Skip to Main Content

Biostatistics Seminar - 6.2.2020

June 08, 2020
Biostatistics Seminar - 6.2.2020
  • 00:00- Hi everyone, welcome to the sixth seminar
  • 00:04of our seminar series on COVID-19
  • 00:06organized by the Department of Biostatistics
  • 00:08at Yale University.
  • 00:10I'm very pleased to have here today Nicholas Christakis.
  • 00:14He's a senior professor of social and medical science
  • 00:19at Yale university.
  • 00:21He's very well know for his research on social networks
  • 00:26and his recent work focuses on how essentially human biology
  • 00:31and health affect and are affected
  • 00:36by social interaction, social networks.
  • 00:39So today, he's gonna talk a little bit
  • 00:42about the epidemiology of COVID-19.
  • 00:45He's gonna give us an overview and updates
  • 00:48and then he's gonna talk about his recent paper
  • 00:52just published in Nature
  • 00:54on how to use mobility data and population overflow
  • 00:59from Wuhan to predict the spread of the COVID-19
  • 01:03in all the areas of China.
  • 01:05And then finally he's gonna talk about the new shining app
  • 01:08called Hunala which is gonna use network science
  • 01:14to essentially develop network sensors
  • 01:18to four epidemic forecasting.
  • 01:23So Nicholas is gonna take question anytime.
  • 01:26So you're welcome to write questions in the chat box.
  • 01:30And I will try to monitor it and read them to him.
  • 01:34Or you can just unmute yourself and ask questions any time.
  • 01:39- Well raise your hands,
  • 01:41I can monitor the participant list
  • 01:43for raised hands electronically raised.
  • 01:46That's easy for me.
  • 01:47- All right.
  • 01:48So Nicholas, thank you for participating
  • 01:52and why don't you take it from here.
  • 01:55- Thank you Laura.
  • 01:56Thank you so much.
  • 01:57I see many names that I recognize
  • 01:59on this big panel in front of me.
  • 02:02I'm gonna talk without slides
  • 02:03because I find it very stressful
  • 02:05and weird to use slides on Zoom.
  • 02:08I find Zoom like probably many of you
  • 02:10do pretty weird already.
  • 02:12It's so disembodied and taxing in some ways.
  • 02:15So I'm just gonna tell you a little bit
  • 02:17about the epidemiology of Coronavirus
  • 02:19as it has come to be known by many people around the world
  • 02:22in the last few months since the epidemic started.
  • 02:26And some of these things may be very simple or known to you,
  • 02:30others will not be perhaps known to you,
  • 02:32I hope I will tell you some things you don't know.
  • 02:34And I'm happy to take questions at any time.
  • 02:36And then towards the end,
  • 02:37I'm gonna tell you a little bit about some of the projects
  • 02:40in my lab that have raised a number,
  • 02:42are raising a number of difficult statistical questions
  • 02:45that we are absolutely eager
  • 02:46to collaborate with people about.
  • 02:49And Laura has been interacting with us now
  • 02:52for quite a number of years
  • 02:53as I've certain others of you
  • 02:55that I can see on this list.
  • 02:57So we are experiencing something very unusual
  • 03:01in our species that happens from time to time,
  • 03:04which is the introduction of a new pathogen.
  • 03:07We happen to be alive at a moment
  • 03:09when a new germ is entering our species
  • 03:12and having what is known as an ecological release.
  • 03:16It's just like when the rats were first introduced
  • 03:18to New Zealand, and they found you know, Terra Incognita,
  • 03:22and could just take over and do whatever they wanted.
  • 03:25This virus spent decades evolving in barks,
  • 03:28probably spent some time in pangolins
  • 03:30that's still being worked out.
  • 03:32And then in an unseen way,
  • 03:34almost surely in October or November in Wuhan China,
  • 03:37leapt into human beings and gradually spread among them
  • 03:41and then spread around the world.
  • 03:44This pathogen SARS-CoV-2
  • 03:47bears a strong similarity to other pathogens
  • 03:50that have been long circulating in bats.
  • 03:52And it's the seventh such Coronavirus species
  • 03:57that afflicts us.
  • 03:58There are four species of Coronavirus
  • 04:01that just cause the common cold,
  • 04:03they cause about 20 or 30% of the common cold
  • 04:06that people get.
  • 04:07The other viruses that cause the common cold
  • 04:09are other species of viruses.
  • 04:11And two of these Corona viruses also came from bats.
  • 04:15In addition, there are two prior Coronavirus,
  • 04:18a serious Corona viruses that have afflicted us,
  • 04:22what is the so called SARS-1 that was pandemic in 2003.
  • 04:27It was a kind of limited pandemic, which I think,
  • 04:30on the one hand, gave certain Asian countries
  • 04:33a taste of what could happen so they prepared well.
  • 04:37But on the other hand,
  • 04:39because the pandemic petered out,
  • 04:40it kind of lulled the rest of the world
  • 04:42into a false sense of security.
  • 04:44And then the seventh, before the current SARS-CoV-2,
  • 04:49the seventh Corona virus
  • 04:50is something called Middle Eastern Respiratory Syndrome,
  • 04:53or MERS, which is a virus that has a R naught
  • 04:57which we'll talk about in a moment
  • 04:58of less than one we think.
  • 05:00Each infection yields about .9 new infections.
  • 05:05So that epidemic self extinguishes,
  • 05:07which is one reason
  • 05:08that MERS has not become as serious as SARS-2 has become.
  • 05:14Anyway this SARS-CoV-2 leapt into humans
  • 05:17sometime in November, started causing cases in December.
  • 05:22And by the middle of January,
  • 05:23the Chinese knew that it was extremely serious.
  • 05:28I was contacted by some colleagues in China and Hong Kong
  • 05:31on January 23 or 24th,
  • 05:33about the possibility of collaborating to do some work.
  • 05:36We had been working for a long time,
  • 05:38using phone data from China to look at the impact of things
  • 05:42like earthquakes on shaping people's social interactions,
  • 05:46or the building of high speed rail lines.
  • 05:49So we had a well established collaboration
  • 05:51and well established procedures for handling data.
  • 05:54And they contacted me
  • 05:55and we decided to study the impact
  • 05:58or something to do with phone data
  • 06:00and the pandemic.
  • 06:02And so we began working in earnest on the 24th of January,
  • 06:07and all the one 58,
  • 06:08it reminded me of when I was a graduate student,
  • 06:10because we worked non stop for three weeks,
  • 06:13it was very exciting.
  • 06:14And of course, because they were on the other side
  • 06:16of the world, you know,
  • 06:17I would work during the day and then hand it to them,
  • 06:19and then they wake up
  • 06:20and they work during their day while I slept.
  • 06:22And then it would come back to me.
  • 06:24When we submitted the paper on February the 18th,
  • 06:26and it was ultimately published two months later
  • 06:29in the middle of April.
  • 06:31That's the paper that Laura mentioned.
  • 06:32And in this paper, what we did is is we had phone data
  • 06:35on 11 and a half million transits through Wuhan.
  • 06:39We could track people as they transited through Wuhan
  • 06:42and spread out around the country.
  • 06:46And we had the misfortune as a species,
  • 06:49that this epidemic left to us at a moment in time
  • 06:53and in a place where the one of the largest I think,
  • 06:57annual human migration takes place.
  • 06:59During that annual Harvest Moon Festival in China,
  • 07:04the new year festival.
  • 07:06So there are 3 billion translocations of people
  • 07:11that take place in China in the run up to this holiday,
  • 07:15which was on January 24th or 25th this year.
  • 07:19So the virus steps into us at a time
  • 07:22when people are spreading out,
  • 07:23and millions of Chinese moved throughout the country,
  • 07:26including transiting through Wuhan.
  • 07:28And unbeknownst to them, they carry the virus with them.
  • 07:31And what we were able to do
  • 07:32is using simply the movement of people,
  • 07:35track with phone data, the aggregate number of people
  • 07:38that left Wuhan between January 1st and January 24th,
  • 07:42and spread out to the other 296 prefectures of China.
  • 07:46By tracking the number of people who left
  • 07:49and carry the germ with them,
  • 07:50we were able to build a model that allowed us
  • 07:52to predict the timing, intensity
  • 07:55and location of the epidemic up through late February.
  • 08:00And this model we believe,
  • 08:02and I'll return to a little bit later,
  • 08:03this model we believe could be useful
  • 08:06in other sorts of situations
  • 08:08in which there is a risk source or risk sources
  • 08:11that one is trying to assess in terms of its impact
  • 08:16on this spreading of an epidemic,
  • 08:18especially if there's data available.
