 Hi, this is Luke Leven. Welcome. Thank you. So thank you for doing this. I will introduce you and moderate the talk. And so the colleagues will, there are about 30, 40 colleagues sign up and they will ask them to mute first of all. And then they will be able to ask questions through the chat button and I will then consolidate those and direct them to you. So we plan normally on like 45 minute talk and then 15 minute Q&A, but we'll take it as it comes along. So with that, this is another COVID-19 webinar. Today we are welcoming Jessica Metcalf. Jessica is an assistant professor at Princeton University. She's an expert in demography, disease dynamics and public policy. So we're very much looking forward to learn more about what's behind this huge sun boom that I see on the screen. So that's over to Jess. Great. Thanks very much and thanks for inviting me. Can you see my slides? Yes. Great. So sometime in November or December of 2019, a coronavirus spilled over from a zoonotic reservoir into human populations in China. To understand subsequent spread, epidemiologists use two key values, one of which is referred to as R0, which I'm going to illustrate using the value of 2 and the other of which is the serial interval. And so as a thought experiment, imagine that an initial case has dropped into a population. Then one week later, that initial case will have infected two individuals because R0 is 2 and two weeks later, those two individuals will have infected four and so on. So with an R0 of around two and a serial interval of about a week, what we expect is that the number of cases will double every week. And this leads to effectively exponential growth. And exponential growth is what we've seen for this outbreak. So I'm showing the trajectories of confirmed cases on the bottom and deaths on the top for countries across Europe. One of the important things to remember about these sorts of graphs, which you've seen many of, is that we're seeing the impact of effects that happened two to three weeks in the past, because it takes about two weeks for cases to enter the system. And it takes about three weeks for people to die. So all of the actions that we're taking today, all of the lockdowns, all of people staying at home will be seen in two to three weeks in the future in terms of their impact on the epidemics trajectory. Of course, you might be wondering, well, why is this pandemic so much worse than anything else we've ever seen? It's a major public health issue, I think, for two reasons. The first is that it combines asymptomatic transmission. Estimates suggest at the moment that maybe 30 to even 50% of cases are not showing serious symptoms, although the exact concept of asymptomatic is a nebulous one. And this means it's very hard to contain the infection. In 2003, the SARS pandemic that occurred that kicked off also in China was containable in part because individuals who were transmitting the infection were also showing very serious symptoms. So it's very easy to identify them and to slow spread down by containing them and reducing transmission from them. In this particular pandemic, the virus can travel invisibly across very long distances. But of course, if it was just traveling invisibly, this wouldn't be a problem. The reason it's a problem is that this invisible transmission, a fraction of the cases is combined with very severe outcomes for many, which can overwhelm health systems, as we've seen in China and now in Italy. The main correlate of severe outcomes based on the early trajectory of the pandemic seems to have been age. So this shows you data on the case fatality ratio for 44,000 cases from CDC China and about 4,000 cases from the CDC US that came out more recently. And you can see that the two lines look pretty similar. As you increase in age, the case fatality ratio increases to going up to 20% in individuals of around 80. And I want to stress that this does not mean that if you acquire the infection and you are even 60 years old, you have a 5% chance of dying. Because cases are different from infections. Cases of the individuals who turn up with symptoms and bump into the health system and therefore get counted as cases. But as mentioned on this slide as well, many infections may be asymptomatic. So epidemiologists distinguish between the infection fatality ratio and the case fatality ratio. And the infection fatality ratio by definition will be much lower than the case fatality ratio. Effectively, by focusing on the case fatality ratio, we're ignoring a huge sort of lower side of the iceberg in some sense. There's a huge number of individuals who are infected who are not entering the denominator of those estimates of the case fatality ratio. This is an estimate of the case fatality ratio that was produced, sorry, of the infection fatality ratio that was produced by the imperial team relatively recently by integrating data on the case fatality ratio and what was known at the time about rates of asymptomatic carriage, et cetera. Of course, we learn something new about this virus every day. These estimates are uncertain and these estimates are changing. But one thing that does seem to be consistent is that the pattern of the trajectory of mobility and mortality seems to increase with age. And early on in the outbreak that gave us one sort of lever to pull to start making projections, one thing we can identify is that clearly in older populations, given this pattern of the case fatality ratio or the infection fatality ratio, we expect more severe outcomes. So we recently used this combining county-level demography across the United States based on census data with what was known about the number of hospital beds to try and identify which counties in the United States were most at risk. And you can see that almost all counties we were expecting under a range of assumptions, and of course you have to make assumptions, and predicting the future is rather hard, but if you assume that 40% of the population will be infected and 80% show symptoms, then the number of hospital beds was exceeded in almost every