key: cord-0937603-fym8qze2
authors: Yang, Juan; Chen, Xinhua; Deng, Xiaowei; Chen, Zhiyuan; Gong, Hui; Yan, Han; Wu, Qianhui; Shi, Huilin; Lai, Shengjie; Ajelli, Marco; Viboud, Cecile; Yu, Hongjie
title: Disease burden and clinical severity of the first pandemic wave of COVID-19 in Wuhan, China
date: 2020-09-01
journal: medRxiv
DOI: 10.1101/2020.08.27.20183228
sha: b3261b7f087da568641d68e09190eb3b569cb186
doc_id: 937603
cord_uid: fym8qze2

The pandemic of novel coronavirus disease 2019 (COVID-19) began in Wuhan, China, where a first wave of intense community transmission was cut short by interventions. Using multiple data source, we estimated the disease burden and clinical severity of COVID-19 by age in Wuhan from December 1, 2019 to March 31, 2020. We adjusted estimates for sensitivity of laboratory assays and accounted for prospective community screenings and healthcare seeking behaviors. Rates of symptomatic cases, medical consultations, hospitalizations and deaths were estimated at 796 (95%CI: 703–977), 489 (472–509), 370 (358–384), and 36.2 (35.0–37.3) per 100,000 persons, respectively. The COVID-19 outbreak in Wuhan had higher burden than the 2009 influenza pandemic or seasonal influenza, and that clinical severity was similar to that of the 1918 influenza pandemic. Our comparison puts the COVID-19 pandemic into context and could be helpful to guide intervention strategies and preparedness for the potential resurgence of COVID-19.

As of 26 July 2020, 188 countries have been affected by the novel coronavirus disease 42 2019 , with 15,745,102 COVID-19 cases and 644,661 deaths reported 43 worldwide 1 . COVID-19 has a broad spectrum of severity. The bottom of the severity 44 pyramid includes serological-confirmed infections, of which only a fraction will 45 develop symptoms. A fraction of symptomatic cases may seek medical care, when 46 they can be identified via surveillance systems, require hospitalization and die. 47

Hospitalization is an important metric as it determines the strain exerted by an 48 epidemic on the health care system. Further, deaths are highly relevant to planning 49 pandemic response, as mortality is an outcome that health authorities typically aim to 50 minimize. (Fig.1a ) 51 52 Estimates of disease burden and clinical severity of COVID-19 are critical to identify 53 appropriate intervention strategies, plan for healthcare needs, and ensure the 54 sustainability of the health system throughout the duration of the pandemic. However, 55 quantifying these estimates based on surveillance data is challenging due to changes 56 in health seeking behaviors during the pandemic, as well as underdiagnoses. For 57 instance, the detection of a novel pathogen may give a high rate of false negatives. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https: //doi.org/10.1101 //doi.org/10. /2020 typically considered as the worst-case pandemic scenario for pandemic planning. In 62 contrast, the 2009 influenza pandemic is considered mild but provides a benchmark 63 for a pandemic in modern times, as the health systems, supportive care, and Wuhan is a particularly well-suited location to assess the health burden of Firstly, Wuhan experienced intense community transmission of severe acute 73 respiratory syndrome coronavirus 2 (SARS-CoV-2); secondly, the first wave has 74 ended, with only seven sporadic cases reported between March 24 and May 18 3 . 75 Therefore, the first epidemic wave in Wuhan (for the period December 1, 2019-March 76 31, 2020) is an opportunity to comprehensively quantify the disease burden and 77 clinical severity of COVID-19. Here we used multiple data sources to estimate age-78 specific rates of symptomatic SARS-CoV-2 infections, medically attended cases, 79 hospitalizations, and deaths, accounting for health seeking behaviors and 80 underdiagnoses. We also estimated rates of medically attended influenza-like-illness 81 (ILI) associated with SARS-CoV-2 infections; hospitalizations with severe acute 82 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

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

The copyright holder for this this version posted September 1, 2020. . https: //doi.org/10.1101 //doi.org/10. /2020 6 respiratory infection (SARI), and pneumonia hospitalizations associated with SARS-83

