key: cord-0685081-3e1ji2mw
authors: Wan, Hui; Cui, Jing-an; Yang, Guo-Jing
title: Risk estimation and prediction by modeling the transmission of the novel coronavirus (COVID-19) in mainland China excluding Hubei province
date: 2020-03-06
journal: nan
DOI: 10.1101/2020.03.01.20029629
sha: 8c6e523654a7eab291bd150a6894986f76a30b4f
doc_id: 685081
cord_uid: 3e1ji2mw

Background: In December 2019, an outbreak of coronavirus disease (COVID-19)was identified in Wuhan, China and, later on, detected in other parts of China. Our aim is to evaluate the effectiveness of the evolution of interventions and self-protection measures, estimate the risk of partial lifting control measures and predict the epidemic trend of the virus in mainland China excluding Hubei province based on the published data and a novel mathematical model. Methods: A novel COVID-19 transmission dynamic model incorporating the intervention measures implemented in China is proposed. We parameterize the model by using the Markov Chain Monte Carlo (MCMC) method and estimate the control reproduction number Rc, as well as the effective daily reproduction ratio Re(t), of the disease transmission in mainland China excluding Hubei province. Results: The estimation outcomes indicate that the control reproduction number is 3.36 (95% CI 3.20-3.64) and Re(t) has dropped below 1 since January 31st, 2020, which implies that the containment strategies implemented by the Chinese government in mainland China excluding Hubei province are indeed effective and magnificently suppressed COVID-19 transmission. Moreover, our results show that relieving personal protection too early may lead to the spread of disease for a longer time and more people would be infected, and may even cause epidemic or outbreak again. By calculating the effective reproduction ratio, we proved that the contact rate should be kept at least less than 30% of the normal level by April, 2020. Conclusions: To ensure the epidemic ending rapidly, it is necessary to maintain the current integrated restrict interventions and self-protection measures, including travel restriction, quarantine of entry, contact tracing followed by quarantine and isolation and reduction of contact, like wearing masks, etc. People should be fully aware of the real-time epidemic situation and keep sufficient personal protection until April. If all the above conditions are met, the outbreak is expected to be ended by April in mainland China apart from Hubei province.

belong to the order of Nidovirales, family of Coronaviridae, and subfamily of Orthocoronavirinae ( [1] ).

In December 2019, an outbreak of novel coronavirus disease (COVID-19) emerged in Wuhan, China and has swiftly spread to other parts of China and a number of oversea countries. By February 27th, the new virus has infected more than 78,900 people and killed at least 2791 in China ( [2] ). Besides China, more than 4440 people have been infected and 67 died in at least 48 countries and regions ( [3] ). Currently, there exist no vaccines or anti-viral treatments officially approved for the prevention or management of the diseases. The outbreaks are still on-going.

At the early stage of the outbreak, estimation of the basic reproduction number R 0 is crucial for determining the potential and severity of an outbreak, and providing precise information for designing and implementing disease outbreak responses, namely the identification of the most appropriate, evidence-based interventions, mitigation measures and the determination of the intensity of such programs in order to achieve the maximal protection of the population with the minimal interruption of social-economic activities [4, 5] .