county in the United States. The gray areas on this plot indicate that there are no beds in that particular county, no hospital beds, and so the cases were allocated to neighboring counties. So this is what happens in the short term, and how one could have projected need over the short term just based on this age profile. The next question is, of course, what happens in the slightly longer term? And here we can turn to another set of concepts from epidemiology, which is known as the Susceptible Infected Recovered Model. And what this captures is the fact that what we've talked about so far is susceptibles becoming infected, but we haven't accounted for the fact that once you've been infected, it's very likely that you'll be recalcitrant to infection for at least a little while, so you'll enter a recovered state that will not be vulnerable to infection for at least a while. And this will have a number of dynamical consequences on the trajectory of the outbreak. So if you imagine, again, we're going to drop one infected individual into a population, red dot on this slide, and in this particular example, I'm going to assume that R0, this magical number, which is the number of new infections per infected individual in a completely susceptible population, this time I'm going to estimate, I'm going to assume that it's around three. So I had two on the previous slide, I'm going to use three on this one. Estimates for SARS-CoV-2 to date range somewhere between two and three. So if I drop that one infected individual into this population, where 100% of individuals are susceptible at the start, what happens is that that infected individual will infect three others, and the number of the susceptibles will fall, this will repeat, obviously the three others will go into infect another three and so on, and the number of infected individuals will rise. It will come a point however, where the number of recovered individuals grows to a point where the number of susceptible individuals available for an infected individual to infect is actually quite low. So you can see in the cartoon below the graph here, if you have one infected individual, they're sort of inverted commas wasting two of their infections on individuals who are recovered and therefore recalcitrant to infection. And so since this one infected individual is infecting one other, or potentially even less than one other, what we expect to see is that the number of infections over time will decline. And that's shown by that red curve reaching a peak and then turning over. And as you proceed forwards, you reach the stage where the number of infected individuals falls to nothing, but there are still some susceptibles in the population, and these are individuals who are indirectly protected via the fact that the individuals surrounding them are recovered. So if you were to drop an infected individual into the population, they'd be unlikely to be infected by them. This broad trajectory clearly depends on the r-naught of the infection. So the height of the peak will be shaped by how transmissible this infection is, how many new infected individuals result from one infected individual. And it's also the number of the total height of this peak is one of the things we've been most concerned about as this trajectory has unfolded. And so many of you will have seen this concept that's been floating around referred to as flatten the curve. And the idea is that once the number of cases exceed a certain threshold, health systems are overwhelmed and we see the sorts of outcomes that we see in Wuhan, or in northern Italy, or in Italy in fact. But what we can do is we can work to try and decrease the height of this curve, so as you can move it to a place where the amount of time spent above that threshold, where health systems are overwhelmed, where just the sheer number of cases far outstrips the number of hospital beds, by taking steps including case-based self-isolation, social distancing, public events banned, etc. And these work because what you're doing is effectively you're reducing the R of the population, you're reducing the number of infected individuals infected by one infectious individual, simply by the fact that individuals are not encountering each other. And the benefits of this are one that we hope to protect our health systems from being overwhelmed, but secondly we're buying time and that time will give us the opportunity to test the array of therapeutics that might help reduce the outcome, the negative outcomes for this infection, which will also reduce burden on the health system, but also will give time for the development of a vaccine. Now realistically I think all experts agree that it's unlikely that we'll have, well that's not true, all experts never agree on anything, but it's unlikely I think that we'll have a vaccine before 12 to 18 months from now, it just takes time to develop vaccines. Okay so this gives us, so I spoke about the short term, that rapid exponential growth. I spoke about what we might expect in the longer term, one thing that's received a lot of press is whether we'll see an effect of seasonality on this infection, whether as summer comes around we can expect a change in the transmission dynamics. Now it's true that coronaviruses are winter pathogens and we think that, so this is the family from which this SARS-CoV-2 is from, there's a number of other of them, there are about four that circulate quite regularly in populations in the United States and Europe and manifest generally as the common cold. So we know something about those and those seem to be winter viruses and it's expected that the reason that they're winter viruses is there's something about reduced humidity and lower temperatures that may increase their transmission, so it may be something about the fact that drier air allows droplets to be suspended in the air longer, it may be something about what drier air does to our mucus membranes, we don't know the exact mechanism but it does seem likely. And so if that's the case, if that also applies to this coronavirus, which we don't