 mentioned just now, not only because it's recent, but also because I think it's sort of focused on many of the issues that are in people's minds. And then to talk about a few other areas of work that we have been doing or are doing that I think are particularly important and then I think some of the more general questions about what to do are worth discussions, tried to leave enough time for. So this is the work of quite a lot of colleagues, mainly those who are highlighted. Sounds like someone is not on mute. We've been doing a lot of work in our center and let's talk about some of this. So first topic I want to just briefly mention because it sets the stage for some of the other key points is the background in Wuhan. And this reflects a piece of work that's still finishing up peer review I hope led by Rohan Lee where we just did a very simple thing which was to extract from Wuhan CDC online database the number of individuals in critical care, sorry, the number of individuals critically ill over prevalent on each day in Wuhan and compare that on a per capita basis to for example the US intensive care empty beds and the US intensive care total beds and this in my mind this single figure is kind of the motivation for what we're all trying to do with social distancing which is to keep the hospital system and particularly the intensive care system from being overwhelmed because you see that the peak demand for critical care in Wuhan would have filled every essentially every bed in the United States per capita. And the other so that I think is fairly well appreciated by everyone what I think is maybe less widely understood is the lag times involved in the system. So the lockdown of Wuhan was on January 23rd, we'll come back in a few minutes to the controversy about whether that was fully effective in turning the curve over or not. That's the last thing I want to talk about in this talk but the next ground of interventions was on February 2nd. And what's clear here is that there's a really long lag in the system due to the fact that people stay in intensive care for a while but especially the fact that they take some time to get sick enough for intensive care so that the control measure implemented here takes something like four weeks to manifest as a peak in intensive care demand. So those who think about control theory understand that dealing with these really long lags is a challenge especially if we are having a hard time monitoring mild cases as is the case in many places. But we do need to and this motivates the need for testing for surveillance as opposed to just testing for medical care purposes in order to have a sense of the mild cases and even the asymptomatic cases and what direction they're trending because we can't wait until we see the problem at the intensive care setting. So with that as background, we embarked almost immediately after this became obvious as a problem on trying to understand what we could learn from other seasonal, other coronaviruses. There are four seasonal coronaviruses that circulate every year in temperate climates and peak in the winter. Two alpha coronaviruses and two betas are the group that includes SARS-CoV-2 as well as SARS and MERS viruses but these were the ones on which we focused. And just to try to understand what's the seasonal pattern of change. So one challenge of that is that you need to have some kind of a proxy or incidence which is the thing you would like to model. Namely the number of new cases per population per week or per day. And there is no perfect proxy for incidence because nobody is systematically measuring the same people over and over or even a known sample from the same people over and over. But nonetheless there are better and worse proxies and some of them are shown here and you can see that they kind of qualitatively at least look the same although quantitatively some are a little bit different. And using standard methods developed by Wallinger and Tunis in the Netherlands during the SARS pandemic we estimate a daily reproductive number that is the number of secondary cases generated by each case assuming that this is a linear proxy for incidence and assuming certain things about the serial interval, the time between infections because that's also not known for these coronaviruses. But rather fairly robustly to various assumptions and you can see these in the different lines here of the same color which are varying assumptions. What you find is that there is a peak of infectiousness or of transmissibility in the late, the scale is a little funny. I'm not sure what exactly where that's not the beginning of 20, yes it is, sorry, yeah. So there's a peak in the late fall in transmissibility and because that is essentially the derivative of the case curve, the case curve peaks in the winter around January and declines and you can see suggestion of competition between these two coronaviruses in the sense that they tend to be one or the other in any given year with this exception, well they often tend to be one or the other and dominated in a season. What we can then do is try to decompose that seasonal fluctuation in the reproduction number into a seasonal component which is assumed to be the same every year by month or by week and that's this yellow spline. And the other factor that contributes to the decline of cases as the winter progresses which is that susceptible hosts are depleted and so those are the blue curve showing the depletion by HKU1 and the red curve showing the depletion by OC43, the other coronavirus. And you can see that there is some evidence of cross immunity, for example here, the reduction in the OC43 transmission by both incidents of OC43 and of HKU1. So these are two different coronaviruses that seem to provide cross immunity to each other although less than the immunity