CoV-2 infections by dividing the number of ILI consultations, SARI hospitalizations 84 and pneumonia hospitalizations by the number of symptomatic SARS-CoV-2 85

infections. Moreover, we estimated the clinical severity of COVID-19 including the 86 symptomatic case-fatality risk (sCFR), medically attended case-fatality risk (mCFR), 87 hospitalization-fatality risk (HFR), symptomatic case-hospitalization risk (sCHR), and 88 medically attended case-hospitalization risk (mCHR). The rates of symptomatic cases, 89 medically attended cases, hospitalizations, and deaths with SARS-CoV-2 were 90 calculated by dividing the number of cases at each level of severity by population 91 size. Clinical severity was obtained by dividing the numbers of cases in the 92 corresponding severity pyramid. (Fig.1a is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint 7 symptoms, and radiographic evidence of pneumonia. Severe cases refer to cases with 104 any breathing problems, finger oxygen saturation, and low PaO2/FiO2 (PaO2 denotes 105 partial pressure of oxygen in arterial blood; FiO2 denotes fraction of inspired oxygen), 106 etc. Critical cases refer to cases having any respiratory failure, shock, and any other 107 organ failure that requires ICU admission. Clinically-diagnosed cases included 108 suspected cases with pneumonia as indicated by chest radiography, but without 109 virological confirmation of infection 6 . (Supplementary Information File 1) These 110 clinically-diagnosed cases were included in our study, recognizing the value of a 111 clinical definition at the peak of a pandemic and in the context of limited laboratory 112 testing capacity. A total of 50,333 COVID-19 cases were reported in the four-month 113 epidemic in Wuhan. Of them, 32,968 (65.5 %) were laboratory-confirmed cases. As 114 of July 20, 3,869 cases have died, and all others recovered. These cases were recorded 115 from passive surveillance which was launched at the start of the outbreak in late 116

December 2019 in Wuhan 9 , and from active door-to-door and individual-to-individual 117 screenings for fever (Supplementary Information File 2) 10, 11 . 118 119 Estimated disease burden of COVID-19 120 RT-PCR sensitivity for SARS-CoV-2 detection varies based on the interval between 121 symptom onset and laboratory testing, which was highest (97.9%) at an interval of <7 122 days 12 . A population-based telephone and online survey conducted in Wuhan found 123 that 35.4% (95%CI 28.4%-43.9%) of patients with acute respiratory infections (i.e., 124 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

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

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint 8 fever with any symptoms of cough, and/or sore throat) sought medical care during the 125 epidemic of COVID-19 13 . All cases from passive surveillance were considered as 126 medically attended cases. In the baseline analysis, we assumed that a proportion of 127 mild cases, and all moderate-to-critical cases (had radiographic evidence of 128 pneumonia) captured by active screening in the community would eventually seek 129 medical care given that the health system was not overwhelmed. It was assumed that 130 the cases from passive surveillance had the same health seeking behavior as those 131 captured by active screening in the community. Laboratory-confirmed cases 132

(moderate-to-critical) and clinically-diagnosed cases had radiographic evidence of 133 pneumonia, and thus were considered as requiring hospitalization. (Fig.1b per 100,000 individuals respectively. A consistent increasing trend with age was 143 observed across all metrics, with the highest rates occurring in adults aged 60 years 144 and over (Fig.2a, Fig.3a, Fig.4a and Supplementary Information File 3) . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. The hospitalization rates of COVID-19 in Wuhan were 3.1-fold higher than that of the 155 2009 influenza pandemic, and 1.8-2.6 times that of seasonal influenza 16, 17, 18, 19 . The overall sCFR of COVID-19 was 4.54% (95%CI 3.70-5.14%), which is 165 comparable, if not higher, than that of the 1918 influenza pandemicfrom the 166 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

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

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10. 1101 /2020 analysis of data from eight US localities, the sCFR was estimated at 1.61% and 1.98% 167 for the first and second wave, respectively 23, 24 . Such a figure is substantially higher 168 than that of the 2009 influenza pandemic (<0.1% in the US) 25 . The sCFR of COVID-169 19 was higher for adults aged ≥60 years than for the other age groups (9.09% vs. affected during seasonal influenza epidemics 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26 . 