Recently, some papers have been released as pre-prints or undergone peer-review and published to estimate R 0 and the risk of outbreak. Li et al. ([6] ) analyzed data on the first 425 confirmed cases in Wuhan and determined the epidemiologic characteristics of COVID-19. Based on their estimates, the mean incubation period was 5.2 days, and R 0 was 2.2, which is in line with the result estimated by Riou et al. ([7] ). Zhao et al. ( [8] ) assessed the unreported number of COVID-19 cases in China in the first half of January with the estimation of R 0 2.56. Considering the impact of the variations in disease reporting rate, Zhao et al. ( [9] ) modelled the epidemic curve of COVID-19 cases, in mainland China from January 10 to January 24, 2020, through the exponential growth and concluded that the mean R 0 ranged from 2.24 to 3.58 associated with 8-fold to 2-fold increase in the reporting rate. Li et al. ([10] ) conducted a mathematical modeling study using five independent methods to assess R 0 of COVID-19. Their results illustrated that R 0 dropped from 4.38 to 3.41 after the closure of Wuhan city. Over the entire epidemic period COVID-19 had a R 0 of 3.39. Moreover, Tang et al. formulated a deterministic compartmental model. Their estimations based on likelihood and model analysis showed that R 0 with control measures might be as high as 6.47 ( [5] ). Most recently, Chen et al. ( [11] ) developed a Bats-Hosts-Reservoir-People transmission network model to simulate the potential transmission from the infectious sources to human. The estimated values of R 0 were 2.30 from reservoir to person and 3.58 from person to person. We noticed that the estimations of R 0 in varied studies are different. As mentioned in [5, 12] , variability in the estimation of the basic reproduction number is a general recognized methodological issue, and standardized methods both for calculating and reporting R 0 are still missing. Furthermore, the value of R 0 may vary with key clinical parameters inferred from data which depend on the time period, quality, accuracy, and reliability. To better quantify the evolution of the interventions, Tang et al. fitted the previously proposed model in [5] to the data available until January 29th, 2020 and re-estimated the effective daily reproduction ( [13] ). There are also some literatures focusing on the prediction of COVID-19 development trend. Wang et al. formulated a complex network model and analyzed the possible time node and 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 March 6, 2020. . https://doi.org/10.1101/2020.03.01.20029629 doi: medRxiv preprint the risk impact of resumption on secondary outbreak in Wuhan and surrounding areas ( [14] ). Roosa et al. ([15] ) utilized several dynamic models to forecast the cumulative number of confirmed cases in the coming 5, 10, and 15 days in Hubei province, and the overall trajectory of the epidemic in China excluding Hubei.

With the gradual alleviation of the epidemic situation in mainland China excluding Hubei province, considering the pressure of economic operation and the needs of people's normal production and life, some places have adjusted the primary response of epidemic prevention and control to the secondary response [16] . In this situation, some critical questions need to be answered promptly. Does the reduction of the emergency response level mean that people can fully or partially relieve from self-production? When can people return back to normal life?

The aim of this work is to evaluate the effectiveness of varied interventions and selfprotection measures, estimate the risk of partial lifting control measures and predict the epidemic trend of the virus in mainland China excluding Hubei province by establishing a COVID-19 transmission model incorporating the intervention measures implemented and fitting the data obtained from the National Health Commission of China (NHCC).

Based on the clinical progression of the disease, epidemiological status of the individuals and intervention measures (including travel restriction, body temperature measurement, close contact tracing, self-isolation and protection, etc.), we propose a novel deterministic COVID-19 transmission model. We parameterize the model using data obtained until February 27th, 2020 in mainland China excluding Hubei province, and estimate the control reproduction number as well as the effective daily reproduction ratio of the disease transmission.

The population was grouped into various compartments, namely susceptible (S), exposed (E), infectious with symptoms (I), infectious but asymptomatic (A), isolated susceptible (Si), quarantined infected pending for confirmation (Q), hospitalized (H), and recovered (R). We assume that recovered individuals have immunity at least during this epidemic period. Let N (t) = S(t) + E(t) + I(t) + A(t) + R(t) be the total number of individuals in the free community. In order to fit the data, we explicitly generated additional two groups, i.e. the cumulative number of recovered R h (t) and dead cases D(t) from hospital . The total number of cumulative reported cases is set to be T (t). All the state variables are summarized in Table 1 .

⟨ Table 1 is near here ⟩ Due to the travel restriction, migrations from/to Hubei province and other regions are ignored. Birth and nature death are also neglected.