they provide each provides to itself. So it's sort of consistent with what you would expect. This is the regression framework in which we did it also developed by Echolonga and his colleagues. And the take home message is that indeed seasonality is as for flu a combination of seasonal favorability for transmission which is highest in the late fall early winter and depletion of susceptibles that then as the winter wears on makes it even less favorable than the seasonal conditions would allow. And we can then simplify that in the seasonal driving sense into a transmission model where we look at the two strains and cross immunity between them with a seasonal amplitude that's the best fit, that's not exactly the fit that's shown here, but the best fit in our final model was an amplitude of about 21% reduction from peak to trough, about 10% up and down from the mean transmission in a cosine function. And this transmission model looks like some kind of eastern mystical thing or something but it's just showing the progression of people from susceptible to exposed to one virus strain to exposed to the other virus strain and infectious with the other. So it's an SEIR type model for two strains. And when you fit that model to the data from the U.S., you get reasonable fits from the simulated in dots and the data in solid lines that shows the alternating pattern by season for, or especially the seasons that are dominated by OC43 and you get a fairly good fit especially during the times when there's a lot of data, which is the darker colored dots here, you get a fairly good fit to the sort of raw estimates that we had before from just this sort of sine wave of transmissibility overlaid with the decline in susceptible hosts. Though it seems like these two factors reasonably capture the seasonality of the viruses and then allow us to ask the question, is that amount of seasonal forcing a lot or a little? And as you have probably heard, the estimates for the reproduction number of COVID-19 are around 2 to 3, some higher, but maybe around 2 to 3 is the close to consensus estimates. And those were measured mostly in the winter. So a 21% decline from 2 to 3 is not enough to by itself reduce transmission of COVID-2 below the critical threshold of 1 in the summer, although it may provide some help. So that's some data analysis. I can take questions on that now or move on and continue, sort of the set up to the next set. One question on this mark is, it seems that the outbreak of these previous viruses was somewhat earlier in the season, sort of in the fall. I guess as much as we have information, I mean that may have been the case again in China, but not in other countries this time around. Does that matter or how we should apply these results to the new virus? You mean that you mean that this peaks in January, which is sort of the close to the beginning of the of the COVID-2 emergence? Is that what you're saying? Exactly. Yeah. So I think as a general point, the dynamics of seasonal viruses, flu or coronaviruses we now see having analyzed it, nobody had ever looked really before, but the dynamics are really driven by two different things. One is this seasonal variability and susceptibility in transmission, which may be heat or maybe humidity or it may be schools, possibly most likely a combination of those and maybe other factors. For flu, it's clearly schools plus humidity that drives it. And then on the other hand, the depletion of susceptibles. So seasonal viruses are seasonal not only because of the environment, but because they're only just building up enough susceptibles to have an epidemic as the environment becomes more favorable. They use that small pool up and then they're done. And so seasonal viruses are seasonal the way they are, importantly because of lack of susceptibles. But when you start with a new virus where susceptible hosts are very plentiful, it breaks all the rules. And we see that in pandemic flu that it can be almost any time of year. So there's no reason why this virus needed to wait for a favorable time. Having said that, if you take literally the timing of the peak transmissibility here for these other viruses, you're certainly right that subtracting 20% might be too generous. It might be really that you should subtract 10% or something from, maybe by the time we were into the real transmission of coronavirus of SARS-CoV-2, it was already in the winter rather than the fall level of transmissibility. But I wouldn't, as you can see from these confidence bounds, I wouldn't put tremendous faith in the exact timing or magnitudes. But I think it is probably a good guess that transmissibility was not at its very peak when, in December and January, it was probably a little bit off its peak. Okay. So I'll continue the next part of this work and all of these figures that I'm showing for this first part of the talk are in the paper that was in Science yesterday. So the next part of this is to then think about what happens with and without an intervention. So to that transmission model, we added a third strain, which is SARS-CoV-2. And we made varying assumptions about the duration of immunity and the extent of cross-immunity to the seasonal coronas. Just try to see what could happen. And of course, we really have no idea other than by assuming it's the same as things we do know about because there's no data yet on immunity or cross-immunity. But so among the scenarios that can happen, if you have an introduction here in early 2020, as in places outside of China in the Northern Hemisphere, you have, if unmitigated, a big outbreak, followed by continued annual outbreaks, assuming that the duration of immunity is short, around 40 weeks. And that's in the same range that we assume immunity to be from the seasonal viruses. It could be considerably longer from a