know for sure but seems not implausible, what we expect is they will be an outbreak in the endemic situation, maybe an outbreak every winter to here. So does that mean that we expect the summer will reduce transmission? I think the chances are no and the chances are no because the difference between that sort of endemic setting where you're getting a regular outbreak every year is that that's a setting where the susceptibles have been depleted by previous years, where nobody in the world prior to the introduction of this pathogen late last year had seen the pathogen, so every single person in the world single person in the world was likely to be susceptible to this pathogen and this huge pool of susceptible individuals will overwhelm the effects of modulations and transmission driven by climate change. If you remember the previous cartoons, the number of new cases in each instance is predicated on infected individuals encountering susceptible individuals and then transmission occurring. If there's just tons and tons of susceptible individuals that makes the risk very large. So you can do the thought experiment to suggest this is true but you can also turn to history and maybe our best guide for pandemics from history is previous influenza pandemics. So this is a figure from a paper that appeared in New England Journal of Medicine in 2009 looking at the 1918 influenza pandemic. This is the plot from Copenhagen and the red bars show you where the typical flu season is and the blue bars show you the percentage of total mortality occurring in the periods and you can clearly see that this in 1918 in Copenhagen the influenza pandemic appeared in July and sort of bused through in September and November and so occurred outside of the typical flu seasons. So to expand on this thinking our group in Princeton has been working to put together with climate scientists at Princeton to try and understand the nuances of this a bit more and what we did was we took parameters describing the climate dependence of all known coronaviruses for which estimation was possible but also influenza so all of these winter viruses with climate models of the trajectory of humidity and temperature all around the world. And so this map shows you the week of maximum specific humidity in different locations in the world from some January up to well up to January because that's how seasons work and so if you do that you plug that into a model of infectious disease dynamics what you find is that focusing on cities all around the world there's very little difference in the timing of that first peak so that's the left hand peak on these on these trajectories and shown for New York, London, Delhi, Melbourne, Buenos Aires, Johannesburg etc. Once you start looking at the second peak so we do expect a second wave of this outbreak because just to put this in context once there has been a first peak of this outbreak not everyone will have been infected necessarily and therefore the people that you know the susceptibles aren't going anywhere so they're still at risk of infection. The second wave is slightly altered in timing possibly by climate and we would also expect that control efforts that reduce transmission will increase the relevance of climate but overall it's sort of trying to ban the problem by looking across the full span of what we know about winter varices nothing to us suggests that we will see a reduction in transmission over the summer. Okay so what can we expect for the coming months so I think you know this is plotting that same data from the Johns Hopkins page on the global numbers of cases we've seen in many different settings that the interventions do seem to have flattened the curve and in fact the number of cases are declining in many places and there are lots of issues with how we count cases and how testing has been wrapped up in different settings but this is at least suggestive that the interventions have been working but of course as many of you are probably thinking about such interventions are incredibly costly to sustain or seem to be costly to sustain and so a really important next question is how do we what do we need to know to start setting a pandemic recovery trajectory and I think there are two big gaps in this in the in around setting a pandemic recovery trajectory there are probably many others but the two that I've been thinking about a lot is the fact that we don't know how many new infections are possible we don't know quite where we are on that epidemic curve if you think back to that plot I showed you of susceptibles infected and recovered and this is in part because we don't really know how many cases there have been because the number the fraction of asymptomatic cases makes it rather hard to estimate that we also don't know which measures are the most effective so what can we do to start grappling with these two questions so if we go back to thinking about the trajectory of the susceptible infected recovered plot I explained that individuals start susceptible and then after a certain amount of time they bump into an infected individual and they become infected and then following infection individuals can recover and the data I plotted for you on the first couple of slides was the numbers of cases and the numbers of deaths and that reflects what we expect to come from this infective class but of course this provides an incomplete window onto the pandemic it doesn't tell us anything about the individuals who are susceptible and it doesn't tell us anything about the individuals who are recovered and since cases will definitely be under reported because of this issue of many individuals who are infected showing no symptoms and even deaths are potentially under reported because of misdiagnosis of deaths because of people dying at home and because we don't know how deaths translate into infections so it's actually very hard to map from one to the other we have a very incomplete window onto the epidemic if we just have these two sources of information and with a colleague Brian Grenfell we wrote a paper when was it was