line and points show the fitted regression of arrival times up to the shutdown on day 23 (23 January, vertical dashed line). Grey points show the expected arrival times after day 23, without the shutdown. The red line and points show the fitted regression of delayed arrival times after the shutdown on day 23. Each observation (point) represents one city. Error bars give ±2 standard deviations.

Result interpretation

Panel A shows the cumulative number of cities reporting cases. The figure is the number of towns that had confirmed their first COVID-19 cases by the time indicated on the horizontal axis. This figure saturates after the fortieth day, meaning the arrival of COVID-19 in most of the reachable cities surveyed in this study.

Panel B presents the fitted arrival time for two groups of cities. The first group of towns that had reported their first COVID-19 cases are shown by blue dots, where those that reported the first cases after the Wuhan lockdown are indicated by red dots. The difference between the red and blue lines indicates the average delayed arrival time by 2.91 days.

Further questions

There seem to be nearly no new cities reporting confirmed COVID-19 cases in the first two weeks after the report of the first case in Wuhan. What would this observation imply?

The total number of cases reported from each province by 30 January, one week after the Wuhan shutdown, was strongly associated with the total number of travellers from Wuhan (r = 0.98, P < 0.01; excluding Hubei, r = 0.69, P < 0.01) (Fig. 2, B and C). COVID-19 arrived sooner in those cities that had larger populations and had more travellers from Wuhan (Table 1 and table S1). However, the Wuhan travel ban was associated with a delayed arrival time of COVID-19 in other cities by an estimated 2.91 days (95%CI: 2.54 to 3.29 days) on average (Fig. 3B and Table 1).



Table 1 Association between the Wuhan travel ban and COVID-19 dispersal to other cities in China. The dependent variable Y is the arrival time (days) of the outbreak in each city.

Variables explained

This linear regression involves multiple variables.

The authors are concerned about the effectiveness of the Wuhan travel ban on slowing the spread of COVID-19. The spread was indicated by the time when a city reported its first confirmed case, i.e., the arrival time. This dependent variable has been defined with respect to 31 December 2019, on which day Wuhan reported the first COVID-19 case. The independent variables considered in this study include the geographical location of a city (represented by the longitude and latitude), its population, and the inflow of travelers from Wuhan. The implementation of the travel ban has also been included as a separate independent variable.

Two points are worth noting here. First, the data for travels from Wuhan to a city were those for Spring Festival travel 2018. Second, the population and the total inflow went through log transformations. These transformations allow us to zoom into the (comparatively tiny) differences between these big numbers. For more detailed explanations, check this link

Types of variables

All variables in this regression, except the implementation of the travel ban, are quantitative variables, meaning that they take numbers. The only exception is a categorical variable by nature. To fit it into the linear regression, the authors used a binary dummy variable ("Shutdown" as in the model presented in the Supplementary Materials). This dummy variable was set to 1 if the Wuhan travel ban was implemented before the breakout of COVID-19 in one city, or, equivalently, no cases were reported in one city before 23 January 2020 when the ban was activated. For the opposite, the dummy variable was set to 0.

Construction of the regression model

The regression model used here was chosen out of five candidates the authors had proposed initially. For the complete list, see Table S1 in the Supplementary Materials. The best-fit model should explain the greatest amount of variation in the dependent variable using the fewest possible independent variables. This model gives the lowest Akaike’s information criterion (AIC) among all proposed candidates. If you are interested in learning more about AIC, check out this introductory video

Further investigation

Can you think of other factors that may have affected the arrival time of COVID-19 in different cities? What data will you need to test your guess? What type will your newly added variable be?

Are you ready to try your models? The data and code used in the present study can be found here. Building on the data set and modifying the codes will be a good starting point.

This delay provided extra time to prepare for the arrival of COVID-19 in more than 130 cities across China but would not have curbed transmission after infection had been exported to new locations from Wuhan. Figure 1 shows the timing and implementation of emergency control measures in 342 cities across China (see also figs. S2 and S4). School closure, the isolation of suspected and confirmed patients, plus the disclosure of information was implemented in all cities. Public gatherings were banned and entertainment venues closed in 220 cities (64.3%). Intra-city public transport was suspended in 136 cities (39.7%) and inter-city travel was prohibited by 219 cities (64.0%). All three measures were applied in 136 cities (table S2).

