Qianying Lin – COVID-19 outbreak in Wuhan, China: in retrospect and in prospect

– Okay, so let’s

Yes Shall we start? Yeah So good afternoon, my name is Qianying Lin, I’m gonna present this topic but first of all there’s a few things that I should tell all of you guys So during the presenting everyone except me

will have their microphone or camera turned off and if you have questions or comments on my presentation please enter it into the Q and A section at the bottom of the screen and I will answer the questions once I’m done with the slides and we get into the Q and A section A copy of the slides will be sent after the presentation, a copy of the recording with captions will be available in our website in a week or so So if you want an un-captioned version please reply to the email with the slides and just let us know The information about this webinar including what to do if there is a technical issue, links to publications et cetera can be found on the MIDAS website That is midas.umich.edu on the calender under events Yeah so if everything, like everybody’s okay so I will just start Hi everybody, good afternoon, welcome to this webinar It’s my first time doing this so yeah This topic is Coronavirus 19 outbreak in Wuhan, China because I’m from China, in retrospect and in prospect So first of all I will give a little bit background introduction of myself Right now I’m a Data Science Fellow of Michigan Institute for Data Science in University of Michigan, Ann Arbor and before I joined this position I gained my Ph.D degree in Applied Mathematics at Hong Kong Polytechnic University last year My research interests focus on mathematical epidemiology including various kind of infectious diseases including influenza, MERS-Coronavirus, HIV et cetera and also I study some human factors, the impacts on epidemics and the trends So here is the outline of this presentation First of all I will give an introduction on the Coronavirus infectious disease and we’ll build on the epidemics in Wuhan, China and also the government’s actions and then I will give a basic concept in epidemiology for better understanding of this study or our mathematical models Then I will talk about two of my previous publications on reporting rates and R naughts that will explain a little bit later and then also the impact of governmental actions and individual which is closely related to the situation right now in United States Last is the Q and A section which may be like 10 or 15 minutes So here is the simple introduction on the virus for all of you to understand I have some knowledge on the viruses It’s first confirmed in December 2019 in Wuhan, China It is the novel Coronavirus which means at the very beginning no one knows things about this, everything is unknown It’s genetically close to the other two, the previous Coronavirus, which are SARS in Hong Kong in 2003 and the Middle East Respiratory Syndrome Coronavirus basically in Saudi Arabia in 2015 This virus would cause pneumonia and the symptoms of infection is non specific It could be cough, fever, shortness of breath, diarrhea, swollen eyes et cetera So it could be transmitted through respiratory droplets through contact with other people especially with infected individuals and until now, the intermediate host is still unknown As of today there are over 81,000 confirmations and with

over 3000 deaths in China So there are ongoing major outbreaks in Italy, Iran and South Korea The outbreaks in United States right now are at the very early stage, although as of today, just like three minutes ago, there are 1990 confirmations with 12 confirmations in Michigan with 41 deaths in United States So before we go deep into the research I will give a brief review of the events and actions taken by the government in Wuhan throughout this outbreak in China So the first hospital admission in Wuhan was on December 16th although the first confirmation can be dated back to December 1st, 2019 which was not related to the seafood market that is reported to be the so called source of the outbreaks Then the supermarkets shut down on January 1st, 2020 and then the genome sequences of the novel Coronavirus released on January 12th There’s a first household human to human transmission confirmed in Guangdong Province in China Before that it was reported that the transmission between humans were limited On January 23rd the Wuhan government shut down the public transportation and then they shut down the whole province The Hubai Province publication and also the government banned all cars in downtown On February 11th WHO officially named the disease caused by the novel coronavirus as Coronavirus Infectious Disease 2019 So the reason why I go through this review of the events and actions because one of my focus in the studies was to model the trends of the outbreaks incorporating some human factors including individual response to the outbreaks and also governmental actions So here are some basic concepts in epidemiology in case someone may be unfamiliar with them These are very crucial concepts Let’s go to the primary case first The whole duration of infections, which is from the start of the infection and so called the end of the infection can be divided into different periods So the first category is we divide into latent period which is from the start of the infection to the start of infectiousness which means this patient has the ability to spread the disease to infect other people after that Then from the start of the infectiousness to the end of infectiousness that is so called the infectious period So during this period, this primary case can infect other people which can be the secondary case By the end of the infectiousness then it’s the end of infection The second way to divide this whole duration of infection is from the start of infection to the start of onset of the symptoms which is called the incubation period So by comparing the latent period and the incubation period we can see that these individuals can spread the disease even before the symptoms appear So from the onset of the symptoms