With the increasing of cumulative number of confirmed cases, the probability of contact transmission among the informed susceptible populations would certainly reduce ( [17, 18, 19] , etc.). To better quantify the varied interventions and self-protection measures, we assume the contact rate to be time-dependent c(t) = q 1 (t)c 0 , where c 0 is the initial contact rate and q 1 (t) is the intervention coefficient with respect to contact. Here we assume that q 1 (t) = e −δT (t) , which is dependent on the total number of cumulative confirmed cases T (t) and is monotone 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 March 6, 2020. . https://doi.org/10.1101/2020.03.01.20029629 doi: medRxiv preprint decreasing with T (t), so as to well reflect the impact of media coverage on people's psychology and behaviors. c(t) = c 0 for T = 0 and lim T →∞ c(T ) = 0. It should be mentioned that the contact function c(t) in [13] is also assumed to be time-dependent, but it is not dependent on state variables. Let the transmission probability be β. Thus, the incidence rate can be given by βc(t)(I + ξA) S N , where ξ is the correction factor of transmission probability with asymptomatic infectious individuals.

By body temperature measurements everywhere and diagnoses in hospitals, symptomatic infectious individuals can be detected. The detection rate is assumed to be q 2 I(t). Infected individuals in Q class can be confirmed at the rate of ηQ(t) by nucleic acid testing. Additionally, close contact tracing followed by quarantine and isolation is a critical control measure. We assume that, once a case is confirmed, q 3 individuals would be traced. Therefore, q 3 (q 2 I(t) + ηQ(t)) individuals would be traced in a unit time, which is dependent on the number of new confirmed cases q 2 I(t) + ηQ(t). We also assume that, among these traced individuals, S N fraction parts are susceptible, E N fraction parts are exposed, I N fraction parts are infectious with symptoms and A N fraction parts are infectious but asymptomatic. R N fraction parts are recovered, which are not needed to be isolated and are still in the R class.

The disease transmission flow chart is depicted in Figure 1 and other parameters are summarized in Table 2 . Based on the above assumptions, we formulate the following model to describe the transmission dynamics of COVID-19.

(1)

⟨ Figure 1 is near here ⟩ According to the concept of next generation matrix in [20] and the basic reproduction number presented in [21] , we calculate the basic reproduction number with control measures, i.e. the control reproduction number, R c , of COVID-19, which is given by 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 basic reproduction number is the average number of secondary infections due to an infective during the infectious period when everyone else in the population is susceptible [22] . It can depict the transmission risk in the early phase of disease transmission. With the spreading of the COVID-19, increasingly intensive intervention measures have been implemented and people gradually enhanced self-protection. In order to quantity the daily reproduction number and evaluate the transmission risk changing over time, inspired by Tang et al. [13] , the initial contact rate c 0 in the formula of R c is replaced by the aforementioned time-dependent contact rate c(t) to reflect the changes of intervention measures and people's behaviors. Thereby, we define

as the effective daily reproduction ratio, the average number of new infections induced by a single infected individual during the infectious period at time t. ⟨ Table 2 is near here ⟩ Data COVID-19 data excluding Hubei province were archived from NHCC from January 20th (the first day that the number of confirmed cases was reported) to February 27th, 2020, as shown in Figure 2 ([2] ). The data include the cumulative confirmed cases, the cumulative number of deaths, newly confirmed cases and the cumulative number of cured cases.

We used the Markov Chain Monte Carlo (MCMC) method with an adaptive Metropolis-Hastings (M-H) algorithm to fit the model. The algorithm is run for 110,000 iterations with a burn-in of the first 80,000 iterations, and the Geweke convergence diagnostic method is employed to assess convergence of chains. At the significance level of 5%(the critical value of z is 1.96), all parameters and initial values estimated do not reject the original hypothesis of convergence to a posterior distribution (Figure 3 ). ⟨ Figure 2 is near here ⟩

The population of mainland China excluding Hubei province is around 1,336,210,000

[23], which can be set to be the value of S(0). 21 confirmed cases were reported on January 20th, 2020 for the first time and the numbers of recovered and dead individuals were both 0. It can be assumed that nobody has been traced by contact tracing at the beginning. Therefore, we set S i (0) = Q(0) = R h (0) = D(0) = 0, and H(0) = T (0) = 21. The quarantined susceptible individuals were isolated for 14 days, thus µ = 1/14 [2] . According to [6] , the incubation period of COVID-19 is about 5.2 days, thus ϕ can be set to be 1/5.2. ⟨ Figure 3 is near here ⟩ 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 March 6, 2020. . https://doi.org/10.1101/2020.03.01.20029629 doi: medRxiv preprint Table 1 and Table 2 . By fitting the data, we estimated the control reproduction number to be 3.36 (95% CI 3.20-3.64).