To assess the robustness of our findings, we conducted four sensitivity analyses: In 182 scenario i) we assumed that moderate cases had the same health seeking behavior as 183 mild cases, i.e., only a proportion of moderate cases sought medical assistance; in 184 scenario ii) we excluded clinically-diagnosed cases; in scenario iii) we used the upper 185 limit of 95%CI of the probability of seeking medical care; and in scenario iv) we used 186 the lower limit of 95%CI of the probability of seeking medical care. Compared to the 187 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

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

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint baseline analysis, the mean rates of symptomatic cases for COVID-19 increased from 188 796 to 935 per 100,000 persons in scenario i) and 960 per 100,000 persons in scenario 189 iv), while rates decreased to 634 per 100,000 persons in scenario ii) and 719 per 190 100,000 persons in scenario iii). The sCFR decreased from 4.54% to 3.87% in 191 scenario i) and 3.77% in scenario iv), while it increased to 5.38% in scenario ii) and is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10. 1101 /2020 This study uses multiple sources of data to estimate different levels of the COVID-19 208 severity pyramid. We find that the mean rates of symptomatic cases, medical 209 consultations, hospitalizations and deaths were respectively 796, 489, 370, and 36.2 210 per 100,000 persons in Wuhan from December 2019 to March 2020. All burden 211 metrics increased with age, with adults ≥60 years of age most affected. Similarly, the 212 highest sCFR and HFR were found in older adults. 213 214 Our study is strengthened by adjustment for several potential biases. First, rates of 215 medical consultations were adjusted by the sensitivity of RT-PCR assays 12 . 216

Sensitivity was only 30-40% before January 23 due to delayed detection, which could 217 lead to important underdiagnoses and has not been considered in previous studies. 218

Second, we accounted for the health seeking behaviors among the Wuhan population 219 during the epidemic 13 . The probability of seeking medical treatment conditionally on 220 symptoms of acute respiratory diseases is a critical parameter to estimate the true 221 number of COVID-19 cases in community. Accordingly, our estimates of disease 222 burden may be the most accurate for Wuhan so far. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint 13 affected province of Canada. Variation in testing strategies likely contribute to the 229 difference in rate of symptomatic cases, in addition to true difference in epidemic 230 dynamics. Unlike in Wuhan, only individuals with signs or symptoms consistent with 231 COVID-19, and asymptomatic individuals with suspected exposure were 232 preferentially tested in the US. Moreover, in contrast to our study, the US and 233

Canadian estimates were not adjusted for the sensitivity of RT-PCR assays and health 234 seeking behavior, and thus may be underestimated. 235

Our estimated hospitalization rate for a four-month COVID-19 outbreak was much 237 higher than that for a three-to-six-month COVID-19 outbreak in the US and Qué bec 238 (370 vs. 47-114 per 100,000 persons) 27, 28, 29, 30 . We estimated that 76% of medically-239 attended cases were hospitalized in Wuhan, while only 18% were hospitalized in the 240 US 30 . The difference between these estimates could be explained by the potential 241 different clinical thresholds for hospitalization. We assumed that moderately ill cases 242 with radiographic evidence of pneumonia and more severe cases would be 243 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. Our estimates of sCFR (4.54% vs. 1.2-1.4%) and mCFR (7.40% vs 5.91%) for Wuhan 263 are higher than in prior modeling studies 35, 36 . This is likely explained by the addition 264 of revised statistics on cases and deaths, and a more complete dataset with no right-265 censored outcomes in our study. Large variations in mCFR were observed between 266 countries, which have not been systematically analyzed. Qualitatively, these 267 variations could be explained by differences in the sensitivity of surveillance systems 268 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

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

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10. 1101 /2020 to detect cases at different levels of the severity pyramid, differences in clinical care 269 of severe and critical patients, and in age structure and underlying conditions of the 270 population. 271

Our HFR estimate (9.77%) is higher than the estimate obtained by Wang et al. in a 273 highly censored sample in Wuhan in the very early stages of the epidemic (4.3%) 37 . 274

However, it is much lower than the 28% estimate obtained in two COVID-19-275 designated hospitals for severe COVID-19 cases in Wuhan, probably due to the 276 particularly high proportion of severe and critical patients hospitalized in these 277 facilities (64% vs. 28%) 38 . Our HFR estimate was lower than the 18.1% estimate in 278

France 39 , probably due to aforementioned loose threshold for hospital admissions in 279

China and preference of seeking care in hospitals rather than outpatient settings. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint burden and clinical severity of seasonal and pandemic influenza using Wuhan age 290 profile as a reference (Supplementary Information File 5 and File 8). Comparison of 291 severity estimates between pandemics was difficult to standardize, particularly for 292