like 2016 perhaps where we pointed out that there is actually a way around this one could use serological surveys to try and understand the changing global landscape of infectious diseases so what do I mean by that serology is the measurement of antibodies that are found in our blood and in our saliva and these are markers of the immune system so these are molecules that our immune system generates once one has been infected by a pathogen and so I could collect blood samples from all of you and it would give me information for example on whether you had experienced previous flu pandemics so in particular I am unlikely to show antibodies to the Hong Kong 68 influenza pandemic because I wasn't born then but if we look at more recent influenza pandemics so the 2009 flu pandemic it's very likely that there would be markers in my blood that indicated that I had been exposed to this infection so we can get this window into the past of infections around the globe but we also so which gives us a sort of invaluable readout of you know where how many cases there had been it also gives us a lot more nuance onto what might be happening in this particular pandemic so just to illustrate this if you imagine you have a trajectory of cases over time and these will be underreported so we'd see the number of cases there shown on the left in red but in the current way that surveillance works we will not see the number of the proportion of the population that's susceptible we don't know who is at risk of becoming infected and if you imagine that it's relatively early on in the outbreak and you're trying to fit models to this data you could fit a model that looked a lot like the early trajectory but that corresponded to much lower transmission and a longer infectious period right so it maps perfectly to the first half of that curve but it goes much much higher so you would be anticipating way more cases would be possible and the serology would give you some power to discriminate between these two scenarios because you'd see if you had a readout of that solid blue line of the proportion susceptible in the population of different times you might be able to identify the discrepancy earlier conversely exactly the same data you might fit a different trajectory which would show you that that would suggest that there was high transmission but a short infectious period and this would deplete cases sort of earlier sort of would deplete cases more rapidly earlier but then would settle out to a much lower level because the infection would burn it out much faster burn itself out much faster and so data on serology will just give us one more way to triangulate on what's going on under the hood as these epidemics progress there's been an awful lot of debates I'm not sure if you have you have all been seeing it as people are starting to do serological surveys there was one that was just published out of Santa Clara in California and the author suggested that from this analysis that 50 to I think it was 600 cases that had been reported had actually occurred and there I think there are quite a few that study was rich in flaws in the ways that they did the estimation but it suggests you know this is why people are so keen on doing it it's because it gives us one way of handling an estimation of the cumulative numbers of cases that we have observed because that's one more triangulation and unless we know that it's very hard to know where we are and the reason that's one of the reasons I mean one it's interesting to know how many cases have occurred already but the reason that it's a question of such burning interest is that recovered maybe recalcitrant to reinfection not just for a short period but maybe up to a year possibly even up to two years so to be seropositive so to test seropositive on one of these serological assays could mean that you were immune to infection so you could go out and live your life without being a risk either to yourself or to others we don't know that I want to stress absolutely that we still don't know that for this infection but we do have evidence from other coronaviruses so previous other coronaviruses so which are in the same family not identical to this one but in the same family people have done experimental tests on even on humans and shown that it's impossible to reinfect individuals who have been infected within a couple of years so that immunity is long-lasting sterilizing immunity is long-lasting people have run mathematical models Mark Plipsich's group in particular has done some really lovely work in this area and looked to see signatures of immunity and so aspects such as biennial dynamics where you have a big outbreak one year and then a small outbreak the next year and then a big outbreak the next year these features suggest that there is some immunity in the population and that individuals who were infected in that first big outbreak year were protected the next year which is why you get a smaller outbreak finally increasingly we have data measuring the profile of the various immune molecules and the bloods that people are showing after infection with SARS-CoV-2 or COVID-19 and this looks a lot like many other infections that we know in this space so we see the sort of same sort of trajectory that we would expect there doesn't need to be anything super weird about immunity to this infection in individuals who've recovered so all of this provides some reason to hope although again to absolutely emphasize this is not certain that this is at least suggestive that individuals who have been infected will be protected from infection for at least a little while and you can see how this is just extraordinarily important to resolve because if this was the case then what we could do is we could test healthcare workers we could figure out whether other essential professions could go back to work this would be really game-changing in terms of how what the pandemic recovery trajectory looks like but again to absolutely emphasize we don't know that this is true yet it's going to take a while to figure it out we're going to