Using the estimated parameter values and the number of cumulative cases T (t), the effective daily reproduction ratio R e (t) can be calculated ( Figure 5 ). The result demonstrates that R e (t) has dropped sharply from 3.34 (R e (1) = 3.34) on January 20th to 0.89 (R e (12) = 0.89, less than 1) on January 31st, 2020, which implies that the integrated control strategies implemented in mainland China excluding Hubei has successfully reduced transmission intensity and prevented the epidemic growth in a short time frame.

⟨ Figure 4 is near here ⟩ Under the current rigorous integrated control and self-protection measures, the time series of T (t) depicted in Figure 4 (a) shows that the cumulative number of confirmed cases will continue growing slowly for some duration and tend to its predicted maximum, which is 13155. Besides, although the number of hospitalized individuals has peaked on around February 12th, 2020, but it will not shrink to zero in the near future (Figure 4 (c) ). Obviously, new infections would occur as long as the infectious individuals who have not been detected exist once people start relieve self-isolation and protection. It is the number of undetected infectious individuals that determines when people's lives are able to return back to normal. Hence we should closely follow the total number of I(t) and R(t). Figure 4 (d) displays that the number of infectious individuals has been decreasing gradually since the end of January. However, it will not descent down to 1 until late March, which infers that people should be fully aware of the real-time epidemic situation and keep personal protect before April.

⟨ Figure 5 is near here ⟩ Using the estimated parameter values and the expression of the effective reproduction ratio R e (t), the threshold value of the intervention coefficient with respect to contact, q 1 , can be calculated, which is 0.3. This implies that in order to block the continuous spread of the virus, the value of q 1 must be less than 0.3 to guarantee R e (t) is below 1. In other words, the contact rate should be kept below 30% of the normal level.

To examine the impact of partial lifting control measures on disease transmission, we plot the predicted time series of the number of cumulative confirmed cases, T (t), and the number of the infectious individuals, I(t)+A(t), with different contact rates ( Figure 6 ). Assuming that the adjusting time is from March 5th, 2020, Figure 6 (a) and (b) illustrate that contact rate with 20% of the initial value c 0 will not cause the disease re-bounce. However, the epidemic period will be extended for about 40 days until early May and the cumulative number of confirmed cases will increase by around 0.5% to 13227. While if the starting time of the adjusting is postponed to March 20th, the epidemic time of disease will be extended for about one week and the cumulative number of confirmed cases will increase by only around 0.05% to 13161, compared with the scenario of no changes. Nevertheless, if the contact rate is half of the initial value, i.e. q 1 = 0.5 and R e (t) = 1.68, COVID-19 will re-bounce 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 March 6, 2020. . https://doi.org/10.1101/2020.03.01.20029629 doi: medRxiv preprint on a large scale in a short time frame, even if the starting time is postponed to March 20th.

⟨ Figure 6 is near here ⟩

In this paper, we proposed a novel COVID-19 transmission model incorporating the intervention measures implemented in China. Particularly, we adopted two novel function forms which dynamically captured the real-time endemic situation. The first one is the contact rate, c(t), which was assumed to be dependent on the cumulative number of confirmed cases to better quantify the varied interventions and self-protection measures. The other one is the contact tracing rate, which was dependent on the number of new cases. We evaluated the impact of partial lifting control measures on COVID-19 transmission. Our results show that relieve self-protection too early may lead to the spread of the virus for a longer time and more people would be infected, and may even lead to the mass transmission of the virus again. The reduction of the emergency response level does not mean that people can be off guard. At least until the end of March, life will not be able to return to normal.

In the process of recovery of production and life, we should pay attention to take protective measures to minimize the contact between people, such as wearing masks and trying to avoid crowded places to cut the risk of catching coronavirus. By calculating the effective reproduction ratio, we assert that people's contact rate should be kept below 30% of the normal level and the lower the better.