1918 influenza pandemic 23 . The 1918 sCFR is based on data from a single US study 293 from more than a 100 years ago, at a time when awareness of viral diseases was 294 inexistent, case ascertainment and disease surveillance were limited, and definition of 295 clinical outcomes varied. Therefore, our comparison was not intended to quantify the 296 absolute value of differences, but to put the COVID-19 pandemic into perspective. 297

To put our results in perspective, it is important to stress that our COVID-19 estimates 299 refer to the first epidemic wave in Wuhan -a four-month long period. The epidemic 300 was controlled by intense interventions 4 . If the epidemic rebounds, as one would 301 expect if the infection was reintroduced in a population with low immunity, the 302 disease burden would rise. Moreover, given that the epidemic lasted only four months, 303 the stress on the healthcare system was tremendous, as severe cases and 304 hospitalizations were concentrated over a relatively short period of time. Furthermore, 305 neither seasonal nor pandemic influenza outbreaks were controlled, as vaccination 306 was either low or delayed until after the main wave had passed, and no social 307 distancing was put in place. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint Using a simple data-driven approach, we quantitatively assessed the impact of 310 COVID-19 on the healthcare sector using the local number of ILI consultations and 311 SARI/pneumonia hospitalizations in the absence of COVID-19 as a reference. The 312 number of COVID-19 hospitalizations was several folds higher than that of baseline 313 SARI hospitalizations and 25-132 folds higher than that of pneumonia 314 hospitalizations among adults ≥20 years of age. This indicates that during this time 315 period, the Wuhan healthcare system considerably exceeded surge capacity, 316

highlighting the importance and necessity of preparedness for sufficient healthcare 317 resources. Moreover, there is a winter peak of consultations and hospitalizations 318 related to respiratory diseases such as seasonal influenza and respiratory syncytial 319 virus 14, 40, 41 , which may have contributed to overwhelm the healthcare sector during 320 the first wave of the COVID-19 epidemic. 321

Our study has some limitations. Firstly, health seeking behavior maybe not constant 323 throughout the epidemic. In this survey, study participants in Wuhan were asked to 324 review their history of ARI between December 2019 and March 2020, and whether 325 they sought medical assistance for these symptoms 13 . However, since we did not 326 obtain the onset date of these symptoms, and hence we could not stratify health-327 seeking behavior by COVID-19 epidemic phase. Instead, we calculated the overall 328

proportion of ARIs cases who sought medical care during the epidemic. If the 329 distribution of onset dates of ARIs cases in our sample was skewed towards the early 330 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

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

The copyright holder for this this version posted September 1, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. triage, and treatment, and health seeking behavior over time and across locations. A 360 modelling study has revealed that containment has proved to be successful to control 361 the local COVID-19 epidemic in Wuhan. Without containment efforts, the number of 362 COVID-19 cases would have been an estimated 67-fold higher than that has been thus 363 far 46 . Therefore, our estimates in Wuhan could represent the disease burden and 364 clinical severity in a region with 1) wide-spread community transmission of SARS-365

CoV-2; 2) strict non-pharmaceutical interventions, referred as to "wartime measures" 366 in the study by Leung et al. 47 ; 3) extensive detection of all outpatients with fever 10 ; 4) 367 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https: //doi.org/10.1101 //doi.org/10. /2020 exceeded that in Wuhan by far, the pandemic in other countries is still ongoing and 373 any estimate is bound to be revised. Our estimates represent the full impact of a short 374 but intense first wave, and could be considered as benchmarks to plan intervention 375 strategies for a potential second wave of the pandemic. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint 26 Methods 454

Case definitions for laboratory-confirmed-cases were issued by the National Health 456

Commission of China, and included mild, moderate, severe, and critical cases. Cases 457 were confirmed by real-time reverse transcription polymerase chain reaction (RT-458 PCR) or by viral sequencing indicating genomes highly homologous to SARS-CoV-459 2 5, 6, 7, 8 . Clinically-diagnosed cases included suspected cases with pneumonia as 460 indicated by chest radiography, but without virological confirmation of infection 6 . 461