more severe infection like this one. So in that category, in that case, you have a longer period in which the population is not receptive to further viral spread and then outbreaks sporadically maybe on a longer time scale thereafter. And this is classic, what's known as post-honeymoon dynamics for a, or it's almost the same thing as post-honeymoon dynamics for an infectious disease where a very large outbreak overshoots the number of susceptibles a lot. And so keeps things under control for a while. And then as births or waning immunity replenish the susceptible pool, you get a rebound. And people see my cursor. I'm gesturing at things, but I don't know whether I need to use some other button or something. We can follow the cursor. Okay, great. If we assume greater degree of seasonality, then the outbreaks become more punctuated, because again they very much deplete the susceptible pool and then need to wait for a long enough to get immunity down in the population. And if immunity is permanent, then, and there's some cross-immunity, then we could in fact see elimination. And then in the most perverse scenario, if there is enough cross-immunity from these other coronaviruses to SARS-CoV-2, then you might have a situation where SARS-CoV-2 disappears effectively for several years and then resurges afterwards because the immunity from the direct immunity that's the strongest has been slowly waning, but being propped up by the cross-immunity from the seasonal viruses. So there's really quite a range of scenarios, and I think it really suggests that we need to understand immunity and cross-immunity better, a theme to which I will return at the end. So then we began to consider what happens with interventions. And we did that now, for now, since we're thinking in a more short-term fashion. Now, disregarding the seasonal coronaviruses, which means either that they, well, which is essentially assuming that they do not provide any cross-immunity to SARS-CoV-2, it doesn't really matter if SARS-CoV-2 causes immunity to them, as long as they don't affect SARS-CoV-2 strongly. And we created then a very, very simple SEIR-type model in which individuals are sort of fated at the time they get infected, either to have a mild illness or a hospitalized illness or critical illness. And these assumptions are quite important to the quantitative outcomes. And these were, in fact, taken directly from the Imperial College report because they had reviewed the literature more carefully than we had time to, so we just took their numbers. And we'll come back to that also as an issue because those numbers may be revised as better data come in from different places. But the assumption is that the large majority are mild and then 3% require hospitalization and only in another percent and a half almost require ICU. These durations of stay are also important. I don't know if I put a slide in this, but a clever reviewer pointed out that distribution of the time that you stay in ICU is really important. If it's exponential as we sometimes, as we usually model it, there's less congestion in the ICUs, but if it's a more peaked normal type distribution, then the ICUs stay fuller for longer with the same inputs. And so that's a sensitivity analysis we explore in the paper, but I won't talk about in detail. So that's another data need is actually to understand the distribution of stays in ICU in order to do capacity planning. So then we modeled the effects of one-time social distancing and here we took a little bit different approach from what some other groups have done. It's our view or at least my view that we don't really know what the impact of school closures or social gatherings or any of these other individual interventions are on transmission of a virus when people are behaving in a way they've never behaved in modern history and when it's a virus we don't really understand yet. So rather than trying to model individual interventions, we simply made the strong simplifying assumption that the interventions collectively made some reduction in the reproduction number, the total, the transmissibility of the virus. And we modeled the proposal that seemed to be coming from our White House at the time of one-time social distancing for a period of time and looked at what that would mean with varying levels of reduction in the reproduction number, with the green being the largest reproduction of 60%, blue being 40% and red being 20% compared to black with no intervention. And the curves show the case numbers on the left axis and the critical cases on the right axis lagged by the delay in requiring critical care of several weeks. And what you see is that compared to no intervention, one-time social distancing can make a difference and actually in some cases can make a fairly large difference, but it's not a way of just getting a single peak. It always leads to, well, it's not a way of getting a single peak during the distancing. It either pushes it off into the future if it's short and or relatively ineffectual or you get a single peak and then it pushes it into the future, sorry, having a hard time reading my own figures. And then in the longest period we considered here, you can get a second peak later on after the social distancing has ended. You drive cases way, way down and then get a second peak after the distancing has ended. So that led us to consider the possibility of intermittent social distancing and there we wanted to look at the impact of seasonality on our results. I'm just trying to see, sorry, on my previous slide we were assuming that there was seasonality. And sorry, that was one thing I meant to emphasize and just missed. The width seasonality, the timing of a peak matters. And so if you have seasonality of the magnitude we estimated for the other coronaviruses, there are situations