  • 08:20And I'll come back to that later.
  • 08:24So of course, these people in China in Wuhan,
  • 08:26then of course spread out throughout the world.
  • 08:28The Chinese were criticized for closing
  • 08:31their internal borders by January the 25th,
  • 08:36the Chinese had imposed stay at home orders
  • 08:39on prefectures in China,
  • 08:41that encompassed past 930 million people.
  • 08:46So beginning on January 25, nearly a billion people
  • 08:50were under some form of home isolation.
  • 08:52And this really got my attention,
  • 08:54because the Chinese had judged
  • 08:56that in order to combat this pathogen,
  • 08:58that the enemy that they were facing
  • 09:00and the virus required them to basically detonate
  • 09:04a social nuclear weapon.
  • 09:06This was how strong they rightly in my view
  • 09:09felt that the epidemic was.
  • 09:12So they close down their own country,
  • 09:15but they lagged a little in closing down travel
  • 09:17and leaving Wuhan.
  • 09:19Some people have have said
  • 09:20that there was some conspiracy to do that.
  • 09:22I see no evidence of that.
  • 09:24I just think they were scrambling to cope with a pandemic,
  • 09:26they closed internal travel
  • 09:28but didn't close external travel till a week or so later.
  • 09:32And without of course the germ you know,
  • 09:34spread around the world.
  • 09:35Although it would have spread no matter what.
  • 09:38It's in the nature of these pathogens
  • 09:40once they take root.
  • 09:42There's really no stopping them
  • 09:45as I alluded to earlier.
  • 09:49So the Chinese quarantined Wuhan
  • 09:52and then Hubei province surrounded home
  • 09:54to 58 million people on January the 24th.
  • 09:59Now, the first paper about this pathogen,
  • 10:02regarding the first 41 cases
  • 10:04appeared in The Lancet on the same date
  • 10:06around January the 24th.
  • 10:09And that very first paper noted the extreme likelihood
  • 10:12of interpersonal spread and the severity of the infection.
  • 10:15So the nature of what we were confronting
  • 10:18was well understood by scientists early in January.
  • 10:23I don't think we can claim that we had no idea
  • 10:27there was interpersonal spread,
  • 10:28or that it was serious.
  • 10:30And the virus we now know from genetic studies
  • 10:34arrived in Seattle already by the middle of January.
  • 10:37And this is one of the reasons that border closures
  • 10:39are so ineffective as other scientists
  • 10:42including Neil Ferguson's group, and Mark Lipschitz's group
  • 10:46have also looked at
  • 10:48is that by the time you're aware of what's happening,
  • 10:51and you try to close the borders, it's too late.
  • 10:54The pathogen has spread, you know, surreptitiously
  • 10:57and cross the borders.
  • 10:59And in fact, it had arrived in Seattle
  • 11:01by the middle of January,
  • 11:02and via Italy, in New York City by the middle of February.
  • 11:07And by then, after that point,
  • 11:10most of the cases throughout the rest of the United States
  • 11:13actually were seated from internal cases.
  • 11:17And eventually, community transmission took over
  • 11:20at importation, whether from abroad or from other states
  • 11:23became a progressively tinier fraction of the size
  • 11:26of operates at any particular location.
  • 11:29This is, again, typical of what happens with epidemics.
  • 11:33You know, some cases move in, epidemic starts,
  • 11:36and then it just takes off
  • 11:37and it's not doesn't matter
  • 11:39how many more people come in to a location.
  • 11:44And of course, it's spread into other countries
  • 11:46around the world as well.
  • 11:48Now right from the beginning,
  • 11:49there was a lot of effort
  • 11:51to estimate key epidemiological parameters
  • 11:54about this pathogen.
  • 11:55And I suspect everyone in this group knows about this,
  • 11:58but I'll just review quickly what's known,
  • 12:00and then highlight one other interesting parameter
  • 12:02that not as many people pay attention on
  • 12:05but I know this group will be interested in.
  • 12:08So the so called R naught, the R0
  • 12:11is the number of new cases
  • 12:13in a fully susceptible non-immune
  • 12:16normally interacting, typical population.
  • 12:20That's an attempt to measure something intrinsic
  • 12:22about the virus in a kind of typical human population
  • 12:26where no one is immune,
  • 12:27the virus is brand new to us,
  • 12:30people are interacting normally,
  • 12:31we haven't yet taken any protective action.
  • 12:34So this is known as the R naught.
  • 12:37And for this germ, for SARS-CoV-2,
  • 12:39it's probably around two and a half,
  • 12:42and it could be as high as three.
  • 12:45This is high, actually.
  • 12:47The seasonal flu has an R naught of about 1.3 to 1.6.
  • 12:53Chickenpox has an R naught of 3.5 to 6.
  • 12:56Ebola has an R naught of 1.5 to 1.9.
  • 13:01And of course the champion pathogen is measles,
  • 13:04which has an R naught of 18,
  • 13:06which is why vaccination rates for measles
  • 13:09have to be so high,
  • 13:11because the pathogen is so infectious,
  • 13:13there's a relationship between the amount of the pathogen
  • 13:16and the required vaccination rate to stop it,
  • 13:19before you could get herd immunity for example.
  • 13:23Now, the so called Re, R sub e
  • 13:26is the effective reproductive rate.
  • 13:28This is the number of new cases as the epidemic proceeds,
  • 13:32and as immunity rises, or as people take action.
  • 13:36And this number can fall and change.
  • 13:38So for example, if we all of a sudden became hermits,
  • 13:42nobody interacted with anyone else,
  • 13:44the Re would fall below one.
  • 13:46This is actually what happened to China.
  • 13:48They were able to track the Re
  • 13:50and find that whereas it started
  • 13:52at around 3 in Wuhan in January.
  • 13:54After their national lockdowns,
  • 13:56it fell to about 0.3,
  • 13:58each new case only created a third of a new case.
  • 14:02So that's, you know, when the epidemic extinguishes.
  • 14:06So this Re is very sensitive to the natural rise
  • 14:10of immune people in the population,
  • 14:12and also the human behaviors
  • 14:13or ultimately for lucky vaccination that we might implement.
  • 14:18But there's another very important parameter
  • 14:20that I think will interest this group
  • 14:22and that many that perhaps not all of you have heard about,
  • 14:25which is the variance in the R naught,
  • 14:27or the variance in the Re.
  • 14:30And there was a landmark paper
  • 14:31that was published by Lloyd Smith
  • 14:33and his colleagues in Nature in 2005,
  • 14:36that quantify this using a dispersion parameter
  • 14:39they called Kappa,
  • 14:40which seeks to quantify the interindividual variation
  • 14:44in the R, in the reproductive rate of the pathogen.
  • 14:48So imagine a situation where,
  • 14:52for everyone, every single person in the population,
  • 14:55the R is two, each person infects two other people,
  • 14:59and another situation in which it is zero
  • 15:02for many people, but let's say 50 for one person,
  • 15:05imagine the population, a small population.
  • 15:08The average R in these two situations could be the same.
  • 15:13But the ability of the epidemic to establish itself
  • 15:17actually could be quite different,
  • 15:19and would be much easier in the former case.
  • 15:22In the latter case, you have more super spreading events.
  • 15:26There's a variance in the R
  • 15:28so you've got that right tail distribution,
  • 15:30some situations where one person might infect 50
  • 15:33or 100 people, but also most of the cases
  • 15:36are dead ends, most people infect no one.
  • 15:39So in a population of such people infected
  • 15:41with such a germ,
  • 15:42if one person leaves and goes somewhere else,
  • 15:45most of the time they won't be able to establish an epidemic
  • 15:48in the new location.
  • 15:50So the random movement of people from one population
  • 15:52to another, from a risks source to another place
  • 15:55won't be able to establish an epidemic.
  • 15:58So this dispersion parameter is actually quite important
  • 16:02for what might happen in these types of a situation.
  • 16:05And it turns out that the dispersion parameter for SARS-2,
  • 16:10what we're currently facing
  • 16:11is smaller, the variance is smaller
  • 16:14than the variance was for SARS-1.
  • 16:17And this actually is one of the things
  • 16:19that's making SARS-2 worse for us.
  • 16:21Even though there are super spreading events now,
  • 16:24they are fewer than they were for the previous pathogen.
  • 16:28And more often now, a move of a person
  • 16:31from one place to another starts the epidemic
  • 16:34and can result in it taking off.