we have the end of the symptoms and then the end of the infection So there are two concepts defined by these periods One is called the generation time which is the time intervals between the primary infection and the secondary infection And another one, which is equal to the generation time in value is called the critical onset serial interval which you can find everywhere It is defined as the onset of the symptoms in the primary case to the onset of the symptoms in the secondary case So these two concepts, these two time intervals are equal in value but obviously the serial interval could be much more observable and then from previous literature the median incubation period of Coronavirus 2019 is about 5.1 days Reported by the news it could be as long as 14 days That’s why people will suggest 14 days self isolation when you contacted with infected individuals So another important or maybe well known concept is called R naught or R zero, like the basic reproduction number So it has the expected number of secondary cases infected by one primary infection in the pure susceptible population What is a pure susceptible population? Which means that these people are healthy and immune not protected by vaccine or maybe by other antibodies So when these properties, when R naught is less than one the disease will die out in the long run Otherwise it will probably spread across the population So this is a very critical property to evaluate the contagiousness and the severity, contagiousness of the infectious diseases and the severity of the epidemics So here I share some information on the severe intervals and R naughts of other diseases So, for example, measles is high contagious and very severe disease which has severe intervals of 10 to 13 days and R naught of 12 to 18 MERS in human, I mean in humans, has severe intervals of six to 7.8 days and R naught of 0.3 to 0.8 I emphasize humans because it’s reported that there’s very limited transmission between humans The outbreak was due to the zoonotics, that is the camel to human transmissions So one thing that I should emphasize is the influenza, like the Spanish pandemic in 1918 has severe intervals of two to four days and R naught two to three So please remember these two quantities I will compare it with the coronaviruses because they are both pandemics right now My previous publication, my first publication mainly focused on estimating R naughts, the basic reproduction number of Coronavirus Disease 2019 in Wuhan So I highlight the key components that we use to estimate the R naught which is first the growth rate We assume that the cases increase exponentially when one individual could infect more than two the cases grows exponentially and then that is the severe interval that I mentioned before because it differs from people to people

so we assume that it follows the distribution and the third one is the reporting rate which is a pretty crucial factor here both in modeling the epidemics or estimating the R naught That is a ratio of reported confirmations over the actual infections So when reporting rate is 100%, which means it’s perfectly correct in reported new cases, but when it’s lower than 100% there’s some under reported cases that is like potential cases that can spread the disease or over reported which means that it is larger than 100% It can be due to the previous under reported cases So we use this formula to calculate, to estimate the R naught So here are some results So I showed this figure in order to mention that the reporting rate is pretty crucial in estimating the basic reproduction number So actually we assume different reporting ratios increments and these two are the extreme cases here So first of all from the first row we assume 100% reporting rate, all the time there is no under reporting, no over reporting, every cases are correctly reported So here is the figure in the middle is the daily new cases Blue dot is the so called adjusted cases Green circle is the reported cases They match perfectly and we can use the curve and the formula I mentioned before to estimate the basic reproduction number as 5.71 with 95% confidence interval as 4.24 to 7.54 Then in the second row we assume an eight fold increment which means that at the beginning of the outbreaks there is only 12.5% reporting rate and then it increase gradually to the end to peak at 100% on January 21st So from the middle figures we can see that daily new cases, the first one is about 41 cases It’s under reported, the actual number here is eight times of the reported cases which is over 300 here Then this one is zero cases and from here we increase the reporting rate and then on January 21st it’s 100% So here is the figure of cumulative cases then using this increment of reporting rate So we have these blue dots here We use these so called adjusted actual infections to estimate the basic reproduction number so we can estimate the R naught at 2.24 with a 95% confidence interval 1.96 to 2.55 So these two examples show that the reporting rate is very, very important in estimating the basic reproduction number which is critical property on evaluating the contagiousness and severity of the epidemics So as we all know that at the beginning, because nobody is familiar with this virus, everybody knew nothing about the virus, due to various factors, for example lack of the toolkits to testing results or like people are not aware of these diseases so there will be many under reporting here But when the public aware of this disease