in which especially very effective social distancing at the beginning can push the peak off into a time that is actually more favorable for transmission and thereby increase the total area under the curve as well as the height of the peak. So if you believe that seasonality affects this virus, then one time social distancing is a potentially worse than nothing intervention, especially if it's very effective. If it's moderately effective then you sort of split the transmission into two curves and it can actually, you can get lucky with this red curve. But if it's very effective and you squash it down and just then to stop, then you get a bigger peak in the next cold time, the next winter and fall. And of course I don't have to tell economists that bad in the future is better than bad now, but it's actually worse in the future under the seasonality scenario. In the non-seasonal scenario you can't do any harm, you can only do good with social distancing. So then we ask the question if we're going to think about social distancing not as a one-time thing but as a multiple thing, multiple round thing, how would that look? And so we considered a scenario in which the goal of the social distancing is to protect critical care capacity and so we set some threshold at which you would turn on social distancing when cases, not critical cases, but when cases pass some threshold and another threshold at which you would turn it off when they get below a certain number and those are the dashed lines in this figure. And so what we find is that we have a period and you set those thresholds in order to sort of you work backwards from what do we need the thresholds to be so that we don't exceed the critical care capacity knowing that there's going to be this lag. So you have a period of on social distancing, a period of until cases get below this threshold and then you turn off social distancing, they rapidly increase, you turn on social distancing again and so forth and you can see just a little bit that the on periods get shorter and the off periods get longer and that's because the reproduction number is declining slowly due to the build-up of herd immunity. So even without seasonality if 20% of the population is immune that means that the reproduction number is down 20% from its baseline and so transmission is slower in the off periods and the reduction in cases is faster in the on periods. You begin to build up momentum but it's a fairly long and painful process over several years. If there's seasonality we saw in the one-time situation that seasonality could make things worse in the in the repeated situation seasonality actually makes things better because you have these periods of respite that create that allow you to stay open longer in the summer while not building up cases as fast and that allows you to then gain some herd immunity but while staying open for a longer period before you have to close again. And then Mark there's a question from one of the colleagues this is Philip Hartman on the previous slide what's the intuition for the necessity of the second green peak for highest reduction of R0? This one? Yes sort of the intuition here why is it? So the intuition behind all of this is that what regulates transmission is interventions plus number of proportion of susceptible hosts. So the interventions slow transmission but they also slow the acquisition of immunity and the decline of susceptibility in the population. So justice I mean and this is sort of the 1918 flu lesson as well if you put off transmission putting aside seasonality you've done good in two ways this was sort of the the flatten the curve idea that was circulating in in February and March of this year if you've done good in two ways you've delayed the bad bad thing from happening and you've also reduced the area under the curve but you don't but once you let up on social distancing or whatever the control interventions are the virus is still there and the people are still susceptible so you're back where you were at the beginning and then if there's seasonality you're in a potentially worse position that answer the question yes thank you yeah so that is really the premise of all of this is that that when effective reproduction number the average number of cases caused by one case exceeds one um you will have increasing cases that's how the models all work and when it's less than one you'll have decreasing cases the way to get it below one is either to intervene to to reduce contacts or to um have more immunity in the population so that each contact is less likely to cause a transmission because it's within an immune person or to have seasonal conditions just reduce the the efficiency of transmission those are the three influences that we consider so it's the the whole concept is premised on the idea that if we don't have immunity then there is no long term if we don't have herd immunity in the population there is no long term solution to continued spread because summer doesn't last forever and and the interventions aren't in place um that underlies all of this uh one part we found interesting and slightly unanticipated especially in advance uh about this model is that if you increase ICU capacity of course that's good for the the people uh and for the health care system but it also means that this regime if you if you have to follow it is um is more off time and fewer cycles because essentially you are now staying below a higher threshold and so you can afford to let more cases build up and so the acquisition of herd immunity here on the right uh up to the point where it would by itself stop transmission occurs more quickly and and if you overlay that with seasonality that's even more pronounced because you get all these cases below threshold in the summer and so so having greater ICU capacity