  • 16:38Now, superspreading depends not only on the pathogen,
  • 16:40but also of course on the host,
  • 16:42attributes of the host that are immunity to the pathogen,
  • 16:46how irritable like some people,
  • 16:48let's say my cough more than other people,
  • 16:50so that might make me more likely
  • 16:52to be a super spreader than you.
  • 16:55Super-spreading events also have to do with the environment.
  • 16:57This is why a pact conferences of people
  • 17:01are more likely to cause super spreading events,
  • 17:04then open air concerts and so forth.
  • 17:08So people try quickly to get a sense
  • 17:11of the reproductive rate of this pathogen
  • 17:13and they were successful.
  • 17:15There have been like dozens of studies now quantifying this
  • 17:17and the summary statistic
  • 17:19is around two and a half that I told you.
  • 17:21And also the dispersion parameter
  • 17:23they tried to quantify.
  • 17:25Distinctly, people tried to quantify the case fatality rate
  • 17:28or the infection fatality rate of this parameter.
  • 17:31And there's still ongoing of this pathogen
  • 17:34and there's still ongoing debate about the CFR
  • 17:37and the IFR.
  • 17:39The CFR is the fraction of people who die conditional
  • 17:44on their coming to medical attention,
  • 17:46or a little bit better definition,
  • 17:49conditional on their developing symptoms,
  • 17:52something which is called the S-CFR,
  • 17:54the symptomatic case fatality ratio.
  • 17:57And we think this number is about between 0.5 and 1% still.
  • 18:01It could be as low as 0.3%.
  • 18:05But I doubt that it's any lower.
  • 18:07And notice that the case fatality rate
  • 18:09is very sensitive to people's behavior.
  • 18:12You know, do people seek medical care?
  • 18:14You know, if they have mild symptoms from the disease,
  • 18:16they might never tell anybody.
  • 18:19Or it's sensitive to the ability of the healthcare system
  • 18:21to save their lives.
  • 18:23So this is not something that's sort of written in stone,
  • 18:26but it's something that attempts
  • 18:27to quantify how lethal a pathogen is it
  • 18:30that we have on our hands.
  • 18:32The case fatality rate for the seasonal flu is about 0.1%.
  • 18:37So on average, about one out of 1000 people
  • 18:39who get seasonal flu will die,
  • 18:41and SARS-2, the current pathogen we're facing
  • 18:45is less deadly than SARS-1.
  • 18:47The case fatality rate for SARS-1 was about 10%.
  • 18:51Yeah, was about 10%.
  • 18:53So it's about about 10 times as deadly
  • 18:56as the current pathogen.
  • 18:58And similarly the case fatality rate
  • 19:00for the 1918 flu pandemic,
  • 19:03which was very bad, was about 4 to 5%.
  • 19:07Now, one of the things that's interesting about this,
  • 19:09as many of you may know
  • 19:10is that actually a less fatal disease is more difficult
  • 19:15to treat, to stop,
  • 19:20because when the disease kills us rapidly
  • 19:22like Ebola, the victim dies
  • 19:25before they can transmit the disease.
  • 19:28But if the disease's less deadly,
  • 19:30and the person is walking around
  • 19:32for a longer period of time,
  • 19:33while sick, they can infect more people.
  • 19:37So this this difference between SARS-2,
  • 19:39what we're facing and SARS-2 in 2003,
  • 19:43I should have mentioned that the SARS-1 petered out
  • 19:45there were only eight and a half thousand cases worldwide.
  • 19:48You know, it was a trivial pandemic
  • 19:51compared to what we're facing now.
  • 19:54So and I forgot how many deaths but it was in,
  • 19:57you know, I think 500 or six 700 deaths
  • 20:00from the SARS-1 pandemic.
  • 20:03So the lower the fatality of this pandemic,
  • 20:06ironically makes it more dangerous,
  • 20:09lower the fatality on a per case basis,
  • 20:11because it can spread farther
  • 20:12and ultimately cause many deaths.
  • 20:15And in general, it's an evolutionary biology principle
  • 20:18that the pathogens don't want to kill us.
  • 20:21That is to say, pathogens do better
  • 20:24when they're not as deadly because they can spread.
  • 20:27And also variants of the pathogen
  • 20:30that don't kill us or don't kill us fast,
  • 20:33typically outstrip variants that do kill us fast.
  • 20:38So that's one of the reasons in general
  • 20:40we tend to see the evolution of pathogens
  • 20:43to be less severe as time goes by.
  • 20:46And I'll come back to this point as well in just a moment.
  • 20:52And then the infection fatality rate
  • 20:55as distinct from the case fatality rate,
  • 20:57is the fraction of people who get infected and die.
  • 21:02Not the ones that come to medical attention.
  • 21:04So we think that about 50% or develop symptoms,
  • 21:08we think that about 50% of people
  • 21:10who get SARS-2 are asymptomatic.
  • 21:13And so this means that in this case,
  • 21:15because of that 50% number,
  • 21:18it means that the IFR is about half the CFR in this case.
  • 21:22So, half the people that get infected,
  • 21:26don't get any symptoms at all.
  • 21:28And so this makes the IFR lower
  • 21:30by a factor of two than the CFR.
  • 21:34Now you can take these two parameters,
  • 21:36the reproductive rate and the case fatality rate,
  • 21:39and you can put them on a little graph
  • 21:41and then you can plot all of the pandemics
  • 21:44that have occurred, let's say in the last hundred years,
  • 21:47this is a typical exercise that epidemiologists engage in.
  • 21:51And if you do that, you find very distressingly
  • 21:55that the SARS-2 pandemic
  • 21:57falls between the 1957 influenza A pandemic,
  • 22:02which was the second deadliest pandemic
  • 22:04we've had in the last hundred years.
  • 22:07And the 1918 pandemic, which is the deadliest.
  • 22:10So, this is a serious pathogen SARS-CoV-2.
  • 22:14It's right there in between the upper right corner is 1918,
  • 22:17it's not as bad as that,
  • 22:19but it's worse than 1957
  • 22:22when you look at these two numerical parameters.
  • 22:26And in fact, it was clear to many people in certainly
  • 22:30by February, that without action,
  • 22:32many people would die.
  • 22:34I think hundreds of thousands of Americans would have died,
  • 22:37had we done nothing.
  • 22:38And unfortunately, I still think
  • 22:40that hundreds of thousands will die.
  • 22:42We've already had 100,000 deaths,
  • 22:44I think we're gonna very likely have at least another couple
  • 22:47of hundred thousand deaths
  • 22:49before the epidemic ultimately winds down
  • 22:52in two or three years.
  • 22:55And this partly relates to the fact
  • 22:56that we're gonna have more waves,
  • 22:58which is a point I'll come back to.
  • 23:00The disease itself has very wide range of presentations,
  • 23:04from asymptomatic to mild to critical
  • 23:07and can affect many organ systems,
  • 23:09not just the upper airway or the lungs,
  • 23:11but also the heart and the kidneys
  • 23:16and the intestinal system and so forth.
  • 23:18And the symptomatology is very protein as well.
  • 23:21People manifest a great variety of symptoms.
  • 23:22There are three clusters of symptoms,
  • 23:25most are respiratory cough, shortness of breath, fever.
  • 23:28Some are the musculoskeletal system, fatigue,
  • 23:32muscle pains, joint pains.
  • 23:34And some are intera, diarrhea, vomiting,
  • 23:37nausea, and again, maybe a fever.
  • 23:42The case fatality rate for this respiratory,
  • 23:45for respiratory diseases in general
  • 23:47typically varies with age.
  • 23:50And most most of the respiratory pandemics
  • 23:54of the last century have had a U-shaped function.
  • 23:57So the very young and the very old
  • 24:00are at the greatest risk of death.
  • 24:03Famously, in 1918, there was a W-shaped function.
  • 24:07There's some interesting theories
  • 24:08if we have time and you're interested,
  • 24:10I can tell you what some other scientists
  • 24:12have speculated as to why it was a W, the very young,
  • 24:15very old were killed and middle aged,
  • 24:18sort of working age young adults, 20s and 30s were killed.
  • 24:24And then finally, there's an L-shaped
  • 24:25or backward L-shaped curve.
  • 24:28So polio has a regular L-shape curve.
  • 24:31So polio pandemics, they kill the young
  • 24:33and they sort of spare the old.
  • 24:35But Coronavirus has a backward L-shaped,
  • 24:38it spares the young and kills the old
  • 24:41and this is unusual,
  • 24:43very unusual actually for a pathogen,
  • 24:46and the fatality rate rises
  • 24:48from about one out of 3000 people younger than 20.