The "clinical" definition was only used for one week in Hubei province as laboratory 462

testing capacity was insufficient, and led to a large number of clinical cases to be 463 isolated and treated without delay. These clinically-diagnosed cases were included in 464 our study, recognizing the value of a clinical definition at the peak of a pandemic and 465 in the context of limited laboratory testing capacity. The laboratory-confirmed cases 466 include mild-to-critical cases, while the clinically-diagnosed cases include moderate-467 to-critical cases. Definitions are presented in detail in Supplementary information file 468

Our study aimed to account for underdiagnosis associated with the sensitivity of 473 laboratory assays, which is strongly dependent on the time lag between symptom 474 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

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

The copyright holder for this this version posted September 1, 2020 . . https://doi.org/10.1101 /2020 27 onset and diagnostic test 12 . The distribution of lags varied at different phases of the 475 epidemic in Wuhan due to laboratory testing capacity 4 . Accordingly, the daily number 476 of COVID-19 cases by symptom onset date was preferred to the aggregated 477 cumulative data. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020 . . https://doi.org/10.1101 /2020 All of these datasets were registered through a surveillance system, which was 496 launched to record information on COVID-19 cases in China at the start of the 497 outbreak in late December 2019 in Wuhan 9 . These data were collected from passive 498 surveillance, and active door-to-door and individual-to-individual screenings for 499 fever. The active screening was implemented twice in Wuhan on a daily basis from 500 January 24-February 10, and February 17-19 10, 11 . A total of 16,781 laboratory-501 confirmed cases were identified through active screening (D7, Details shown in 502

Supplementary information file 2) 10, 11 . 503 504

A study retrospectively analyzed the RT-PCR assays of 301 patients with 1,113 506 specimens in Wuhan, and found that RT-PCR sensitivity varied at different phases of 507 the epidemic due to the difference of interval between symptom onset and laboratory 508 testing (Pse) (Supplementary information file 9). The sensitivity of RT-PCR assays 509 was highest (97.9%) at an interval of <7 days 12 . 510 511

A population-based telephone and online survey was conducted to understand the 513 health seeking behaviors of patients suffering from acute respiratory infections (i.e., 514 fever with any symptoms of cough, and/or sore throat) during the epidemic of 515 COVID-19 in Wuhan. Of patients with acute respiratory infections, 35.4% (95%CI 516 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

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

The copyright holder for this this version posted September 1, 2020 . . https://doi.org/10.1101 /2020 28.4%-43.9%) sought medical care, by adjusting for the age structure of Wuhan 517 population. Children had a higher probability of medical attendance than adults 518 (Pmed.care) 13 . 519 520

A total of 10.7 million persons lived in Wuhan during the epidemic 52 . The age profile 522 of the Wuhan population was obtained from the China Statistic Yearbook 53 . To 523 compare the burden of COVID-19 to baseline activity of acute respiratory infections, 524

we obtained refence historical data on ILI surveillance in Hubei province and SARI 525 surveillance in Jingzhou city, Hubei province 14, 17, 21 . Additionally, we collected the 526 annual number of consultations in pediatric and internal medicine departments in 527

Hubei, and the national number of pneumonia hospitalization rates from the Chinese 528

Health Statistics Yearbook 54 . All these data were collected from publicly available 529 sources and did not contain any personal identifiable information. Summary of data 530 were presented in Supplementary information file 10. 531 532 Statistical analysis 533 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint

In the baseline analysis, we considered COVID-19 cases in Wuhan as those with 539 laboratory-confirmation or with a clinical diagnosis (for the brief period where the 540 clinical definition was in place) and tabulated data by symptom onset date. The 541 interval between symptom onset and diagnosis was obtained from data D1 4 . Then, we 542 randomly simulated 10,000 draws from a gamma distribution representing these time 543 intervals to estimate onset dates for laboratory-confirmed cases reported between 544

March 9-April 3 (data D3), and added laboratory-confirmed cases (data D5). This 545 allowed us to impute onset dates for cases that did not have this information. 546 547

All cases from passive surveillance were considered as medical attendance (data 549 D1+D3+D5, and D2). In the baseline analysis, we assumed that a proportion of mild 550 cases, and all moderate-to-critical cases captured by active screenings in the 551 community (data D7) would eventually seek medical care given that the health system 552 was not overwhelmed (Assumption 1). The health seeking behavior of mild cases was 553 assumed to be the same as aforementioned patients with acute respiratory infections 554 during the COVID-19 epidemic (Pmed.care) 13 . Hence, to estimate medically attended 555 cases, we only excluded a proportion of (1-Pmed.care) mild cases identified by 556 community screening from the total reported COVID-19 cases. Moreover, the number 557 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