is beneficial not only for the system but also for the for the progress of this kind of approach to social distancing and importantly and I'll come back to this at the end if you could have a treatment that that cut demand in half for the intensive care unit though a treatment for mild cases that that reduced their chances of progressing to critical that would have as you might imagine almost the same effect of of protecting almost the same effect overall so and of course would be much better because the goal if if you have to do this kind of project the goal is to build up as many immune people in the population with as few serious illnesses and deaths as possible and the the drug treatment would do that the the ICU capacity increase would be a less effective way to do that mark yeah mark so before you conclude on this few questions so one is in everything you're showing us you are of course assuming that there is no vaccine available in the meantime so it's all about building up herd immunity so just asking you whether that's correct and be whether you also sort of route it out that before the next season that's the next winter when when you get your second peaks and the simulations there will be a vaccine one question okay yeah the assumption is there's no vaccine I think the numbers people are giving for when there might be a vaccine are all over the place but with a seeming trend downward for which I'm cautiously optimistic the latest thing that was said by a scientist at our national institutes of health is that health care workers might have emergency use in the fall that strikes me as not impossible but extremely unlikely well very unlikely let's say and she said possibly general use in the spring I think that's conceivable if everything goes right and there is a good vaccine and and a lot of sort of international agreements come together to make it possible so I think it's that's the that would be unprecedentedly fast for an unusually for a not easy vaccine to make to make a good one of at least but I think widespread availability of a vaccine sort of even for a billion people say by the fall my is sorry by the by a year from now is it's going to be a challenge but yes that that is the assumption and I can take other questions I I don't have a slide although I should have I was just have been rushed in putting the talk together but but I do want to actually talk about sort of putting this more in context of outside the model what's the what's the reality and what might what might make the model wrong because I would like it to be wrong for obvious reasons so there are a few more questions coming in so one is indeed outside of your model so this is from Diana Garcia Lopez she's asking in your model you assume a large fully mixed population to what extent does the population structure matter are you or other research groups considering running individual based simulations maybe with explicit population structures considered I'm also thinking here that I mean a lot of discussion is to especially among economists of how to get the labor force mobilized again so if I were to sort of move away from lockdown to having the some of the social distancing measures still in place for the inactive population but maybe relax for the active population would that work and how have you thought about that or others working on this what would it mean for these simulations yeah so yes the assumption is of a well-mixed single population that makes things worse in general and if you include either structure or heterogeneity and risk or both it tends to reduce the total attack rates and people are doing that the imperial model which which has results similar to these for the overlapping cases they don't consider seasonality to my knowledge but they consider a lot more details about the population structure does get qualitatively similar results and others are also using more complex models which I think will get qualitatively similar results but if there is significant heterogeneity of various sorts such that some people just are very hard to infect then that would be all for the good compared to the scenarios that we're considering on the other hand if there's a lot of if there are pockets of people who remain uninfected by chance and then and then get infected that could sort of prolong the period when there's significant infection so it's mostly it's mostly in answer could only get happier as you incorporate more structure and then in terms of right in terms of sort of alternative approaches to getting out of this I think that is obviously the question of the moment I don't have anything model-based or or very intelligent to say about that right now except that I think there has been a strong feeling in the modeling community that the problem with the you know lock away the everybody over 60 and and just restart the economy with everyone else approach is that first of all although cases although severity is lower in in the young and healthy it's not zero and it's not even close to zero so that would be a potentially quite there would still be a lot of morbidity and mortality from that although obviously less and with the benefit that we would have more economic activity and then the second problem just about the practicalities of how well you can insulate if you're allowing widespread contagion in the in the under whatever under 60 population those people do have to have some contact with the at-risk people just to get them food and and care and medicine and all the things they need and so sort of letting contagion go while protecting the most vulnerable seems like a challenge but putting making that a harder number I think is going to be difficult and more likely we'll have somebody will try to do it and then we'll find out how well it works thank you mark then there