  • 24:52The case fatality rate so conditional on getting sick,
  • 24:55one out of 3000 people will die
  • 24:57to about one out of 100 for people
  • 24:59in their late 50s, early 60s
  • 25:01to about one out of five for people who are older than 80.
  • 25:05So pretty sharp L-shaped curve.
  • 25:10And I found it very poignant,
  • 25:13almost biblical actually
  • 25:15and sweet that this epidemic spared the young
  • 25:19because, you know, the young,
  • 25:20the leading killer of young children is infectious disease,
  • 25:23something like 60% of kids
  • 25:25under five worldwide die of infections.
  • 25:30And the fact that this virus spared them
  • 25:33was very pleasing to me.
  • 25:35And moving actually, and as a parent,
  • 25:39I didn't have to worry about my college age kids
  • 25:41or we have a new child but Eric
  • 25:43and I do that's 10 years old,
  • 25:44and so we didn't have to worry about him, which was helpful.
  • 25:49Now, another important part of the epidemiology
  • 25:51of this condition is something known
  • 25:52as the incubation period.
  • 25:54Incubation period is the time between being infected
  • 25:59and developing symptoms.
  • 26:01And that is between two and 24 days,
  • 26:04I'm sorry, two and 14 days,
  • 26:07the incubation period varies between two and 14 days,
  • 26:10more precise estimates of this have shown recently
  • 26:13and people are studying this a lot, that 97.5% of people,
  • 26:19if they're going to get symptoms,
  • 26:21after being infected, get them by 11 and a half days.
  • 26:26So, people are still studying the details of this,
  • 26:28but the gist is that early on,
  • 26:30it was established that 95% of cases got symptoms
  • 26:33within the first 14 days of infection.
  • 26:36And this is the origin of the 14 day quarantine
  • 26:39that we have all been practicing.
  • 26:42There's a different quantity known as the latency period.
  • 26:46This is the time from infection to infectiousness,
  • 26:48how long between when you're infected
  • 26:50and can infect others and very sadly for us
  • 26:54in this pathogen, unlike SARS-1 in 2003,
  • 27:00this latency period can be a couple of days
  • 27:02shorter than the incubation period.
  • 27:05That means that people can spread the infection
  • 27:08when they're asymptomatic.
  • 27:10And many estimates from China and Italy
  • 27:12and England suggest that the majority of cases,
  • 27:16the bare majority maybe or sometimes the great majority,
  • 27:20arise from this type of transmission.
  • 27:23So most people become infected from other people
  • 27:26who don't have symptoms let's say.
  • 27:29The difference between these two
  • 27:30is known as the mismatch period.
  • 27:33Actually, in veterinary medicine,
  • 27:34there's some veterinary scientists
  • 27:35that call it the Omega period,
  • 27:37which I think is kind of interesting
  • 27:39when you think about the implications for us.
  • 27:44In some cases, the latency period
  • 27:47is shorter than the incubation period,
  • 27:50for example, like in HIV.
  • 27:53So people with HIV can be infectious for years
  • 27:56before they have symptoms,
  • 27:58that makes the disease difficult to control,
  • 28:01or the latency period can be equal to
  • 28:03or longer than the incubation period, like smallpox.
  • 28:08You have to get smallpox vesicles
  • 28:10on your body before you can infect other people.
  • 28:12So we can see who's infected.
  • 28:15And that makes quarantine so much easier
  • 28:18and so much effective.
  • 28:21So the fact that there is a negative mismatch period,
  • 28:24that is to say that the latency is shorter
  • 28:26than the intubation on average.
  • 28:27And this condition is one of the things
  • 28:30that makes it so nasty and difficult to treat.
  • 28:34In terms of transmission modes,
  • 28:36there's a lot of ongoing research on this,
  • 28:37it's clear that the primary mode
  • 28:40is through respiratory droplets, people coughing,
  • 28:44or speaking loudly or singing.
  • 28:47There have been a number of super spreading operates
  • 28:48associated with singing or yelling.
  • 28:52But there's also evidence
  • 28:54that there's airborne transmission
  • 28:55which is small little parts of droplets
  • 28:57come out of your mouth and fall down,
  • 28:59which is why wearing a mask is effective.
  • 29:01Airborne droplets can stay suspended
  • 29:04in the air and spread farther.
  • 29:06There is airborne transmission.
  • 29:07But it's not so bad.
  • 29:10We don't think in this condition.
  • 29:12Although well, I won't go into it.
  • 29:13There's some examples.
  • 29:15There's also spread by fomites,
  • 29:17that's surfaces that we touch.
  • 29:19Although this is increasingly not seen
  • 29:21as a major vehicle for transmission,
  • 29:24there's also fecal transmission
  • 29:25although again, this is not a major explanation
  • 29:28for what's happening in the epidemic.
  • 29:31Now, how do humans respond to epidemics?
  • 29:36Well, the broad division is pharmaceutical interventions,
  • 29:39and so called non-pharmaceutical interventions.
  • 29:42We don't have any pharmaceutical interventions
  • 29:44really for this pathogen.
  • 29:45We have no vaccines for it,
  • 29:47although we're working on it.
  • 29:48We have no drugs, although Remdesivir
  • 29:51has recently been felt to have some benefit
  • 29:53it's modest benefit and we don't have drugs,
  • 29:56and in general viruses are very difficult
  • 29:57to treat, antiviral medications generally
  • 30:00are weak in their effectiveness.
  • 30:03So, just like plague is an ancient threat to human beings,
  • 30:09we have to respond to a familiar enemy
  • 30:12with a familiar response,
  • 30:15which is physical distancing.
  • 30:17People have been physical distancing
  • 30:19in times of plague for centuries.
  • 30:21And unfortunately, that's what we have to do.
  • 30:24We have to engage in a non-pharmaceutical interventions.
  • 30:29There are two broad kinds
  • 30:30of non-pharmaceutical interventions,
  • 30:32individual interventions, things like hand washing,
  • 30:36or mask wearing, or self-isolation
  • 30:40and collective interventions
  • 30:41that required the action of groups of people
  • 30:44or the state, border closures, collective hygiene,
  • 30:48you know, cleaning the subways for example,
  • 30:50testing and tracing, bans on gatherings,
  • 30:54school closures, and ultimately stay at home orders.
  • 30:59And these two classes,
  • 31:01these non-pharmaceutical interventions
  • 31:02can be divided into individual and collective,
  • 31:06but they can be divided in a different way
  • 31:09in what are known as transmission reduction
  • 31:12and contact reduction.
  • 31:14So transmission reduction are things
  • 31:15that try to reduce the likelihood
  • 31:17that conditional on my interacting
  • 31:19with you, I give you the germ.
  • 31:20So wearing a mask or washing my hands
  • 31:22or sanitation measures
  • 31:24might be transmission reduction measures.
  • 31:28But contact reduction or condom use
  • 31:31in the case of HIV is a transmission reduction intervention.
  • 31:35And contact reduction
  • 31:37is when you try to reduce the amount of social mixing.
  • 31:40So gathering bands, self isolation, school closures,
  • 31:44or in the case of HIV reduction and partner number,
  • 31:47those are examples of contact reduction interventions.
  • 31:51And the point of these interventions
  • 31:54however we taxonomise them, is to flatten the curve.
  • 31:58We've all heard about that now,
  • 31:59but why are we trying to flatten the curve?
  • 32:01We're trying to spread out,
  • 32:03we're trying to this wave
  • 32:04is about to hit us with a new pathogen
  • 32:06for which we have no immunity.
  • 32:07And the force, the compressive force
  • 32:09of the wave is gonna hit us,
  • 32:11what we're trying to do is deaden the wave, slow it down,
  • 32:14like build breakwaters offshore,
  • 32:16so maybe even if the same amount of water comes ashore,
  • 32:20it will come ashore with lower intensity.
  • 32:23So that's what we're trying to do.
  • 32:24We're trying to flatten the curve.
  • 32:27And what we mean by that
  • 32:29is that we are going to allow the healthcare system
  • 32:31and the supply chains time to work.
  • 32:34By flattening the curve,
  • 32:35maybe we can save more lives
  • 32:37by not overwhelming our healthcare system.
  • 32:39That's one bit.
  • 32:41The second reason we flatten the curve
  • 32:43is that it postpones some cases
  • 32:45and deaths into the future,
  • 32:48at which time we might have a vaccine,
  • 32:50that might prevent some of the deaths
  • 32:52or we might have better knowledge
  • 32:53of how to treat the condition.