the reporting rate will increase So the reporting rate is not a constant It varies through time So here is the main conclusion here The changing reporting rate over time had a great impact on estimation of the basic reproduction number, R naught as I showed before So the under reported could make an over estimation of the basic reproduction number Also the estimation of R naught dropped from 5.7 to 2.24 when we adopt different ranges of increment in reporting rates So one of our publications actually estimates the reporting rate can be as low as 5% at the beginning and with that result we estimate the R naught at 2.56 with a 95% confidence interval of 2.49 to 2.63 So next we should talk about the relations between the Coronavirus Disease 2019 and the 1928 influenza pandemic So they share a lot of similarities here For example, they have a very similar R naught which is two to three, suggested by the WHO actually They have a relatively short mean serial interval For example, Coronavirus have around four to five days and influenza have 3.6 days So here I highlighted the four to five days If you guys remember I mentioned that the mean incubation period of the Coronavirus is about 5.1 days, so a relatively short serial interval with a relatively long incubation period could actually imply asymptomatic transmission, that is like a patient can infect other people even before the symptoms appears which is a pretty crucial property of these viruses here They also have shared relatively low fatality rates, it’s around 2% and also they have a significant proportion of death due to pneumonia after infection So with all these similarities between these two pandemics we can refer some properties of the 1918 Spanish pandemic to model the dynamics right now in Wuhan, China or maybe around the world So first I will introduce the epidemiological compartmental models So basically in here, basically this model is like divide the whole population into different classes which is different stages of infection and then we assume the rates of transition between these classes or interactions between these classes and we modeled these transitions, the size of these classes in ordinary differential equations So by these definitions it seems that it’s very, very flexible because we can divide the classes as detailed as we want as long as we have the rates of transmission between them So it could be very complicated and very realistic to the data And then let me introduce the conceptual model for the Coronavirus 2019 outbreak in Wuhan that we used in our paper just published There’s a few assumptions we make for the models for better fitting, or not better fitting, for more realistic and also

much easier to understand So first we have the zoonotic, that is the animal to human transmission at the first month because we have about 41 cases at the very beginning So then we assume these first 41 cases are due to animal to human infections and after that all infections were caused by human to human transmission And then because the outbreak is during the period of Chinese New Year, so everybody is rushing to go back to their hometown, so there would be around five million people left Wuhan from December 31st 2019 to January 22nd 2020 So the whole population in Wuhan City is 14 million which means that the outbreak after the city was locked down, there are only nine million people stayed at the city So also we assume that 10% of the population are not susceptible to the Coronavirus because from previous reports the infected rate among children, especially among children under 15 is pretty low so we assume that children is not that susceptible to the virus than adults or than elderly people So here two components are not included here due to limited knowledge For example, we did not include asymptomatic transmission because right now we do not know how large is the proportion of the population could have the asymptomatic transmission and also we did not include the temperature because the impact of the temperature on the outbreak was still unknown Here is a detailed explanation on the model I know this is a little bit advanced and too much information but it’s very crucial so please pay a little bit attention to this So as I previously mentioned the compartmental models divided the whole population into different classes or different stages For example in this model it’s a very basic and classic SEIR model framework For example this population was divided into susceptible, S, E is the exposed or latent which is infected but not infectious, and I is infectious That portion of population can have the ability to spread the disease, and R is for recovered or removed Sometimes just eliminated from the population by isolated, quarantined et cetera So F here, as I mentioned, we included zoonotic introduction the previous 41 cases and then D here represents the daily number of severe cases It also represents the perception of risk regarding the severity of the epidemic which play a crucial role in human response, human behavior or response towards the diseases Because when you find out the number is pretty high, for example the number of deaths is pretty high, you could be very cautious about your daily life For example, frequent hand washing and maybe keep some distance when contacting with people and C here is the number of all infections including the reported or unreported