is a question from Gerhard Rinsler this question is whether you and other researchers will be able to produce reliable estimates of the seasonality of the new virus within the next months ideally before the winter season I mean what kind of data needs do you have for that at what point in time do you think this is realistic that's a good question excuse me for eating I just haven't had time to do so it's going to be really hard because we're not in a normal year and we don't I mean in contrast to the seasonal coronaviruses we don't even have a decent proxy for cumulative incidents although we may begin to have serologic surveys that would help us to nail down that piece of the change but but assuming that there's ongoing control measures in place I think it's going to be really hard to say much about the seasonality especially if it's in this sort of 20% range I mean if it was 80% it would probably be quite obvious somewhere but 20% is consistent I think with with the fact that it's spreaded you know that it has been spreading everywhere in the world where it's been introduced to a first approximation and maybe a little faster some places and slower other places I think I think it's seasonality when you have everybody susceptible and weird behavioral patterns I think it's going to be really hard to tease apart it's a good question I haven't thought hadn't thought about actually trying to do that because it's it's it seems too hard but it's probably worth to figure out if it's possible yeah it's a question from one of our researchers thinking about future research so then there's um also a question still on your modeling approach um just to make sure we understand uh is this more more or less a closed model whereby you assume that these viruses will survive so there are these cycles but viruses don't disappear entirely and if so is this also what we should expect for this new virus so it's it's basically I mean it's the conclusion here that we should count on there will be another peak it's just a matter of how big and when so one time distancing is not enough but also we should count on this virus to be around for a while yeah I think so as a mathematical question the model is ordinary differential equation so it never goes to zero yes and so in that sense uh it can always come back in the model but that's a statement about the model um in terms of the uh in terms of the uh whether it comes back um in real life um you know we have two hemispheres so it's always it's always reasonably favorable for transmission somewhere um uh and so I think uh the scenario that we model with permanent immunity to this if that were so I think we might in fact have elimination at least for long enough so that we would not be worrying about racing to get a vaccine we could just get the vaccine well we'd probably be racing because we wouldn't know but but we might uh we might have real elimination but I think permanent immunity to coronaviruses doesn't seem to be uh the rule even even SARS and MERS coronaviruses the limited data that exist uh and I wrote about this in the york times yesterday uh with citations that's the new thing as you put citations in op-ed pieces it's weird but but true um uh the the MERS and SARS it seems that immunity begins to diminish in in two or three years at least the neutralizing ability of the serum from infected people so it's not um it's not uh reasonable to I mean it's reasonable to hope but it's not reasonable to expect that immunity would be permanent to this coronavirus but you know nobody nobody cared about coronaviruses until SARS uh and then they started to care and then got distracted and then MERS uh and people cared again and got distracted so I think you know the field of coronavirus immunology is probably 100 papers or something uh it's not it's not a well understood thing um and there is another question from Diana uh just a clarifying question um as to whether the viruses that you have considered in your analysis uh whether they also all have been uh present in europe yeah um uh I believe that all of these four coronaviruses are present in essentially everywhere HKU1 stands for Hong Kong University 1 and one of the other ones one of the alphas is called NL 63 which I assume stands for Netherlands 63 I'm not sure um uh so I think uh that all of these are and and actually most of the research on the immunology of this of those viruses was conducted in the UK so I think I'm 95 percent sure that all four viruses are essentially everywhere more seasonal in the north and south and less in the tropics okay um so you are on your concluding slide right yeah uh well I have uh I have some other things to talk about but yeah please maybe I'll so the conclusions you've now seen uh and I've already said I think it is worth just sort of saying um some of the things that were in the discussion of this paper that I should have put on a slide but just didn't get to um the first is that uh we wanted to be clear that we're not saying that this multiple rounds of social distancing is a recommendation or a or a good idea or or in any way desirable uh except in the way that it seems to be one tool to keep the health care system intact while not uh while not being closed all the time but we recognize that it's both probably practically challenging and a lot of time on social distancing so this is a sort of epidemiologic analysis in the absence obviously of considering the other impacts uh and therefore can't be an all things considered recommendation but rather a recommendation that one time social distancing doesn't accomplish the stated goal and that multiple rounds would uh but but as we say in the paper we're really trying to stimulate both efforts to bring the problem to an end through faster vaccines and also creative solutions to other types of ways