Cities that implemented a Level 1 response (any combination of control measures) (figs. S2 and S4) pre-emptively, before discovering any COVID-19 cases, reported 33.3% (95%CI: 11.1-44.4%) fewer laboratory-confirmed cases during the first week of their outbreaks (13.0, 95%CI: 7.1-18.8, n= 125) compared with cities that started control later (20.6 cases, 95%CI: 14.5-26.8, n = 171), with a statistically significant difference between the two groups (Mann-Whitney U = 8197, z = –3.4, < 0.01). A separate analysis using regression models shows that, among specific control measures, cities that suspended intra-city public transport and/or closed entertainment venues and banned public gatherings, and did so sooner, had fewer cases during the first week of their outbreaks (Table 2 and table S3). This analysis provided no evidence that the prohibition of travel between cities, which was implemented after and in addition to the Wuhan shutdown on 23 January, reduced the number of cases in other cities across China. These results are robust to the choice of statistical regression model (table S3).


Table 2 Associations between the type and timing of transmission control measures and the number of COVID-19 cases reported in city outbreaks (first week), evaluated by a generalised linear regression model.

Poisson regression

This table summarizes how the suspension of intra-city public transport and closure of entertainment venues are associated with the growth of COVID-19 in one city. Two more factors, the arrival time of COVID-19 and the distance of a town from Wuhan, have also been considered. The results were derived through a Poisson regression, i.e., the generalized linear regression model.

Poisson regression is used when the response (dependent) variable is assumed to follow a Poisson distribution, as in this case. The dependent variable also needs to take values of small positive integers. A linear correlation between the logarithm of the dependent variable’s expected value (its weighted average) and other independent variables is to be found through this regression.

If you are interested in the details of Poisson regression, watch this video

Coefficients explained

This model has two binary variables: the implementations of suspending intra-city public transport and closing entertainment venues. The binary variables take one if the corresponding measure was implemented before or during the first seven days of the outbreak in a city and zero for the opposite cases. The negative coefficients for both indicate the negative correlation for both measures with the number of confirmed cases, meaning cities taking these measures had fewer cases. The positive coefficients for the implementation timings imply that cities that responded sooner had fewer cases. All these coefficients come with p-values smaller than 0.01; they are thus significantly different from the values as if no correlation were present.

Intercept explained

This regression returned a negative intercept. The intercept, however, should not be interpreted as there would be no cases if all independent variables in the model were evaluated at zero. This misinterpretation is not allowed because zero values for the arrival time, the logarithm of distance from Wuhan, and timings of implementing measures are out of the plausible range. For instance, the arrival time can never be zero as virtually all cities included in this study were infected, as shown in Figure 3A. 

The reported daily incidence of confirmed cases peaked in Hubei province (including Wuhan) on 4 February (3156 laboratory-confirmed cases, 5.33/100,000 population in Hubei), and in all other provinces on 31 January (875 cases, 0.07/100,000 population) (fig. S1). The low level of peak incidence per capita, the early timing of the peak, and the subsequent decline in daily case reports, suggest that transmission control measures were not only associated with a delay in the growth of the epidemic, but also with a marked reduction in the number of cases. By fitting an epidemic model to the time series of cases reported in each province (fig. S3), we estimate that the (basic) case reproduction number (R0) was 3.15 prior to the implementation of the emergency response on 23 January (Table 3). As control was scaled-up from 23 January onwards (stage 1), the case reproduction number declined to 0.97, 2.01 and 3.05 (estimated as C1R0) in three groups of provinces, depending on the rate of implementation in each group (Table 3 and table S4). Once the implementation of interventions was 95% complete everywhere (stage 2), the case reproduction number had fallen to 0.04 on average (C2R0), far below the replacement rate (<< 1) and consistent with the rapid decline in incidence (Fig. 4ATable 3, fig. S5, and table S4).


Table 3 Parameter estimates of the SEIR epidemic model. BCI: Bayesian confidence interval; C1_high, Heilongjiang, Shanghai, Tianjin, Zhejiang, Hubei (exclude Wuhan); C1_medium, Anhui, Beijing, Fujian, Guangdong, Guangxi, Guizhou, Hunan, Jilin, Jiangsu, Jiangxi, Inner Mongolia, Shandong, Tibet; C1_low, Gansu, Hainan, Hebei, Henan, Liaoning, Ningxia, Qinghai, Shanxi, Shaanxi, Sichuan, Xinjiang, Yunnan, Chongqing.

SEIR epidemic model

The S, E, I, and R in this model represent susceptible, exposed, infectious, and recovered, respectively. In this model, susceptible people are first infected through exposure to the infectious agent. They are not contagious at this stage and are thus called the latent or exposed population. The length of this stage is the latent period. After an incubation (latent) period, some of these latent become infectious. Infectious people should usually be isolated and admitted for medical treatments. Having beaten the disease, people are recovered and gain total immunity, meaning that they will not be infected again.