That’s how we include the changing under reporting ratios here For beta 0 is the initial transmission between S and I, like before the government’s action, before everybody know the virus and had some response to it and beta (t) is the time varying transmission rate It’s changing due to the governmental action, due to the individual response I will explain that in detail in the next slide how we model this change in transmission rate So other parameters it’s like the inverse sigma is the latent period in days and inverse gamma actually is the infectious period in days and also because five million people left Wuhan before the lockdown of the city so we have the emigration rate here and we have the rate of severe cases and decay of perception which means that people is really easy to forgot about the severity of the diseases, so just lose their cautions I guess So in this slide I will explain how we modeled the government action and individual reaction and incorporate them into the time varying transmission N is the total population in Wuhan It starts from 14 million and then dropped to nine million after the lockdown of the city and then D, as I explained previously, is the daily number of severe cases, also represent the perception of risk on the virus Alpha here, we call it the strength of governmental action and kappa here represent the intensity of individual reaction So these three components here build up the whole time varying transmission So one minus alpha, which includes the reduction of transmission rates by governmental action, for example governmental action including transportation shut down and quarantine and also hospitalizations, isolation et cetera These personal, individual reaction includes hand washing, decreasing contacts with people for example also represent a reduction of transmission rate by personal reaction to the proportion of the severe cases So right now we go into the finding of this paper So from figure (a) we have the daily new infections with a reporting delay which we assumed there’s 14 days of reporting delay Here please pay attention to the Y axis here as in lock 10 so the axis is a little bit not realistic but I will explain why So there we have the gray dotted line here represents the reported cases So we have assumed three scenarios here One is called naive which means the government did nothing to control the diseases as well as the individuals did nothing to prevent infections So you see that if everybody did nothing it will go up as a peak over 100,000 cases, one day That is the daily new infections So the second scenario here is we only include the individual reaction, which means only depends on individuals to wash their hands, cancel any gathering, decrease contacts with other people

So you become about 10,000 a day here and like keep for a long time Once we include both factors which is the individual reaction and also the governmental action here, which is the green line, you can see it’s very effective control here, drop vastly and by the end of April the epidemic will be totally controlled So all the parameters are referenced from a previous paper on 1918 influenza epidemic and other official news releases So in figure (b) we also calculated the reporting rate here So reporting rate means the green line, the number in gray dotted line over the number in green line here So at the very beginning, early January here, the reporting rate is very low and then it jump a little bit to about 50% in early February So it shows that the reporting rate is exactly time changing and some of you may ask how about like this over 200% reporting cases here? Because it’s due to the under reported cases in the early stage so the government just assigned examination of the previous cases so that’s why it had the overestimated condition here We also did some sensitivities on the governmental actions which is alpha, the trends of the governmental action and also the individual reactions controlled by kappa From figure (a) also is daily new infections with a reporting delay in lock 10 We use different strength here from weak to strong You may see that is a very, very effective way to control or maybe end the outbreaks with a strong governmental action, for example shut down the transportation, shut down the school et cetera Then in the second figure here, figure (b) we also use a weak intensity of individual response to a strong intensity here and we find now that it helps to control and you have to accelerate the control, the end of the outbreaks under a very strong governmental actions So the second part conclusion here is the outbreak in Wuhan would be completely controlled by the end of April under current policies and restrictions Second, the reporting rate is time-varying At the beginning, in early January, it was below 10% and then increased to around 50% in early February Individual caution in daily life helps to reduce the transmission and accelerates the end of the outbreaks Governmental actions, for example holiday extension, travel restriction, quarantine et cetera are very effective means to control and end the outbreaks So last part right now some discussion here So since United States is actually at very, very beginning of the outbreak United States is actually here, at the very beginning of the outbreak by comparing to stages of other countries For example China is at the end, almost end the outbreaks here South Korea, Japan and Singapore and United States is only at the beginning stages So should the government of United States do the same thing? What kind of factors we should consider