to reopen such as the ones that have been talked about briefly here or perhaps serologic passports as one option or or other things oh and then and then I promised at the beginning to say say something about other ways the model could be wrong so I think these these numbers about the relative proportion of severe and mild illness look like it's already very heavily weighted towards mild but but I think it's worth mentioning that until we have serology there is still the former formal possibility that there's a lot more infection and that this is actually a larger fraction even than than we assume I don't think that that's the case that's why we assume these numbers but but the way I hope that we're one way I hope we are wrong is that is that there are in fact a lot more immune people in the population than we think because they've just slipped through the this whole cascade without ever being detected I just saw on Twitter this morning report from the Netherlands of some some new serology which I assume is probably being done by by Erasmus which is a very careful lab and so I tend to believe it that suggests that we haven't been totally far off in our estimates but but I think that is an open question still until we have more more good data and I very much hope that it's that there are more cases out there that we don't know about the most sensitive piece of the assumptions although there are obviously many others yeah there was some piece put out by the Netherlands but I had a home blood bank donors where they had also tested whether there were any positive antibodies and three percent of the population showed that and so there's numbers are indeed quite high um there's one more question from uh Laffarad Mosby um this is more of an economist question um should we think of COVID-19 based on your experience as a one in a 100 year catastrophic event the reason is asking this is um this would inform us of how much we should invest in future preparedness um I don't know how to answer that in the sense that if you take the last event that was similar to this magnitude I think it is the 1918 I don't know what it was like to be around in 1957 I don't think it was it was quite on this scale but the question is whether that hazard is constant over time or whether it's accelerating certainly people in my field spend a lot of spend a lot of time talking about reasons why the trend might be for this kind of thing to happen more with global travel and with with destruction of habitats that lead to zoonotic emergencies of emergencies of zoonotic disease and increasing drug resistance and also I think I mean I've been thinking about this 1918 maybe was a worse virus than this one all else equal but my sense is that we didn't have an intensive care we didn't have intensive care to protect and so we didn't there wasn't this you know this is a sort of hard thing to think about maybe economists are better at it there was no there was no sense of overwhelming the health care system because there wasn't that much of a health care system at least it was effective um and so now we invest in protecting something that we didn't have in those days so I don't know how to think about sort of as you build up capital that you want to protect whether that changes your investments so I think I would sort of probably go with my professional orientation and say some of the causes of this are accelerating and most of the causes I can't think of any causes of it that are slowing down or that are reducing the hazard every year so I would think that you know if this interval was 100 years given the number of near misses with SARS and MERS it's reasonable to expect that it's accelerating rather than one in 100 years but but I think you know we're getting soft here so we're coming to the end of the talk so I just wanted to ask one last question what what's your professional advice to the economists community to economic policymakers but if you're one one line one you're one liner gosh uh I think you've left me so speechless I don't know what to say I think I think we need solutions please please do listen to your public health colleagues when they say something's impossible and push back on them push back on us but but also listen so my interactions with economists have been some very positive and some very extremely positive and really trying to be constructive in finding solutions that are both possible and ambitious and then with some others they have been more like sort of dismissive in the sense of public health people just think small and don't realize that with enough money you can do anything and I think that's sometimes true and sometimes not but but we should have a discussion about it rather than assumptions but that that's a little bit based on some individual interactions but we're going to need your help desperately because obviously everything we recommend has huge economic consequences and and we got to work together thank you mark a previous speaker earlier this week in the series said that at least for the time being this we play second pedal to you and your colleagues as economists and we are learning a lot that would of course be much better we can work together especially when we get to the recovery phase so I want to thank you very much on behalf of all the colleagues at the ECB and yeah this is of course a reality and a rather gloomy outlook for us but that's what it is so we will circulate also your article together with your presentation and thank you very much for doing this stay healthy thank you thank you all of you as well and I'll share the slope it was a very rich discussion but I think there are a few other pieces that may be interesting and and point have references that people can and look up if they're interested we're going to look at that thank you very much