need measurements on individuals who have been infected and recovered we're going to need to track trajectories very carefully to see whether individuals who are regularly exposed do not seem to be showing signs of infection again you wouldn't want to send someone to work on a COVID-19 ward if you weren't absolutely certain that that once they've been infected they were not at risk of infection again so this is work that's underway all over the world and people are working on this problem there's a huge and I think a new serological test comes out pretty much every day but the hard work well that is hard work the second piece of hard work of identifying what being seropositive actually means whether we're measuring the right antibodies to try and understand whether they map to immunity is underway and I think you know unfortunately this is not going to be the last pandemic we experience and just to hark back to that initial call I think if we had sort of put some money ahead of this towards a global immunological observatory where we had a combination of you know convenient samples blood banks are collecting blood all the time we could be testing that for immune markers we could be combining that with really detailed longitudinal age serology and combining this with efforts you know laboratory techniques it's now possible to take a single drop of blood and test for every single virus that has been known to infect humans to see whether your blood shows markers of being having been exposed to them and you could bootstrap this of course the interpretation is still a question but you could bootstrap this with statistical approaches and modeling approaches and you'd have a much better sense of where the world was in terms of our landscape of immunity and this would have put us in a better place I think ahead of this particular pandemic you would have been in a place where you could have checked for anomalies you could have seen how far along you were much much earlier no country in the world I mean although it's true that the signs were there and there are many other reasons that preparedness failed but this I think would be an important piece of starting to think towards the future so that I suggested was one of the first big gaps is actually knowing knowing what immunity means and having a sense of where we are on this epidemic trajectory how many cases are still to come and the other piece I think that is still a sort of large gap is thinking about which measures are the most effective so as mentioned previously it is clear that interventions have flattened the curve just here's an example from Germany in different countries have implemented different policies and they range across the spectrum from case-based self-isolation mandated to lockdowns being ordered the we can take this trajectory and turn it back into the sorts of cartoons that we that I showed on the first slide and so the red slide the red curve here might be what happened if you did nothing at all so if you let the uncontrolled pandemic occur and the blue curve might be what happened if you implemented the suite of different policies and almost all countries have implemented almost all policies often in mildly different ways with mildly different timings and perhaps with a spectrum of intensity around how strongly they're implemented etc and I'm sure that in the wake of this outbreak there will be a stream of country comparisons that try and identify which the most important which the critical interventions were but I think there'll be so many other things that vary between the different countries in the different context it'll be very hard to interpret what's going on and the issue is that we don't know whether some you know all of these these interventions are differentially costly and they might also be differentially effective but we don't know which world we're in many countries especially in Europe are now starting to think about loosening and of course what you particular policies that have been implemented to again just to remind you these are all designed to reduce contact between infected individuals and susceptible individuals and they work simply by reducing contact between any individuals at all but as we talk about loosening which is increasingly looking like it's necessary well it's necessary it's increasingly there is appetite for loosening across many settings as we talk about loosening of course what you'd want to do is you'd want to relax the interventions and the policies that were the least effective and the most costly and you'd want to maintain the ones that were the most effective and the least costly but we don't know which do what so the question might be how do we evaluate these policies and to some degree because we know what the mechanism of transmission is its encounter between individuals a lot of mathematical models have kind of followed through the logic and tried to infer what the impacts are but I think there's also potential to step into a space where you actually test loosening specific restrictions for short periods of times in some locations this will be happening everywhere anyway and so it might be helpful to try and do it in a systematic fashion of course ideally both systematically and using randomization and monitoring outcomes in pilot and comparison areas clearly as part of this process very careful ethical deliberation is important such approaches such randomization and application are only appropriate where there's uncertainty about intervention effectiveness but there is precedent so the Ebola vaccine trials had made some really important progress in thinking about how you design such trials in contexts where it's very severe outcomes are occurring so what this plot shows here is a scenario where the red curve again is the uncontrolled outbreak the blue curve is the outbreak that occurs when you throw absolutely everything you have are trying to control the the the outbreak and then the green curve shows what happens if you were to release the out if you were to release that particular intervention for