key: cord-0742633-wtsy1ein authors: Lin, Y.; Peng, F. title: Control Strategies against COVID-19 in China: Significance of Effective Testing in the Long Run date: 2020-08-24 journal: nan DOI: 10.1101/2020.08.22.20179697 sha: 5f42cf359e5fd62f657f7d23502f568d2677722c doc_id: 742633 cord_uid: wtsy1ein The coronavirus disease 2019 (COVID-19) outbreak is reasonably contained in China. In this paper, we evaluated the effectiveness of different containment strategies in halting the pandemic spread in both short- and long-term. We combined a networked metapopulation SEIR model featuring undocumented infections, actual mobility data and Bayesian inference to simulate the counterfactual outbreak scenarios removing each one or a combination of the following three policies in place: i) city lockdowns, ii) intercity travel bans, and iii) testing, detection, and quarantine. Our estimates revealed that 11.4% [95% credible interval (CI): 9.7-13.0%] of the infected cases were unidentified before January 23, 2020. The rate grew to 92.5% [95% credible interval (CI): 85.9-94.5%] in early March, thanks to the boost in coronavirus testing capacity. We show that increasing the detection rate of infections from 11.4% to 92.5% alone would explain 75% of the reduction in infections from a no-policy baseline by March 15, 2020. The most pronounced policy implication is that city lockdowns appeared to be the more effective intervention in the short-run but effective testing is essential in containing the COVID-19 spread in the long run. By March 15, restoring within-city personal contact to its 2019 level would lead to a 678% growth in infections with all the other interventions remaining unaffected. Removing intercity travel restrictions and effective detection measures would lead to 3% and 477% growth, respectively. Extending the time horizon to July 15, the counterfactual increase in infections would become 581%, 3% and 30000% had the three classes of interventions been lifted individually. Up to June 30, 2020, more than 10.1 million cases of coronavirus disease 2019 (COVID-19) had been reported, with more than half a million deaths. As the World Health Organization (WHO) declared COVID-19 a pandemic on March 11 2020, the world started preparing to live with COVID-19 as the new normal. All-encompassing lockdowns and travel bans would wreck the global economy in the long run. As a result, countries are looking for middle-ground solutions that would neither dry out national medical resources nor paralyze the economy. So what is there to be learned from China's COVID-19 experiences? Our study aimed to quantitatively evaluate the contribution of the three major types of policies implemented for successful containment of COVID-19: i) city lockdown aiming to reduce within-city contact, ii) cross-city travel restrictions, and iii) effective ways to test and isolate infected persons. Importantly, we followed the full trajectory of COVID-19 in China from January 10, two weeks before the drastic lockdown of Wuhan on January 23, to March 15, when the pandemic was contained. The long period study period enabled us to estimate rather than simulate the policy effects. We based our analysis on a Susceptible-Exposed-Infected-Recovered (SEIR) model adapted from (1) . We explicitly modelled the testing, detection and quarantine process by categorizing infections into documented and undocumented ones and imposing different parameters of 3 transmission between the two categories. A radical ramping up of testing could turn a majority of new infections into the less transmissive class and crucially slow down the spread. We estimated the model for five sub-periods from January 10 to March 15, 2020, and mapped the changes in key parameters to observed changes in different non-pharmaceutical interventions in reality. The impacts of city lockdown and intercity travel bans were modelled as reductions in within-and intercity mobility, which affects the probability of contact and disease transmission across individuals. We obtained daily real-time mobility data from Baidu Migration, a travel map offered by the largest Chinese search engine. To derive a counterfactual scenario where the restrictions on with-and intercity mobility had never been implemented, we aligned the 2019 and 2020 Baidu Migration mobility data on the basis of relative timing to the Spring Festival. For example, we assumed that without intercity travel bans, the counterfactual number of travellers between city pairs on January 23, 2020 (2 days before Spring Festival) would be the same as the observed number of travellers on February 3, 2019. Similarly, reduction in intercity mobility from the 2019 baseline level was used to estimate the effects of city lockdown on contact reductions. A first glance at the mobility data in 5 and 4 revealed significant reductions in population inflow into Hubei cities after January 22, when evidence on human-to-human transmission was revealed. The trend never recovered up to mid-March. The flow into non-Hubei cities also experienced a sudden drop around the same time, but gradually increased after the end of February, when the first wave of pandemic was under control. Similarly, the within-city mobility plunged after January 22 and stayed at a very low level for Wuhan until the end of March. The withincity mobility for cities in other provinces such as Chongqing also decreased significantly around the same period, but gradually caught up to its 2019 level by the end of March, when lockdown regulations had been lifted. 