The forecasts presented are based on the assumption that there are no imported cases from Hubei and other infected regions. The WHO recently upgraded the global risk of the epidemic to 'very high'. New endemic foci outside China are formed, such as South Korea, Italy, Iran, Japan, etc. Although the current endemic situation in China is under control, population migration cross country border should be taken into consideration when we modify the updated controlling policy of COVID-19. With the development of the epidemic situation in other countries in the world, it is very important to maintain and strengthen the quarantine of entry personnel. The impact of international mobility on the transmission of COVID-19 will be studied in our future works.

We concentrated on the epidemic situation in mainland China excluding Hubei province is due to the significant differences between Wuhan, Hubei, and the rest of the country. In Wuhan, Hubei, the origin of the new emerging disease COVID-19, it took much longer time to recognize and understand the disease transmission than other regions. It requires certain time from unknown to known. Therefore, the data accuracy in Wuhan, Hubei is a major issue for parameters' calibration. In addition, the sudden large outbreak in Wuhan, Hubei exhausted all medical resources in a short time. A more targeted model considering medical resource capacity will be anticipated in the future.

In conclusion, to ensure the COVID-19 epidemic ending rapidly, it is necessary to maintain the current integrated control intervention and self-protection measures, 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 March 6, 2020. . https://doi.org/10.1101/2020.03.01.20029629 doi: medRxiv preprint including travel restriction, quarantine of entry, contact tracing followed by quarantine and isolation and reduction of contact, like wearing masks, etc. People should be fully aware of the real-time epidemic situation and keep sufficient personal protection until April. If all the above conditions are met, the outbreak is expected to be ended by April in mainland China apart from Hubei province. Additional file 2 -Sample additional file title Additional file descriptions text. 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. (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 March 6, 2020. . https://doi.org/10.1101/2020.03.01.20029629 doi: medRxiv preprint

Novel Wuhan (2019-nCoV) Coronavirus

National Health Commission of China

World Health Organization (WHO);. Accessed

Epidemic Models of Contact Tracing: Systematic Review of Transmission Studies of Severe Acute Respiratory Syndrome and Middle East Respiratory Syndrome. Computational and structural biotechnology journal

Estimation of the Transmission Risk of the 2019-nCoV and Its Implication for Public Health Interventions

Early transmission dynamics in Wuhan, China, of novel coronavirus-infected pneumonia

Pattern of early human-to-human transmission of Wuhan

Estimating the Unreported Number of Novel Coronavirus (2019-nCoV) Cases in China in the First Half of January 2020: A Data-Driven Modelling Analysis of the Early Outbreak

Preliminary estimation of the basic reproduction number of novel coronavirus (2019-nCoV) in China, from 2019 to 2020: A data-driven analysis in the early phase of the outbreak

Estimation of the epidemic properties of the 2019 novel coronavirus: A mathematical modeling study

A mathematical model for simulating the phase-based transmissibility of a novel coronavirus

The basic reproduction number (R0) of measles: a systematic review. The Lancet Infectious Diseases

An updated estimation of the risk of transmission of the novel coronavirus (2019-nCov)

When will be the resumption of work in Wuhan and its surrounding areas during COVID-19 epidemic? A data-driven network modeling analysis

Real-time forecasts of the 2019-nCoV epidemic in China from

Jiangsu Provincial Governmen of China

Media/psychological impact on multiple outbreaks of emerging infectious diseases

The impact of media on the control of infectious diseases

The impact of media coverage on the dynamics of infectious disease

On the definition and the computation of the basic reproduction ratio R 0 in models for infectious diseases in heterogeneous populations

Reproduction numbers and sub-threshold endemic equilibria for compartmental models of disease transmission

The Threshold Infection Level for Wolbachina Invasion in a Two-Sex Mosquito Population Model

The authors thank graduate students Yuanyuan Yu and Jingyuan Li for their efforts in collecting data during the study.

The authors declare that they have no competing interests.