to make political decisions? In my opinion there are some factors including population density, age structure because I mentioned that young people or maybe young children are less susceptible to the disease than older people So also the transportation patterns because the outbreaks in Wuhan, Wuhan is actually the central of the high rate train in China so hundreds of thousands of travelers, passengers would go to Wuhan and transfer to other cities every day Also the individual reaction here For example in South Korea, once the government announced the outbreaks everybody just like on the streets, nobody on the street and everybody was wearing masks and doing very frequent hand washing So these factors we should consider but I should emphasize that governmental action would be very, very effective tool, effective means to control and end the outbreaks So here is the references that I mentioned or that I used for the parameters Okay thank you for you guys so here we can have the Q and A section or if you want further questions in detailed discussion you can just email me after the presentation Thank you guys for your coming – [James] All right, and this is James Walsh here, I’m administrative support here at MIDAS I’m gonna be helping with the Q and A So we’re just gonna take a second to read through some of the questions that have come through during the presentation and then I will read those out loud and try to answer as many as we can in the time that we have left So we’ll start with the first one that came through So first question we had today was, why would we not allow young, healthy individuals to interact with one another in the community in order to allow for transmission of Covid-19 within this low risk population which would then generate herd immunity, therefore protecting community at large? – Okay so the question here is we do not really understand the mechanics of the disease so it would be very risky to just allow so called young individuals because sometimes young individuals can have very severe symptoms after infected So I would say it’s very risky It would be very risky here to just allow them to contact, yeah – [James] More that have come through here – Yeah this one – This one? All right, so next question Is the mortality rate of Covid-19 likely substantially different than statistic being published by the World Health Organization and the Center for Disease Control? For example data from South Korea where population based testing is being conducted appear quite different from data in places where just symptomatic individuals are being tested – Yeah I would say of course because for a country who only tested symptomatic infections, which means in my opinion there is a lot of under reported rates here You do not know these asymptomatic and actually for the asymptomatic transmission we do not know whether these individuals are asymptomatic or presymptomatic which means that the symptoms does not appear yet and maybe appear later So the fatality rate would be a huge difference and also, you know, South Korea government did a really, really great job

in controlling the outbreaks So it’s a different story in South Korea because the outbreak in South Korea is due to a super infectious event A patient that belongs to a religious group they just wandering around and also the group members wandering around and there is major outbreaks in one of the cities with a relatively dense population After that the South Korea government immediately locked down the religious facilities and also have a very strict testing or examination or quarantine to everybody and the public in South Korea very cautious about the outbreak so they did a very good controlling measures to stop or to end or maybe to stop the disease Actually South Korea is almost at the end of the outbreak – [James] Answering some of the questions that we can (laughs) handle in here so we’ll look through some of the next ones coming through – No, no so skip that (James mumbles) – [James] There for you – Take this – Okay So what is the risk of reemergence of epidemic in Wuhan after a complete control has been achieved by the end of April? – So I would say the reemergence actually is like most of the cases in China or maybe in Wuhan right now are actually imported cases from other countries, for example from Italy or maybe from Iran So the risk of reemergence would totally depend on the testing or the quarantine at the airport So right now I would say the risk is very low because Chinese government did a pretty great job on surveillance or testing the incoming travelers right now (James laughs) Answer, yes – [James] So, from Aaron King You effectively used the data on the time course of the epidemic in Wuhan to infer the effectiveness of the government and popular response With more data on the spread of the disease to other localities within China can you test these inferences? – Yes, I mean with more data including, I would say including the transportation data like how many passengers go and leave Wuhan every day to other locations in China and the population densities and also because the government action varies in different location we could test it and use statistical inference methods to test this data and also compare the strands or compare intensities of the strands of government actions and also the intensities on individual response in different locations within China or even like in other countries Okay so yes

– [James] Good, so from Paul Franco We have many thanks for this important discussion Would you say that the serial interval and the R naught are very similar to the flu of 1918? Is the main difference the higher fatality rate of Covid-19? – Yes I would say that but another major difference between pandemic 1918 and the 2019 is the Covid has substantial or maybe some proportion of asymptomatic transmission which we are not clear about that part and also in accompanying the incubation period could be as long as 14 days So that’s why, because due to this asymptomatic transmission here that’s why I would say the prompt and also in time measure taken by the government would be very crucial and critical to decrease the number of infection or even stop or even end these outbreaks The fatality rate actually depends on many other factors because as far as I know the fatality rate in South Korea is very, very low It’s under 1% but in China is about 2% and also in United States it’s about 3% because the major outbreak in Seattle is around elderly people So yeah I don’t know actually (laughs) – [James] Think it looks like our Q and A has slowed down just a little bit here If we weren’t able to get to your questions either they’re just kinda out of the scope of what we’re working with or there may just not have been enough time to prepare an answer So I think at this point we will wrap things up – Thank you all for coming to this webinar Thank you so much