4 Our estimates revealed that the detection rate of infections grew from 11.3% [95% credible interval (CI): 9.7-13.0%] before January 23, to 92.5% [95% credible interval (CI): 85.9-94.5%] in early March, accompanied by a significant reduction in the transmission rate and the length of infectious period of confirmed patients, evidence consistent with substantial improvement in testing and treatment capabilities. A direct comparison across the three groups of control methods is presented in Table 2 and Figure 6 . We found that drastic suppression measures, such as city lockdowns, were most effective in the short run. In the counterfactual scenario had we lifted city lockdowns after January 23, the cumulative number of infections by February 29 would have been 648% of the reality. Comparatively, keeping the detection rate and transmission parameters at the baseline (before January 23) level produced an additional 69% infections. Restoring intercity travel flows to the 2019 level would lead to a threefold growth in infected cases out of Wuhan but would have limited effects on Wuhan. The three containment measures had strong complementary effects: lifting all three interventions, the number of cases would have been 65-fold higher by February 29, quite close to the estimates presented by (2) . However, as we extended the time horizon of the analysis, the cumulative effects of different control measures reversed. City lockdown by itself was not sufficient to bring the spread under control. In the counterfactual scenario with both city lockdowns and travel bans in place, leaving the detection rate at 30% would predicted 2.22 × 10 8 (95% credible interval (CI): 6.6 × 10 7 -3.2 × 10 8 ) cases by July 15, 595 times the factual number of infections, compared to 1.2 times if the detection rate was set at 70%. On the contrary, with efficient testing, tracing, and treatment, we could afford to relax restrictions on both within-city and intercity mobility. In case the detection rate at 70% could be upheld, restoring within-city and intercity mobility to 2019 level 5 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 August 24, 2020. . since the beginning of the outbreak would only lead to a 7.2-fold growth in infected cases. Perhaps the strongest policy implication emerge from our evidence is that the detection rate of infectious cases has to be higher than 50% to bring the pandemic under contril in the presence of strict city lockdown at the beginning of outbreak, and higher than 70% without. (3) estimated the detection rate of COVID-19 infections across 10 countries based on a demographic scaling model and age-specific infection fatality rates (IFRs). By mid-May, the estimated detection rate was less than 20% in Italy, 45% for the U.S. and 55% for Germany, which could explain the diverging performance in COVID-19 control across these countries. Consequently, investments in testing capacity and contact-tracing systems should be placed in a high priority to prevent ongoing secondary outbreaks of COVID-19 or similar future outbreaks of other emergent infectious diseases. We modelled the transmission of COVID-19 using a metapopulation Susceptible-Exposed-Infected-Recovered (SEIR) framework, which can flexibly generate patterns of spatial transmission (See Figure 1 for an illustration of the model structure). We traced the spatial spread of COVID-19 across cities with mobility data from Baidu Migration, a data service provided by the largest Chinese search engine. Baidu Migration collects information on population mobility from real-time location records of smartphones that use its mapping app. The platform reported bilateral migration indices for 36057 city pairs per day for 365 Chinese cities between January 12 and March 26 in 2019, and between January 1 and March 15 in 2020. It also published daily within-city mobility data for each city during the sample period. The period covers the annual "Chunyun" (Spring Festival travel season) mass migration cycle. To evaluate the role of testing, detection and post-diagnosis quarantine, we divided infec-6 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 August 24, 2020. . tions into documented and undocumented cases (I T and I N T in Figure 1 ). The two types of infections had different rates of transmission (β) and infectious period (1/λ). According to the model, extensive testing and detection (higher detection rate α) would help reduce the transmission risk of infected persons as they would be quickly isolated and treated. To capture the changes in epidemiological characteristics of the outbreak over time, especially after January 23, when serious control measures were implemented, and after Feb 08, when industries were gradually re-opened, we divided the full sample from January 1 and March 15 into five sub-periods: January 10 to January 23, January 24 to February 2, February 3 to February 12, February 13 to February 22, and February 23 to March 15. We estimated the key parameters (α, β documented , β undocumented , γ documented , γ undocumented ) separately for each sub-period, and mapped the changes to factually improvements in control measures. To better characterize the overwhelmed testing capacity at the early outbreak in Wuhan, we allow the detection rate α to differ between Wuhan and the rest cities. We also gathered data on the time gaps between illness and case report date from (4) to estimate a time-to-event delay function. For each of a new case, we drew a reporting delay from a gamma distribution. The purpose is to bridge the synthetic and observed outbreaks and improve the model fit. We estimated key parameters of the model using an iterated filter-ensemble adjustment Kalman filter (IF-EAKF) approach (1). The framework identified the maximum likelihood estimates of key parameters listed in Table 1 . The transmission rate (β) for undocumented infections was less than 22% of that for documented cases from January 10 to January 23, 2020. The ratio further dropped to 11% after February 3. The infectious period for positive cases also dropped from 1.827 days in late January to 1.156 days in early March. Both effects could be attributable 7 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 August 24, 2020. . to improvements in the treatment capacity and practices in managing confirmed patients. A notable example is the introduction of Fangcang shelter hospitals, a rapidly-constructed