  • 32:55Again, reducing the total number of deaths.
  • 32:57So flattening the curve could reduce deaths
  • 33:00in this way as well merely by the postponement function.
  • 33:04And finally flattening the curve
  • 33:06may be beneficial because it postpones some of the cases
  • 33:10to occur at a time when the pathogen might,
  • 33:13if we're lucky, have mutated to be less deadly.
  • 33:17Remember, we mentioned this earlier.
  • 33:19So if the pathogen has become less deadly,
  • 33:21people will become infected in the future,
  • 33:23we'll get a milder variant of the disease.
  • 33:26But to be clear, what flattening the curve does not do
  • 33:32is eradicate the pathogen.
  • 33:34What we are doing is stopping transmission,
  • 33:36not killing the germ.
  • 33:39The pathogen is still there, and it's going to come back.
  • 33:42It's coming back in Asia,
  • 33:44it's gonna come back in the United States,
  • 33:47there is no escaping from this.
  • 33:49The pathogen is now a feature of our environment
  • 33:52with which we must cope.
  • 33:55Now, one of the features of this pandemic
  • 33:57I won't spend much time on it
  • 33:59is the affliction of healthcare workers,
  • 34:01why health care workers were at special risk.
  • 34:04This increased risk of healthcare workers
  • 34:07has been noted since time immemorial.
  • 34:09Thucydides in the plague of Athens
  • 34:11in 430 BC, talks about how doctors
  • 34:14are dying in greater numbers
  • 34:16and knew the reason.
  • 34:17It's because they're having contact
  • 34:19with sick patients.
  • 34:20The same thing is happening in our society
  • 34:22and happened in China and happened in Italy.
  • 34:25And part of the reason they're at special risk
  • 34:27is that health care workers
  • 34:29in the course of caring for path people,
  • 34:31especially when they have not had
  • 34:32adequate personal protective equipment,
  • 34:35the lack of which has enraged me, in our society.
  • 34:40The reason is they get high viral inoculum.
  • 34:43So they're up close working with the patient,
  • 34:44the patient coughs in their face.
  • 34:46So they, you might get the germ from touching something
  • 34:50in the subway, or interacting with a colleague at work
  • 34:54who speaks loudly and some number of particles
  • 34:57leave that person's mouth
  • 34:58and enter your body, by the time those viruses
  • 35:01are able to multiply, your body might be able
  • 35:04to mount an immune system, immune response
  • 35:07and clamp down on the infection
  • 35:09so you don't get a serious infection.
  • 35:12But a healthcare worker getting a high viral load,
  • 35:15a large inoculum actually can't do that, is overwhelmed,
  • 35:20their body is overwhelmed.
  • 35:22And there's a very moving website
  • 35:23that's tracking the needs of healthcare workers
  • 35:26around the world who have died during this pandemic
  • 35:29and it's growing every day.
  • 35:33Other places of outbreaks
  • 35:34have been nursing homes, prisons, ships.
  • 35:39You've all heard about the cruise ships
  • 35:40and of course the aircraft carrier
  • 35:43and meatpacking plants,
  • 35:45which is a very interesting if you want,
  • 35:47we can talk a little bit about the packing plants.
  • 35:49The burden of this illness falls harder on men.
  • 35:52Men are more likely to die to get it and to die.
  • 35:55But equally likely to get it
  • 35:57but they're more likely to die than women.
  • 36:00And as is typical of infectious diseases,
  • 36:02it's socially stratified,
  • 36:04the poor and the marginalized and the sick
  • 36:07are more prone to die of this condition.
  • 36:13Now, let's turn briefly to this issue of waves
  • 36:16of the pandemic.
  • 36:20Because I think
  • 36:22it's a serious problem.
  • 36:25Every respiratory pandemic, in the last century
  • 36:28has had multiple waves.
  • 36:30And these typically recur in the fall.
  • 36:33Not always, I think,
  • 36:34because of all the protests that we've seen and the rush
  • 36:37to even before the protest, the rush to leave the lockdowns,
  • 36:42I think we're gonna see an earlier wave
  • 36:44in the United States.
  • 36:47And these waves typically come every year
  • 36:50for two or three years
  • 36:52until eventually the epidemic becomes endemic in us.
  • 36:56We saw waves in 2009 with a very mild H1N1 pandemic,
  • 37:01there was a pandemic in 2009,
  • 37:03but that pathogen was not very deadly.
  • 37:05So nobody noticed.
  • 37:06It was actually less deadly than the flu.
  • 37:08So it circulated the whole world
  • 37:10there were waves, we can see the waves of H1N1,
  • 37:13but we didn't care because it didn't kill very many people.
  • 37:16The 1918 pandemic,
  • 37:19the second wave famously came out of phase
  • 37:22with the first wave.
  • 37:23There's some interesting theories
  • 37:25as to why and was much four times
  • 37:27as deadly as the first wave.
  • 37:29We have no way of knowing how deadly the Coronavirus
  • 37:32the second wave will be.
  • 37:35But I don't believe it'll be less deadly
  • 37:38than the first wave for various reasons.
  • 37:42Now the reason for the occurrence
  • 37:43of these waves, it's complicated.
  • 37:45It has to do with human behavior in part,
  • 37:48which is the fact that people return to school
  • 37:50and move indoors with the coming of the fall
  • 37:53and it gets colder.
  • 37:54It has perhaps to do with environmental factors
  • 37:57to the extent that heat
  • 37:58and humidity affect the spread
  • 38:00or modify our body's resistance to the pathogen.
  • 38:03And of course, the epidemic right now has gone
  • 38:06to the southern hemisphere and is raging there.
  • 38:09And Brazil is having for a number of reasons,
  • 38:11including that it didn't make any efforts
  • 38:13to do anything about it.
  • 38:14You know, many, many hundreds of thousands of people
  • 38:17are going to die in Brazil.
  • 38:20Incidentally, I should mention.
  • 38:23I'll just take a small digression
  • 38:25that there's a lot of geographic variation
  • 38:29with these respiratory pandemics.
  • 38:30And we don't fully know the reason.
  • 38:32For example, in the 1957 pandemic,
  • 38:35there was a 30-fold variation
  • 38:37in the final attack rate,
  • 38:38the number of people that got the disease.
  • 38:41So Chile was really hard hit and Egypt was spared in 1957.
  • 38:46We're gonna say that see the same thing
  • 38:48with this pandemic, some parts of the country
  • 38:50will be very hard hit,
  • 38:51other parts of the country will not,
  • 38:54some countries in the world will be hard hit, some will not.
  • 38:57Sometimes this will have to do with the temperature
  • 39:00in the region, sometimes it'll have to do with what
  • 39:02the nations did in response.
  • 39:04But mostly, most of the variants will be chance,
  • 39:07as far as we can tell from previous analyses
  • 39:10of geographic variation in the pandemic.
  • 39:13Anyway, Brazil is being hard hit at the moment.
  • 39:17But the point I wanna make about this
  • 39:19is that these waves illustrate the fundamental point.
  • 39:23But this epidemic is going to become endemic among us.
  • 39:27Either we will develop herd immunity,
  • 39:31probably at around 50%, ultimately,
  • 39:33of people will be required.
  • 39:35And we can talk about, there's a subtle detail here.
  • 39:37So, if you can compute the fraction of people
  • 39:41that need to be infected
  • 39:42before you get herd immunity naturally infected
  • 39:45and naturally immune before you get herd immunity,
  • 39:48with a little formula
  • 39:49that relies on the R naught of the pathogen,
  • 39:51as we mentioned earlier, in the case of measles,
  • 39:54this epidemic has an R,
  • 39:56let's say around two and a half,
  • 39:58it means that about 60% of people need to be immune
  • 40:01before the epidemic goes away.
  • 40:06But actually, you can sometimes reach herd immunity
  • 40:11at lower percentages,
  • 40:12because of the fact that human populations
  • 40:15are not well mixed,
  • 40:17they have a structured, a network structure.
  • 40:19So typically popular people are more likely to get infected
  • 40:24and therefore more likely to get immune.
  • 40:26And once they become immune,
  • 40:28they're no longer pathways for the movement,
  • 40:30they're no longer vectors for the movement of the pathogen.
  • 40:33So in fact, if you immunize,
  • 40:36let's say, 30% of the most popular people
  • 40:38in a population, you could reach herd immunity
  • 40:41at lower percentages.
  • 40:43So, in practice,
  • 40:44what we typically find is that herd immunity
  • 40:46is reached at a lower percentage.