just two weeks so say for example you reopen schools for two weeks or you lowered a restriction on lockdowns and some settings for two weeks and you could then track what happens to the numbers of cases in settings where you'd released it in settings where you hadn't and it turns out because we absolutely have very crisp definitions of what the mechanisms underlying these processes are we know that to compare different restrictions to each other so you might release some in some places and others and others all you need to do is measure current infections so infections at the end of the window in the green versus the blue curve will tell you how much that particular intervention is doing to reduce transmission if you wanted to go one step further you could also measure the numbers of infections but also the proportion of the population that was susceptible and if you did this you could estimate the absolute magnitude of transmission and this could give you a sense of what the whole curve looked like and it would also give you a sense of what the future trajectory was as I mentioned previously in the current setting where we don't yet have a vaccine the susceptible individuals are not you know to some degree not going anywhere so understanding what fraction there is will give us a sense of what the trajectory is across the future the of course the reason that such such evaluation and implementation is required is because the interventions are not just beneficial in terms of reducing infections but they're costly in terms of economic and psychological outcomes so ideally this would be measured in tandem with measurement of economic and psychological outcomes that are occurring in populations so as an example of one such you know one of the key uncertainties here I think it is increasingly evident that children do get infected it's possible still that they're getting infected at lower rates than adults although we still don't really understand that we don't know whether individuals sorry children are transmitting less than adults so they seem to be transmitting a bit and this is really at odds with our experience in previous pandemics so for example in the influenza pandemics children were an absolute driver of transmission so closing schools was absolutely central in those settings whether that's the case here but the degree to which closing schools helps is it's still uncertain I mean and it might it might be really important but it might not be so important and we just don't know yet I think that the of course this works both ways you could both test listening specific restrictions or you could also test tightening up specific restrictions to see what the gains were and I've shown an example here where the tightening or loosening occurs at the start of the outbreak but the logic holds at whatever point across the trajectory you're experimenting with tightening or loosening the restrictions and I think it would just give us a much better situational awareness of where we are in terms of these non-parmaceutical interventions which are sort of governing our lives at the moment of course something that has become evidence particularly in the context of psychological and economic outcomes across this outbreak has been that disadvantaged populations are bearing the brunt of it in many settings and it's also probably the case that low and middle income countries are likely to be affected even worse than the countries which we've been thinking about the most so far it may be that social distancing is more difficult because of because of housing situations because of the way living conditions work because of how access to food works I think it's likely to be really important to start thinking about how one engineers efficient spread of trusted information in these settings particularly across healthcare workers etc I think the it's very hard so this is not so much the case in Europe but in many countries in Africa it's very hard to imagine really much one can do in the face of the coronavirus outbreak beyond trying to ensure that individuals are economically safe so I think economic safety nets are going to be a really important part of the intervention around it and there are things we just don't know so the backdrop of health conditions that the pandemic is running across remains very unclear and this might interact with COVID-19 in ways we don't understand yet so we don't really know what'll happen if it comes into contact with populations with malaria or anemia and this might be you know a whole different set of headaches around this pathogen and the trajectory it takes across the world so I think that you know as moving forwards some of the things it would be really nice to see it would be models that titrate the roles of demography so the age structure of populations also obviously matter seasonality which I think will not matter in the near future but might matter as we get our act together in controlling this infection and non-pharmaceutical interventions to particularly to guide vaccine and therapeutic global deployment so I think trying to ensure that equity is maintained to some degree as this virus rips across the world could be informed by models that try and get to grips with this. I feel that serological surveys will be a critical part of grappling with where we go and this is I think widely recognized within the public health community there are serological surveys underway all around the world and again the strength of these approaches they tell you whether you've ever been infected it's not just a test of recent infection like the test for the viruses that have done in many settings so a population representative survey would give us a window onto past incidents and tell us where we are on the epidermic trajectory and how any more cases are still to come. We'd also need to put in tandem with that any convenient samples we can get to grips with from bug banks to anything else as well as longitudinal data so repeated measures on the same individuals and this is really important because we need to be able to