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

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint of laboratory-confirmed cases from official reports (data D1+D3+D5, and D7) was 558 divided by the sensitivity of RT-PCR (Pse) to account for underdiagnoses. 559 560

In the baseline analysis, we assumed the cases from surveillance system (data 562 D1+D3+D5, and D2) had the same health seeking behavior as those captured by 563 active screenings in the community (data D7) given that the health system was not 564 overwhelmed (Assumption 1). Accordingly, the number of mild symptomatic cases 565 was estimated by dividing reported mild COVID-19 cases by the probability of 566 seeking medical care, conditionally on self-reported acute respiratory infection 13 . 567

Adjustment of sensitivity of RT-PCR was considered as well. 568 569

Moderate-to-critical COVID-19 cases had radiographic evidence of pneumonia, while 571 mild cases were defined as those without radiographic evidence of pneumonia 5, 6, 7, 8 . 572

Chest x-ray confirmed pneumonia is a threshold for hospital admissions in China. 573

Accordingly, in our study, estimates for SARS-CoV-2 related hospitalizations 574 excluded patients defined as mild cases in the baseline analysis. (Assumption 2) 575 576

In above analyses, to account for the uncertainty of two parameters (RT-PCR 577 sensitivity and probability of seeking medical care), we conducted a Monte Carlo 578 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

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

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10. 1101 /2020 Simulation by drawing 10,000 samples on the basis of Binomial distributions. We 579 generated 10,000 estimates for the number of COVID-19 cases, based on which we 580 calculated the median, and 95% CIs (the 2.5th and 97.5th percentiles) for the 581 outcomes of interest in this study. 582 583 Additionally, below sensitivity analyses were conducted: in scenario i) for above 584

Assumptions 1) and 2), we assumed moderate cases had the same health seeking 585 behavior as mild cases, i.e., only a proportion of moderate cases sought medical 586 assistance (Pmed.care); and in scenario ii) we excluded clinically-diagnosed cases. Chi-587 square tests were used to compare the estimates of baseline and sensitivity analyses. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10. 1101 /2020 Moreover, for comparison with historical outbreaks, we conducted a narrative review 599 on estimates of disease burden and clinical severity for the 1918 and 2009 influenza 600 pandemics, as well as seasonal influenza in China and USA (Summary of studies 601 shown in Supplementary Information file 4-5) . The age profile of COVID-19 cases 602 was obtained from data D1 4 , in which COVID-19 cases were broken down into 20-603

year age categories. We could not generate disease burden and clinical severity 604 estimates for influenza using the same age stratification because numerators and 605 denominators were not available from the literatures. 606 607

The funder of the study had no role in study design, data collection, data analysis, data 609 interpretation, or writing of the report. The corresponding author had full access to all 610 the data in the study and had final responsibility for the decision to submit for 611 publication. 612 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

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

The copyright holder for this this version posted September 1, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint 695 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

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

The copyright holder for this this version posted September 1, 2020. . https://doi.org/10. 1101 /2020 

Coronavirus (COVID-19) Information and Updates

Health Commission of Hubei Province. Daily report on epidemic situation of COVID-19 619 in Hubei province

Association of Public Health Interventions With the Epidemiology of the 623 COVID-19 Outbreak in Wuhan, China

Evolving epidemiology and transmission dynamics of novel coronavirus 625 disease 2019 outside Hubei Province in China: a descriptive and modeling study

National Health Commission of China. The diagnosis and treatment scheme of novel 628 coronavirus diseases 2019 (Trial version 5th)

National Health Commission of China. The diagnosis and treatment scheme of novel 632 coronavirus diseases 2019 (Trial version 6th)

National Health Commission of China. The diagnosis and treatment scheme of novel 636 coronavirus diseases 2019 (Trial version 7th)

Chinese Center for Disease Control and Prevention. Epidemic update and risk 640 assessment of 2019 Novel Coronavirus

The State Council of the People's Republic of China. Announcement of Wuhan 644 Headquarters for the Control and Treatment of Novel Pneumonia