  • 40:49This is, you know, the pre-pharmaceutical era
  • 40:52at a lower percentage than you would predict
  • 40:55based on the amount of the R naught of the pathogen,
  • 40:57for example in 1957,
  • 40:59the epidemic maxed out at around 40%,
  • 41:02was the final attack rate,
  • 41:03you know from retrospective serology studies
  • 41:07that were done after the epidemic.
  • 41:12So, in our case with this pandemic,
  • 41:14either we will get herd immunity, or we will get a vaccine.
  • 41:18And I've now concluded that for whatever it's worth,
  • 41:23that it doesn't really matter which of those two we get
  • 41:26to first because there'll be approximately at the same time,
  • 41:29the likelihood that we will be able to invent
  • 41:31a good vaccine, fast enough, manufactured
  • 41:36and distributed fast enough to outstrip
  • 41:40the inevitable herd immunity seems low to me.
  • 41:43I do think we will get a vaccine eventually,
  • 41:45but I'm no longer putting my hopes
  • 41:47in that as an exit strategy for this pandemic.
  • 41:50Maybe we'll get lucky.
  • 41:51I hope I'm disproven,
  • 41:54not disprove it, but I hope that doesn't prove
  • 41:57to be the case.
  • 41:59So the attack rate if you multiply
  • 42:01all these quantities together
  • 42:03that I've been telling you,
  • 42:04in the end for this pandemic,
  • 42:05in my view will be 40 to 50%.
  • 42:09Maybe more probably higher if we overshoot,
  • 42:13which is another thing, you know,
  • 42:15the epidemic rages onward
  • 42:17before we have a time to actually catch up with it.
  • 42:21And this partly relates to the issue of who are,
  • 42:24you know, like I already said, the popular people
  • 42:26and the acquisition of immunity.
  • 42:30Now, where do we stand with this pandemic so far?
  • 42:33If you look at Cyril prevalence studies
  • 42:34to date in Sweden,
  • 42:36which has adopted a pretty mild approach
  • 42:39to coping with it.
  • 42:40Nationwide there are about 4% of people have had the disease
  • 42:44and are now immune, I think in Stockholm
  • 42:47with seven or eight or 9%
  • 42:49in the most densely populated part of Sweden.
  • 42:52In New York City, it's about 21%
  • 42:54we know from a good study,
  • 42:56and in various era prevalence studies,
  • 42:59no one has really done a perfect study
  • 43:00at anywhere in the United States.
  • 43:02We've been thinking in my lab of doing such a study
  • 43:05in the Greater New Haven area,
  • 43:06picking a random sample of New Haveners,
  • 43:08and then following them prospectively.
  • 43:11Most people have done others sub-optimal.
  • 43:14And I don't criticize them,
  • 43:15it's difficult to get a random sample of people.
  • 43:19But if I had to guess, in our cities,
  • 43:23probably we are at no more than 2, 3, 4, 5%
  • 43:27around the country.
  • 43:28So if we're gonna get to an attack rate of 40,
  • 43:30or 50%, we have a long way to go unfortunately.
  • 43:36I think it's important to note
  • 43:37that the United States response to the pandemic
  • 43:40has been awful, has been completely incompetent frankly.
  • 43:43And the failures in my judgment have occurred
  • 43:46at multiple levels of government, but certainly,
  • 43:49at the White House,
  • 43:51has, you know,
  • 43:52there's been an appalling lack of coordination.
  • 43:55I think the expertise at the CDC was there,
  • 43:58but there were men with deep expertise
  • 44:00at the CDC and deep expertise at the National Institute
  • 44:03of Allergic and Infectious Diseases,
  • 44:08but it hasn't been deployed properly.
  • 44:10But I also think it's fair to say
  • 44:11that many of the state governments
  • 44:13were caught flat footed.
  • 44:14And let's also acknowledge that many European countries
  • 44:16that didn't have the incompetence
  • 44:18at the level of the White House
  • 44:20also seem to have been caught flat footed.
  • 44:22I don't understand why.
  • 44:24The you know, it's not a mystery.
  • 44:27I can reach over and grab a book on my shelf
  • 44:29that's called National Strategy for Influenza Pandemic.
  • 44:33Many, many experts knew what was happening
  • 44:36in January and February.
  • 44:37And the Chinese bought us time, you know,
  • 44:41by locking down their nation.
  • 44:43We had two months to look at what was happening
  • 44:45in China and become concerned.
  • 44:48Let me tell you briefly and then I'll shut up.
  • 44:51What are some of the projects that are happening
  • 44:53in my lab right now
  • 44:57which we'd be welcome.
  • 44:59Welcome collaborators.
  • 45:00And Laura's cooperating with us on some of these things.
  • 45:04We have ongoing work on using big data techniques.
  • 45:08Laura mentioned this paper on human movements.
  • 45:12Many scientists are working on this right now
  • 45:14and there are labs around the world
  • 45:15that are famous for this.
  • 45:16We're trying to contribute to that in a certain way,
  • 45:19a tracking human movements or symptom reporting,
  • 45:24that will using various Big Data techniques,
  • 45:26including Twitter data that would allow us to forecast
  • 45:29the course of the epidemic, to get ahead of it,
  • 45:31to know where it's gonna strike
  • 45:33based on knowing what's happening.
  • 45:34Another similar project like that,
  • 45:36being spearheaded by another graduate student
  • 45:38in my lab, Eric Feltham is looking at gatherings.
  • 45:41For example, we were very interested in the gatherings
  • 45:44to vote the primary elections.
  • 45:45Did they, you know, people got together to vote
  • 45:48at polling places, did that cause a spike?
  • 45:50This is of course highly relevant to our national security,
  • 45:54we need to somehow have a good vote, a fair
  • 45:56and honest vote in November and for that,
  • 45:59in my judgment, We need to have widespread absentee
  • 46:02balloting to allow for this.
  • 46:05Otherwise, if people stay away from the polls,
  • 46:07because they're afraid of the pandemic,
  • 46:09or if they go to the polls and then become infected,
  • 46:12either one of those outcomes
  • 46:14is a threat to our society in my view.
  • 46:16But similarly, I believe that the recent protests
  • 46:19that we've seen after the appalling murder
  • 46:23that we saw in Minnesota,
  • 46:27and the rioting that we've seen,
  • 46:29are gonna contribute to a spike in cases
  • 46:32and I mentioned this earlier.
  • 46:34Finally, we had just released last week
  • 46:37an app from my lab.
  • 46:38That's called Hunala, hunala.yale.edu.
  • 46:47This app relies on some old ideas of ours,
  • 46:50involving network science
  • 46:52that I previously discussed with Raphael
  • 46:54and others on this call,
  • 46:57which is that if you think about
  • 46:59a contagion that begins stochastically in a graph,
  • 47:03you should have the intuition,
  • 47:04it's obvious that the contagion
  • 47:05as it winds its way a social contagion,
  • 47:08winds its way a biological contagion,
  • 47:10that winds its way through the graph
  • 47:11is gonna reach central people
  • 47:13sooner than it reaches random people in the population.
  • 47:17So if we could identify central people,
  • 47:19and monitor them, they would function
  • 47:22as a kind of canary in a coal mine
  • 47:24forecasting the state of the epidemic at some future time.
  • 47:28We published several papers about this 10 years ago,
  • 47:31that showed that we could do it with each one and one,
  • 47:34that we could track central people,
  • 47:36and that we can monitor them.
  • 47:39And then we would therefore be able to use them
  • 47:42to predict the future state of the epidemic
  • 47:44between two and six weeks in advance.
  • 47:49So this is one of the things that our app
  • 47:52is attempting to exploit.
  • 47:53we're attempting to get people to report their symptoms
  • 47:57and then we are attempting to monitor that
  • 48:00or redraw the graph anonymously or privately.
  • 48:05We don't inform anyone
  • 48:06for example, if you get sick, we do not inform your friends
  • 48:10that you are sick.
  • 48:12But we exploit people's reports
  • 48:13to create a kind of ways for Coronavirus.
  • 48:17So everyone contributes a little information
  • 48:19saying whether or not they have symptoms.
  • 48:22And we can then by manipulating that information
  • 48:26using some machine learning algorithms
  • 48:28in partnership with a mean car buses group
  • 48:30in the electrical engineering department here.
  • 48:34We can predict what your risk of getting the epidemic
  • 48:38is in the future.
  • 48:39So if your friends friends friends
  • 48:41had respiratory disease three weeks ago,
  • 48:45that should modify your risk.