and low-cost medical infrastructure providing basic isolation, triage, medical care, monitoring and referral services to clinically confirmed patients. The development of Fangcang hospitals initiated on February 5, 2020 significantly reduced intra-family transmission associated with home isolation and was considered a critical move toward balancing the strained medical system in the Hubei province (5) . Meanwhile, the detection rate α proliferated from less than 1% in Wuhan before January 23 to more than 70% in early March, as shown in Figure 3 . The detection rate in other cities grew steadily from less than 10% at the onset of the pandemic to more than 90% in early March. The estimated percentage of undocumented cases at the beginning of the outbreak in Wuhan was slightly lower than the ones reported by (1) and (6) We present simulations of reported cases generated by the model in Figure 2 . The simulation matches well with the observed outbreaks for both the Hubei province and the rest of China, even though we did not target at matching the two subgroups separately. The explicit modelling of reporting activities also makes our model more flexible to account for surges in reported cases as a result of changes in case identification criteria. Estimation of epidemiological models on China in previous articles ((2),(11),(12)) usually stops before February 13. On that day, China revised the case definition in Hubei to include patients who met the clinical criteria in the absence of a positive PCR test, purportedly to clear the backlog of COVID-19 tests. To account for this outlier, we manually set α = 1 for Wuhan on February 13, 2020. As is clear from Figure 2 , the surge was well simulated in our model. This operation allowed us to extend our analysis to March 15, the last day when Baidu Migration mobility data were 8 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 August 24, 2020. . https://doi.org/10.1101/2020.08.22.20179697 doi: medRxiv preprint made available. An obvious benefit of extending the period of study is that we could look at changes in key parameters in response to new policy changes after February 13, including the opening of temporary hospitals and tentative re-opening of industries. Assuming that the control measures were kept at similar levels from March 15 to July 15, we borrowed the parameter estimates and mobility measures from the last period of our sample (February 23 to March 15) to perform an out-of-sample model validation exercise. The predicted cumulative infections on July 15 is only 17% higher than the actual ones, a sign that our model could capture the intensity of containment policies well. 9 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 August 24, 2020. . 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 August 24, 2020. . 12 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 August 24, 2020. . https://doi.org/10.1101/2020.08.22.20179697 doi: medRxiv preprint it and N it denote the susceptible, exposed, detected infected, undetected infected and total population in city i and time t.β N T t and β T t are the rate of transmission for undetected infected individuals and undetected infected individuals,respectively.α t is the testing rate in time t.Z t denotes latency period through which patients switch from exposed stage to infection stage.A it is within-city population flow in city i in time t. Spatial spread of the disease is governed by the daily number of people travelling from city j to city i in time t(M ijt ). 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 August 24, 2020. . 14 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 August 24, 2020. . 15 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 August 24, 2020. . 16 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 August 24, 2020. . 17 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 August 24, 2020. . 18 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 August 24, 2020. . 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 August 24, 2020. . 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. and N it are the susceptible, exposed, detected infected, undetected (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 August 24, 2020. . • α t is the testing rate in time t, which is defined as the ratio between the number of documented infections in time t and the sum of cumulated undetected patients carried over from last period and new patients in time t.α t = I T it /( • Z t denotes latency period through which patients switch from exposed stage to infection stage. D t is the infectious period that patients could infect the susceptible population. D T t is the infectious duration for detected infections while D N T t is for undetected infections. D t is typically lower than D N T t provided that detected infected individuals would be properly treated and become well sooner. • A it is within-city population flow in city i in time t. Spatial spread of the disease is governed by the daily number of people travelling from city j to city i in time t(M ijt ). To capture the fact that exposed and undetected individuals might travel less to other cities, likely because of the illness of family members or voluntary avoidance behavior following the reports of epidemic hot spots, we add a multiplicative factor, ζ smaller than one. We further assume that individuals in the tested group who have been admitted by local hospital do not move between cities. In this model, the effective reproduction number (R 0 ) is calculated as Parameter estimation We infer model epidemiological parameters using an iterated filtering (IF) approach ( (1)). In our model, we consider the unreported infection may be tested and reported later, so we add α t I N T it source code. We divided the full sample from January 1 and March 15 into five subperiods: January 10 to January 23, January 24 to February 2, February 3 to February 12, February 13 to February 22, and February 23 to March 15. We estimated the key parameters (α, β documented , β undocumented , γ documented , γ undocumented ) for each period. The first 23 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 August 24, 2020. . Reporting delay Cities in Hubei started to count clinically confirmed cases as confirmed COVID infections from February, 12. Patients who met clinical criteria through chest imaging but may not have had epidemiological links or a positive PCR test results were included in the official confirmed cases. 