Report on results of 3-day screening in each district of Wuhan

Dynamic Profile of RT-PCR Findings

Health seeking behaviors of patients with acute respiratory infections 655 during the outbreak of novel coronavirus disease 2019 in Wuhan

Burden of influenza-associated outpatient influenza-like illness 658 consultations in China

Incidence of medically attended influenza during pandemic and post-661 pandemic seasons through the Influenza Incidence Surveillance Project

Estimating the burden of 2009 pandemic influenza A (H1N1) in the 664

United States

The substantial hospitalization burden of influenza in central China: 666 surveillance for severe, acute respiratory infection, and influenza viruses

Burden of influenza-associated respiratory hospitalizations in the 669

Estimating influenza disease burden from population-based surveillance 671 data in the United States

Estimated global mortality associated with the first 12 months of 673 2009 pandemic influenza A H1N1 virus circulation: a modelling study

Influenza-associated excess respiratory mortality in China, 2010-15: a 676 population-based study

Estimates of global seasonal influenza-associated respiratory 678 mortality: a modelling study

The Epidemiology of Influenza

A Missed Summer Wave of the

Influenza Pandemic: Evidence From Household Surveys in the United States and 682

The severity of pandemic H1N1 influenza in the United States

Hospitalization Fatality Risk of Influenza A(H1N1)pdm09: A Systematic 686 Review and Meta-Analysis

Coronavirus Disease 2019 Case Surveillance -United States

Assessing the burden of COVID-19 in Canada

Hospitalization Rates and Characteristics of Patients Hospitalized with 692

Coronavirus Disease 2019 -COVID-NET, 14 States

321 individuals residing in California 696 and Washington, United States: a prospective cohort study

Analysis and suggestion on the factors leading to the difficulty in the 701 construction of hierarchical diagnosis system

Response Team. Preliminary Estimate of Excess Mortality During the COVID-19

Estimation of Excess Deaths Associated With the COVID-19

Pandemic in the United States

Estimating clinical severity of COVID-19 from the transmission dynamics 710 in Wuhan, China

An investigation of transmission control measures during the first 50 days 712 of the COVID-19 epidemic in China

Clinical Characteristics of 138 Hospitalized Patients With

Coronavirus-Infected Pneumonia in Wuhan, China

Clinical course and risk factors for mortality of adult inpatients with 716

COVID-19 in Wuhan, China: a retrospective cohort study

Estimating the burden of SARS-CoV-2 in France

Genetic variability of respiratory syncytial viruses (RSV) prevalent in 723

emergence of subgroup B and A RSV as 724 dominant strains

Diagnostic techniques for COVID-19 and 726 new developments

Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-728 PCR

Diagnostic performance between CT and initial real-time RT-PCR for 730 clinically suspected 2019 coronavirus disease (COVID-19) patients outside Wuhan

Rapid and sensitive detection of SARS-CoV-2 RNA using the Simplexa™ 733 COVID-19 direct assay

Effect of non-pharmaceutical interventions to contain COVID-19 in China

First-wave COVID-19 transmissibility and severity 37

in China outside Hubei after control measures, and second-wave scenario planning: a 738 modelling impact assessment

Fangcang shelter hospitals: a novel concept for responding to public 740 health emergencies

The State Council of the People's Republic of China. Press Conference of the Joint 742 Prevention and Control Mechanism of the State Council

Active case finding with case management: the key to tackling the COVID-19 745 pandemic

The Novel Coronavirus Pneumonia Emergency Response Epidemiology Team. The

Epidemiological Characteristics of an Outbreak of 2019 Novel Coronavirus Diseases 748 (COVID-19) -China

Health Commission of Hubei Province. The 20th press conference on prevention and 750 control of pneumonia caused by novel coronavirus

China Health Commission. Chinese Health Statistical Yearbook. China Peking Union 756

The R Project for Statistical Computing

. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted September 1, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted September 1, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted September 1, 2020. 

Supplementary Information is available for this paper. 785Correspondence and requests for materials should be addressed to J.Y., and H.Y. 786. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted September 1, 2020 . . https://doi.org/10.1101 /2020 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted September 1, 2020. . https://doi.org/10.1101/2020.08.27.20183228 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprintThe copyright holder for this this version posted September 1, 2020. . https://doi.org/10. 1101 /2020