  • 48:47Or if your friends had a fever a week ago,
  • 48:50that should modify your risk.
  • 48:52And so our app is attempting to collect these data
  • 48:55and forecasts the future course of the epidemic.
  • 48:57We expect to have, if we're lucky
  • 49:00and if the app is widely adopted,
  • 49:04we will have a ton
  • 49:05of very difficult complicated data to analyze.
  • 49:08And just because I haven't been as clear
  • 49:09about this as I hope because I'm rushing,
  • 49:12the app is like waves for Coronavirus.
  • 49:14It warns you just like waves does
  • 49:17that there's a traffic jam two miles ahead,
  • 49:19the app can warn you that there is pathogen
  • 49:23in your social network neighborhood.
  • 49:25And just like in ways you might, you know, take an exit
  • 49:28and avoid the traffic jam, with ways you might say,
  • 49:30you know, I'm gonna stay at home now,
  • 49:32because my risk is high.
  • 49:34And we give people daily assessments of their risk
  • 49:38based on where they live.
  • 49:40And here we include lots of public source data
  • 49:43and the reports of our users.
  • 49:46And we give them a risk of like based on where you live,
  • 49:49is the risk low, medium, high.
  • 49:52Like for example, like a fire like a forest fire prediction,
  • 49:56you know, based on the humidity today,
  • 49:58what's the risk of a forest fire?
  • 50:00And, we also give you a personal risk
  • 50:03based on where you are in the network.
  • 50:05So traffic is bad in New York City in general,
  • 50:08but it's really bad on your block right now,
  • 50:10waves might tell you.
  • 50:12So our app also does that.
  • 50:15People have reported seeing a fire,
  • 50:17it's not just that it's dry and hot,
  • 50:19actually other users in your area have seen a fire.
  • 50:22So your risk now is much higher.
  • 50:26So I'm gonna say one last thing,
  • 50:27which is that the issue is when will this epidemic end?
  • 50:32And I already alluded to the fact that
  • 50:33it's gonna end when it becomes endemic among us,
  • 50:36but it's also gonna end,
  • 50:39or epidemics have a biological end and a social end.
  • 50:43And basically, the epidemic is gonna end
  • 50:45when we declare that it's over.
  • 50:46But we have come to accept it.
  • 50:49And so I think we're still in the early phases
  • 50:51of the Kübler-Ross model of grief.
  • 50:54You know, we have anger
  • 50:58and we have sadness, you know, depression.
  • 51:01And we have bargaining, you know,
  • 51:03but soon we're gonna have acceptance,
  • 51:05which is the only way out from this pandemic.
  • 51:10Thank you.
  • 51:12- Thank you Nicholas for this great talk.
  • 51:15We have time for questions.
  • 51:22- [Jose] Hello, I'm Jose.
  • 51:24So I was curious about the app
  • 51:27that you develop in collecting data.
  • 51:30Are you interested in like, what network properties
  • 51:36are you like hoping to gauge, are you looking into,
  • 51:39like sensitivity, you know, like a core.
  • 51:45And do you have a particular hypotheses about like,
  • 51:50because like when you're in a network,
  • 51:52you receive information,
  • 51:54the more connected you are just like,
  • 51:57forgot the term but I think you tend to get more information
  • 52:01based on your connection,
  • 52:03are you expecting that people in certain types of networks
  • 52:08would be reporting
  • 52:11like symptoms like earlier?
  • 52:13Or I was just wondering like,
  • 52:16what are you hoping to capture using those parameters?
  • 52:19- So we've studied that if you want,
  • 52:21you can look at our 2010 paper on the H1N1.
  • 52:24Networks with low transitivity are at higher risk.
  • 52:28So individuals that have low transitivity
  • 52:30in their neighborhood are at higher risk
  • 52:31for getting the flu.
  • 52:33Unsurprisingly, people with higher degree
  • 52:35are at higher risk,
  • 52:36and people that have high centrality are at higher risk.
  • 52:40So we've shown that before.
  • 52:41And eventually we hope if we reach the appropriate scale,
  • 52:44that these parameters will also be relevant.
  • 52:47But it's not just the structure of the network
  • 52:50as I alluded to earlier.
  • 52:51It's what's happening in the network and when.
  • 52:53So I don't know exactly yet.
  • 52:55We don't have the data yet to know what is the real signal.
  • 52:59is it COVID three weeks ago in your third degree alters,
  • 53:03is it COVID 10 days ago or 12 days ago in your alters?
  • 53:08You know, we don't know all these details yet.
  • 53:10If we get enough data, we will know the answer.
  • 53:13But we do know from principles that we published before
  • 53:16and that are mathematically predicted
  • 53:18what should matter.
  • 53:20Excuse me, I've been dying to do a K cornice forecasting.
  • 53:24And I've been talking to Dan Spielman,
  • 53:26I actually had to (mumbles) about this for years.
  • 53:28We have another project in Honduras
  • 53:30that Laura's involved with
  • 53:31where we're gonna look at K cornice,
  • 53:34I don't know whether we'll be able to do with this app.
  • 53:35It depends on how much use we get.
  • 53:39- [Jose] Thank you.
  • 53:44- Other questions?
  • 53:46- Yeah, I have a question.
  • 53:49Hi, my name's Sam Burma, graduate in genetics.
  • 53:51I'm interested in understanding a little bit more
  • 53:52about the dispersion you're talking about
  • 53:55with the equivalent reproductive,
  • 53:57the effective reproductive rate of the virus,
  • 53:59and I think I made it just missed a little bit there.
  • 54:02But can you just go back and explain a little bit more
  • 54:04about how this relates to the fact that
  • 54:07there's a lower dispersion rate actually is worse?
  • 54:11- Yeah.
  • 54:12It's a wonderful, beautiful paper by Lloyd Smith
  • 54:15at all in Nature 2005 which is it's just,
  • 54:19it's like one of those papers, you read it,
  • 54:21and you get the point that you read and you get
  • 54:23"There's more subtlety here," and then you read it again,
  • 54:24"There's more subtlety here."
  • 54:26It's a wonderful paper.
  • 54:28The gist is, if you think about it,
  • 54:30and also just incidentally, as an intellectual point,
  • 54:32there's something happening in the sciences.
  • 54:34This is an old issue in the sciences
  • 54:36between lumpers and splitters.
  • 54:37No Darwin talks about this, lumpers and splitters.
  • 54:41There are scientists who are concerned about
  • 54:42getting the sense of a thing
  • 54:44and like the average and they're scientists
  • 54:46who are interested in variants
  • 54:48like what's, you know, what are exceptions?
  • 54:50How do things spread out?
  • 54:51And so much of statistics and social sciences
  • 54:54in the last 50 years has been focused
  • 54:56on measures of central tendency, why?
  • 54:59Because we invented regression models and statistical tools
  • 55:02that were easier for us to say that,
  • 55:04whereas variance is also so important.
  • 55:06And so there's a lot across the social scientists
  • 55:08and people that are becoming much more interested
  • 55:10in variants and in variation.
  • 55:12And so this is another example of that,
  • 55:14where for a long time, people were, you know,
  • 55:16trying to estimate the R naught.
  • 55:17And Lloyd Smith comes along at all,
  • 55:19and says, "Wait a minute, the variance is also important."
  • 55:23Why does it matter?
  • 55:24So if everyone in the population
  • 55:26has an R naught of two,
  • 55:29then every single time one person goes from one place
  • 55:31to another, then we'll restart the epidemic.
  • 55:36But if there's variation and some people
  • 55:38or most people have an R naught of zero,
  • 55:40they cannot give it to anyone.
  • 55:42And one out of 100 people can give it to a lot of people,
  • 55:46the epidemic is more likely to extinguish
  • 55:48'cause 99 out of 100 times when one person
  • 55:51in the latter example goes somewhere,
  • 55:53the epidemic stops.
  • 55:54Only if that's superspreader,
  • 55:56the person with a capacity to be a superspreader
  • 55:59goes does it get started.
  • 56:00And if it's something intrinsic to the germ,
  • 56:03then in the next cycle,
  • 56:04most of the transmissions will also be zero.
  • 56:08And so this dispersion, which they in that paper,
  • 56:11they quantified across pathogens
  • 56:13seems to fit with the ability of the pathogen
  • 56:17to get instantiated.
  • 56:22- Thank you so much.
  • 56:25And I also shared that paper in the chat for him.
  • 56:26- Yeah, and also, just so you know,
  • 56:28it's estimated that with with SARS-2
  • 56:31what we're currently facing,
  • 56:33is you need four importations to get one transmission,
  • 56:38one community transmission.