Under this model, the number of exposed people firstly climbs and then decays after a peak. A similar but delayed trend is for the infectious population. The susceptible population, since the outbreak, gradually decreases until reaching zero, parallel to the increase in the recovered population.

For more details of SEIR and other epidemic models, check this link

Bayesian statistics

The authors used Bayesian statistics to study the effectiveness of control measures in lowering the basic reproduction number. In a typical Bayesian workflow, before actually seeing the data, researchers first come up with a prior, an informed guess of the parameters in a statistical model based on literature study and theories. They then take one observation and calculate the probability for this observation given the prior. This probability is called the likelihood. Combining the prior and the likelihood leads to a posterior, which is the updated knowledge (i.e., a revised guess of parameters). The posterior then becomes the prior for the next run, with another observation fed in. Through multiple (a considerable number of) cycles like this, we can reach a likely result that is close to reality. Check this link for one example of Bayesian statistics

In the present study, the authors started with guesses about the basic reproduction number, the mean latent period, and the infectious period before isolation, based on their survey of the literature. They then picked one province and calculated how probable the simulation parameters for this province, based on the SEIR epidemic model, to be valid given the prior. Accordingly, they updated the prior and started another run. A total of 10 million runs (i.e., iterations) were taken to derive the results summarized in this table.

Data interpretation

The lowering of the basic reproduction number is shown by a series of constants (C1_high, C1_medium, C1_low, and C2). The smaller one constant is, the more significant the reduction in the reproduction number is.

Given these hints, how will you interpret the effects of measures taken in the first stage and the collective impact of control measures taken nationally in the second stage? Do your interpretations agree with the authors’?




Fig. 4 The role of interventions in controlling the COVID-19 outbreak across China. (A) Epidemic model (line) fitted to daily reports of confirmed cases (points) summed across 31 provinces. Hubei excludes Wuhan city. (B) Expected epidemic trajectories without the Wuhan travel ban (shutdown), and with (green) or without (red) interventions carried out as part of the Level 1 national emergency response; with the Wuhan travel ban, and with (black) or without the intervention (orange). Vertical dashed lines in both panels mark the date of the Wuhan travel ban and the start of the emergency response, on 23 January. Shaded regions in A and B mark the 95% prediction envelops.

95% prediction envelope

The lines in both panels are presented with 95% prediction envelopes. These shaded bands are equivalent to the 95% confidence intervals for point data we have seen earlier in this paper. The bands are formed by connecting the upper and lower limits of 95% confidence intervals of point estimates (e.g., estimated number of COVID-19 cases on one day in panel A). They show the uncertainty of the simulated model lines.

Validating model

The authors used the SEIR model to fit the case number curve of China in the first 50 days after the outbreak. To validate the model parameters as reported in Table 3, they compare the real case numbers with the estimated values based on the model in Panel A. The consistency between the two indicates the validity of the epidemic model used. Another thing to note: the black curve in Panel A is the same as the black one in Panel B.

Further questions

What would the red, orange, and blue curves in Panel B be like after the studied period? What is the curve of the number of COVID-19 cases in your home country or a country of your interest? Does it have a shape prescribed by the SEIR model? What are the possible reasons for the ups and downs you will see in one curve?

Based on the fit of the model to daily case reports from each province, and on the preceding statistical analyses, we investigated the possible effects of control measures on the trajectory of the epidemic outside Wuhan city (Fig. 4B). Our model suggests that, without the Wuhan travel ban or the national emergency response, there would have been 744,000 (± 156,000) confirmed COVID-19 cases outside Wuhan by 19 February, day 50 of the epidemic. With the Wuhan travel ban alone, this number would have decreased to 202,000 (± 10,000) cases. With the national emergency response alone (without the Wuhan travel ban), the number of cases would have decreased to 199,000 (± 8500). Thus, neither of these interventions would, on their own, have reversed the rise in incidence by 19 February (Fig. 4B). But together and interactively, these control measures offer an explanation of why the rise in incidence was halted and reversed, limiting the number of confirmed cases reported to 29,839 (fitted model estimate 28,000 ± 1400 cases), 96% fewer than expected in the absence of interventions.

In summary, this analysis shows that transmission control (non-pharmaceutical) measures initiated during Chinese Spring Festival holiday, including the unprecedented Wuhan city travel ban and the Level 1 national emergency response, were strongly associated with, though not necessarily the cause of, a delay in epidemic growth and a reduction in case numbers during the first 50 days of the COVID-19 epidemic in China.