1 . As a direct consequence, the number of confirmed cases in Wuhan was more than 12 times higher than that in the previous day, creating a time-varying discrepancy in the case detection rate in Wuhan from other cities. To deal with it, we create a multiplication factor µ W uhan,t for Wuhan and the detection rate in Wuhan was defined to be µ W uhan,t * α t . To simulate the sharp increase case identification in Wuhan, we assume the reported rate of Wuhan equal to 1 on February 12. The initial exposed populationE wuhan and initial undocumented infected population, I wuhan are were from a uniform distribution [0,Seed max ]. Seed max was estimated at [1000,4000] in January 10 ((1)), and we compared the fitting results under different initial values, and found that Seed max = 3000 is the best fitting value (S6 ). We set the initial exposed population and initial undocumented infected population of other cities based on the number of travellers from Wuhan to city i on the first day of Chun- denotes the number of travelers from Wuhan to city i ) In our model, we also considered a reported delay for tested infection. In China, cases are classified as suspected before reported officially as confirmed case, before that they must be tested at least two times. Suspected cases are sent to designated hospitals and quarantined before official confirmation 2 . Therefore, reported delay refers to the time interval between a person was admitted by a hospital and the observed confirmation of that individual infected case. In reality, many cases were confirmed after multiple tests, and the supply of testing reagent was insufficient at the beginning of the pandemic for timely confirmation. We calculated the reported delay in our paper in a slightly different way from the approach employed in (1) 25 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 August 24, 2020. . after February 1 (S7). Therefore, we simulated the confirmed delay before and after February 1 We found that the time interval could be fit by the Gamma distribution(a=3.55, b=1.22, LL=1669.71) before 2.1 and a=3.86, b=0.78, LL=522.82 after 2.1) (S8). Since Hubei province included all clinically diagnosed cases to confirmed infections, we set the reported delay in Hubei equal to 0 in 2.12. Observations of Confirmed COVID-19 Cases We have compiled a city-level health outcome dataset in China for 339 cities from January 10, 2020, to March 15, 2020. From January 24, 2020, onwards, data were obtained from the public dataset Ding Xiang Yuan (DXY) that reports daily statistics across Chinese cities 5 . We used a web scraper program to obtain data from 7 . In our main analysis, we rely on inter-city migration in 2019 to simulate the counterfactual spatial transmission of COVID-19 without traffic bans. Naturally, the reduction in intercity mobility in 2020 from its 2019 level is a combination of policy effects and individuals' voluntary avoidance behaviour as a result of increased awareness. Our analysis is going to capture the composite impact of these two channels. Within-city mobility data Apart from the intercity data, Baidu also provides the daily withincity mobility data for each city in the sample period from a separate data product. The data is generated based on Baidu Map app usage within a city. We rely on this data to describe withincity mobility. Since Baidu's app may not cover all population, we compare it to the mobility data in (13) , which used nationwide mobile phone data to track population outflow from Wuhan from January 1 to January 24. The population mobility measured by mobile phone data sources is 5.5-6.5 times of the Baidu index. Therefore, We multiply inter-city mobility measures from Baidu by the multiplication factor of 6. 27 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 August 24, 2020. . Notes: This graphs show spatial distribution of cumulative infections in July 15,2020 with both lockdown and intercity travel ban implemented (inter-city and within-city mobility were set at 2020 level). Figure (a) 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 August 24, 2020. Notes: The graphs plot simulations using different initial value. The observed cases is best fitted when initial value equal to 3000 (MAE=363) than other initial values ( M AE seed=1000 = 848,M AE seed=1500 = 660, M AE seed=2000 = 489 ,M AE seed=2500 = 373 ,M AE seed=3000 = 354,M AE seed=3500 = 438 ,M AE seed=4000 = 554. 6 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 August 24, 2020. . https://doi.org/10.1101/2020.08.22.20179697 doi: medRxiv preprint Figure S7 : Reporting delay Notes: t0 is set to be January 31, 2020. We observed the change of reporting delay between January 18 and February 15 days in which hospital registration data were available. 7 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 August 24, 2020. . Notes: The graphs plot the histograms of case reporting delay for cases confirmed before and after February 1. Reporting delay is defined as the interval between the day each admitted by the hospital and the day each case was confirmed officially. The data prior to February 1 were fitted with a Gamma distribution(a=3.55, b=1.22, LL=1669.71) and data were fitted with a Gamma distribution (a=3.86, b=0.78, LL=522.82 after 2.1) after February 1,2020. 8 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 August 24, 2020. . https://doi.org/10.1101/2020.08.22.20179697 doi: medRxiv preprint 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