  • 56:40So in fact, in Seattle patient zero,
  • 56:43did not infect anybody else.
  • 56:45The first case that arrived in the middle of January,
  • 56:48when they did contact tracing elaborately,
  • 56:50and now genetic studies,
  • 56:51he didn't infect anyone else.
  • 56:52It was a dead end.
  • 56:54You needed subsequent importations
  • 56:55before the epidemic took root in Seattle.
  • 56:59Same in China by the way.
  • 57:04Other questions?
  • 57:07It's very weird Zoom, 'cause I can't see you.
  • 57:09I don't know if I'm boring you,
  • 57:11I have no way of judging how I'm coming across.
  • 57:14You know, I could be coming across as very aggressive,
  • 57:17which of course not my intention.
  • 57:22- I mean, if no one else has questions.
  • 57:23I was also wondering what are the factors
  • 57:25that influence our ability to detect the difference
  • 57:27between the latency and the infectivity period
  • 57:29'cause I remember at the early days in the US,
  • 57:33CDC was still saying they didn't think
  • 57:34that asymptomatic spread was a major factor for COVID-19.
  • 57:39But now it seems like...
  • 57:42- I haven't dug deep in what the CDC
  • 57:44was saying but they knew
  • 57:47there was a symptomatic transmission
  • 57:48certainly by the end of January.
  • 57:51There's some CDC announcements
  • 57:54that I haven't traced it all the way back
  • 57:56but for sure, by the end of January,
  • 57:57they were already saying this
  • 57:59and I think earlier in January too.
  • 58:00There was some confusion in December still,
  • 58:03but certainly by January, people knew.
  • 58:11The genetics of human susceptibility
  • 58:13are very interesting, by the way.
  • 58:15I think we're gonna find that a small part
  • 58:18of the geographic variation will relate
  • 58:20to the genetics of the pathogen.
  • 58:23Right now, there's no evidence that yet that some strains
  • 58:27of the pathogen are more deadly
  • 58:29or more infectious than other streams,
  • 58:30a lot of interest in this topic right now.
  • 58:33We probably will find some of that.
  • 58:35And there's also some small evidence
  • 58:38that human genes, that some people may be more of you
  • 58:41than others, because of, you know,
  • 58:43their various variants still to be described.
  • 58:47So I think that's gonna be
  • 58:48and I think that'll be a small part
  • 58:50of explaining that geographic variation, not a big part.
  • 58:53- Nicholas, Luke here in the chat box has a question.
  • 58:57He's interested in your thoughts about transmission settings
  • 59:01like nursing homes, meatpacking.
  • 59:03Is it network structures susceptibility
  • 59:06excetera driving transmission?
  • 59:08- Well, I think the meatpacking,
  • 59:10the story for the meat packers was aerosolization
  • 59:13of the pathogen in a cold environment
  • 59:15and high density and I have a long Twitter thread
  • 59:18on this if anyone is interested on why meatpacking
  • 59:22and it's a worldwide phenomenon,
  • 59:24it's not just a phenomenon in the United States.
  • 59:26So I think the explanation by the Secretary of Health,
  • 59:31the Secretary of Health in our country,
  • 59:33that it had to do with the living arrangements
  • 59:35of the immigrants working in these factories,
  • 59:38is not correct.
  • 59:40There are other factories
  • 59:41with other industries with similar immigrant populations
  • 59:44and similar living conditions
  • 59:45and they didn't have the operates.
  • 59:46I think it's to do with the temperature in the environment
  • 59:50where these people work.
  • 59:51It's refrigerated, the very tight packing of the workers
  • 59:54and up aerosolization using saws and other equipment
  • 59:58that create aerosols
  • 59:59that are in the turbulent wind conditions
  • 01:00:01in these factories.
  • 01:00:04The nursing homes is different,
  • 01:00:05I think it's a very customary health care situation
  • 01:00:08where you have very vulnerable elderly people
  • 01:00:10and health care workers that are up close
  • 01:00:12and intimate working with people.
  • 01:00:14So once the epidemic takes root, you know,
  • 01:00:16you get a very rapid spread.
  • 01:00:18I think nursing homes are more like prisons, actually,
  • 01:00:20in terms of their epidemiology,
  • 01:00:23meatpacking is different.
  • 01:00:27You know, ships,
  • 01:00:28ships are close quarters as well.
  • 01:00:30So you have young people in ships,
  • 01:00:32you know, in the US's Theodore Roosevelt,
  • 01:00:35you know, you have young healthy sailors,
  • 01:00:36although one died already,
  • 01:00:38we should acknowledge from the disease,
  • 01:00:41but they're very tight.
  • 01:00:42People are living in bunks,
  • 01:00:43you know, one on top of each other in that situation.
  • 01:00:50- Nicholas I have a question.
  • 01:00:52So we're used to think that
  • 01:00:53these social distancing measure, these lockdowns,
  • 01:00:56what they're really doing
  • 01:00:57is reducing the number of susceptible people
  • 01:01:00in the population
  • 01:01:01so that essentially we plug in this number,
  • 01:01:04this lower number in our serial models
  • 01:01:06and predict the spread.
  • 01:01:08But, I think what in addition to reducing the density
  • 01:01:12of the network, what they're also doing
  • 01:01:14is reshaping the network
  • 01:01:16because essentially the structure of the network
  • 01:01:18because people don't go to school anymore.
  • 01:01:20People don't gather in bars.
  • 01:01:22So what do you think of this?
  • 01:01:24Do you think we should make this information
  • 01:01:26will be useful to include in our model is right?
  • 01:01:30- Yes I do think that.
  • 01:01:31And just if there's a little detail there,
  • 01:01:33which was seen in China and the United States
  • 01:01:36is, ironically household transmission,
  • 01:01:39cases of household transmission
  • 01:01:40typically are more severe than cases
  • 01:01:42of community acquired transmission
  • 01:01:44because of the viral inoculum idea.
  • 01:01:46If I get the pathogen from my wife,
  • 01:01:49I'm gonna get a sicker than if I get the same pathogen
  • 01:01:53from riding in the subway,
  • 01:01:55because I'm writing something
  • 01:01:56I might get a low viral inoculum
  • 01:01:58whereas if I get it from my wife,
  • 01:01:59I'm gonna get a serious case.
  • 01:02:00You know, I kiss my wife, for example.
  • 01:02:04So intra-household transmission is often more severe
  • 01:02:07and more efficacious than out-of-household transmission.
  • 01:02:11So the dynamics will change in quite complicated ways,
  • 01:02:14just like like you're alluding to,
  • 01:02:16and I'm sure other groups are looking at this.
  • 01:02:19It's not something we're actively doing.
  • 01:02:22But again, if you are interested Laura,
  • 01:02:23I'm always eager to work with you,
  • 01:02:24I love working with you.
  • 01:02:27- All right after this (laughing).
  • 01:02:30Now, do you do you wanna take one more question?
  • 01:02:33- I'm happy to take one more
  • 01:02:34but pick one I wanna go so, let's do one more then stop.
  • 01:02:38Anyone else have something?
  • 01:02:42I'm looking around here
  • 01:02:43and all the things I need to monitor.
  • 01:02:50Alright, thank you all very much.
  • 01:02:53- Thank you.
  • 01:02:54Yeah, Luke says thank you for a very interesting point
  • 01:02:58as we think about contact tracing.
  • 01:03:00Okay.
  • 01:03:02- And someone else asked about school reopening.
  • 01:03:04That's a long topic to keep people at the last minute.
  • 01:03:07My wife had a nice piece in The Atlantic about this,
  • 01:03:10if you're interested,
  • 01:03:11I think schools are gonna reopen,
  • 01:03:13I think they need to reopen.
  • 01:03:15They're going to reopen only 'cause we have no choice
  • 01:03:17but it would be better from pandemic point of view
  • 01:03:19if they did not.
  • 01:03:21I think that if they're gonna reopen,
  • 01:03:23a lot of procedures are gonna have to be put in place
  • 01:03:25at Yale, at nursery schools, at elementary schools.
  • 01:03:27It'll be different from place to place.
  • 01:03:30And I think that there will be a second wave.
  • 01:03:35So I think the schools will close again,
  • 01:03:37just what I suspect is gonna happen in October, November.
  • 01:03:43Thank you all very much.
  • 01:03:44- Thank you Nicholas.
  • 01:03:45Thank you all for joining.
  • 01:03:47Bye bye.