The number of people who have developed COVID-19 during this epidemic, and therefore the number of people who were protected by control measures, is not known precisely, given that cases were almost certainly under-reported. However, in view of the small fraction of people known to have been infected by 19 February (75,532 cases, 5.41 per 100,000 population), it is unlikely that the spread of infection was halted and epidemic growth reversed because the supply of susceptible people had been exhausted. This implies that a large fraction of the Chinese population remains at risk of COVID-19; control measures may need to be reinstated, in some form, if there is a resurgence of transmission. Further investigations are needed to verify that proposition, and population surveys of infection are needed to reveal the true number of people who have been exposed to this novel coronavirus.

We could not investigate the impact of all elements of the national emergency response because many were introduced simultaneously across China. However, this analysis shows that suspending intra-city public transport, closing entertainment venues and banning public gatherings, which were introduced at different times in different places, were associated with the overall containment of the epidemic. However, other factors are likely to have contributed to control, especially the isolation of suspected and confirmed patients and their contact, and it is not yet clear which parts of the national emergency response were most effective. We did not find evidence for prohibiting travel between cities or provinces reduced the numbers of COVID-19 cases outside Wuhan and Hubei, perhaps because such travel bans were implemented as a response to, rather than in anticipation of, the arrival of COVID-19.

This study has drawn inferences, not from controlled experiments, but from statistical and mathematical analyses of the temporal and spatial variation in case reports, human mobility and transmission control measures. With that caveat, control measures were strongly associated with the containment of COVID-19, potentially averting hundreds of thousands of cases by 19 February, day 50 of the epidemic. Whether the means and the outcomes of control can be replicated outside China, and which of the interventions are most effective, are now under intense investigation as the virus continues to spread worldwide.




Materials and Methods

Figs S1 to S7

Tables S1 to S4

References (1928)



We thank the thousands of CDC staff and local health workers in China who collected data and continue to work to contain COVID-19 in China and elsewhere. Funding: this study was provided by the National Natural Science Foundation of China (81673234); Beijing Natural Science Foundation (JQ18025); Beijing Advanced Innovation Program for Land Surface Science; Young Elite Scientist Sponsorship Program by CAST (YESS)(2018QNRC001); HT, MUGK, OGP and CD acknowledge support from the Oxford Martin School; HT acknowledges support from the Military Logistics Research Program. The funders had no role in study design, data collection and analysis, the decision to publish, or in preparation of the manuscript. Author contributions:H.T., P.Z., R.F.Y., O.G.P., B.T.G., C.D. designed the study. B.C. and Y.M.S. collected and processed the Tencent’s LBS data. Y.H.L., B.Y.L., B.X., Q.Q.Y., B.W., P.Y., Y.J.C., Q.Y.W. collected the statistical data. H.Y.T., Y.L., C.H.W, and J.C. conducted the analyses. M.K., O.N.B., R.F.Y., O.G.P., B.T.G., and C.D. edited the manuscript. H.T. and C.D. wrote the manuscript. All authors read and approved the manuscript. Competing interests: All other authors declare no competing interests. Data and materials availability: Code and data are available on the following GitHub repository: https://github.com/huaiyutian/COVID-19_TCM-50d_China (18). This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/. This license does not apply to figures/photos/artwork or other content included in the article that is credited to a third party; obtain authorization from the rights holder before using such material.



1. N. Zhu et al.N. Engl. J. Med. 382, 727-733 (2020).

2. R. Lu et al.Lancet 395, 565-574 (2020).

3. F. Wu et al.Nature 579, 265-269 (2020).

4. P. Zhou et al.Nature 579, 270-273 (2020).

5. J. Cai et al.Int. J. Environ. Res. Public Health 16, 222 (2019).

6. C. Wang et al.Lancet 395, 470-473 (2020).

7. S. Chen et al.Lancet 395, 764-766 (2020).

8. M. U. G. Kraemer et al.Science in press (2020).

9. Chinadaily, "Tibet activates highest-level public health alert" (Jan 30, 2020). https://www.chinadaily.com.cn/a/202001/29/WS5e318a36a3101282172739c1.html

10. B. T. Grenfell, O. N. Bjørnstad, J. Kappey, Nature 414, 716-723 (2001).

11. D. Brockmann, D. Helbing, Science 342, 1337-1342 (2013).

12. A. Wesolowski et al.Science 338, 267-270 (2012).

13. N. M. Ferguson et al.Nature 437, 209-214 (2005).

14. K. E. Jones et al.Nature 451, 990-993 (2008).

15. D. M Morens, G. K. Folkers, A. S. Fauci, Nature 430, 242-249 (2004).

16. C. Viboud et al.Science 312, 447-451 (2006).

17. A. Wesolowski et al.Proc. Natl. Acad. Sci. U.S.A. 112, 11887-11892 (2015).

18. H. Tian et al., Zenodo (2020).