interpret these measures we need to be able to figure out what they mean in terms of what we refer to as protective immunity so you might have you know you'll generate an immune response to the virus once you've seen the virus and it'll be to some degree protected if you manage to clear the virus but whether it remains protective as time goes on whether you're protected from the virus over the next year or so it's just still very unclear and that's something that we need data to help us get to grips with. I think that you know an internationally coordinated global immunological observatory would be a really important part of this I think one of the things that this outbreak has made clear is that you know the health of anyone is the health of everyone and it takes that sort of large vision to get a sense of where we are in terms of this landscape of immunity this landscape of risk that we sit in as infections spill over from anything from you know industrial chicken raising to zoonotic infections who knows where the risk will come from and finally I think that the non-pharmaceutical interventions have been extraordinarily important and extraordinarily effective in reducing the burden on our health systems and you know helping countries maintain a steady course through this but as we start thinking about what the next steps are I think we're really we're sitting on the edge of squandering opportunities to learn what works best in these settings these things will be released when we are another and we should be doing the best we can to learn to to extract every drop of information we can from this process as we move forwards and so that's that's what I have for you I just like to thank briefly I've of course done this work with lots of different people from the sorority Michael Miner at Harvard Brian Grenfell here I've worked with Johannes Houseover an economist at Princeton here on the evaluating non-pharmaceutical interventions and then a number of graduate students and uh geophysicists at Princeton uh thinking about climate drivers and low and middle income country projections so and that's what I have thank you very much many thanks just Jessica this terrific um a couple of questions have come in and one question um from Florian Haider relates to the first half of your talk um is asking whether you could comment on the possibility of cross immunity across coronaviruses uh for instance why does not everyone in a family get infected so that's a great question um I think there is some evidence of some cross immunity in some coronaviruses right so to um sort of uh provide a kind of metaphor for how this works the coronaviruses might broadly look the similar to your immune system so if your immune system learns to protect you against one it might protect you to some degree against another this is quite common for many infections uh so to date and then this is obviously something that people have to worry a lot about as they develop the serological assays right you don't want a serological test where you're trying to identify whether people have been infected with SARS-CoV-2 and have it throwing up you know positives which reflect just any version of the common cold um I have not seen any convincing evidence yet of cross protective immunity I'm sure people are looking very hard the question of why household infection rates are so low is a very interesting one um so estimates are I think around there's been about two estimates and they're around 15 percent so what that means is that if someone is infected in your household your probability of being infected is 15 percent I went back to this lovely paper published in 1952 by a someone called Hope Simpson who's looking at measles mumps and rubella or no measles mumps in chicken box um and he you know measles is one of the most transmissible infections in the world um with an r-naught of around 15 and although children in their households had a risk of infection of around 85 percent individuals aged 15 and older had an infection rate uh probability of being infected of around 10 percent right so I think you know we have to calibrate for the fact that the two estimates that we do have for household transmission are um in settings where people knew the infection was circulating we were possibly being careful anyway um and that you know children are just uh little vectors and if this virus is doing different things than children it might just be partly that so but but it's a good question cross protective community is a really is a big question but I just haven't seen anything on it yet good um next question is whether um you are whether you're aware of any serological surveys done in low to middle income countries because most of what we are reading about is in in higher income countries so I think that jack ma has distributed uh serological assays in many uh has donated them to many african countries I'm not um entirely up to date on what the sensitivity and specificity so these are two key measures of the precision of these tests are nor am I aware of any um estimates of what the seroprevalence looks like um my suspicion is that uh the the word that we're still at the relatively early phase of the outbreak of many in such countries but I just I don't know I would um were I designing a survey in such countries I think I would because of the shortage of tests relative to people I think I would focus it on healthcare workers as a kind of canary in the coal mine um you know as and I think that would that would be incredibly informative I think it'd be really really interesting to have a sense of what that looked like and see how the cases search so longitudinal samples on healthcare workers is what I might suggest thank you um and then uh there's another question about herd immunity in in europe um do you have any estimate of what the what the current level of herd immunity already could be in europe it's something like higher like like the assumption that you use in your own modelling at the start of your talk or so the so the concept of herd immunity just to step back is this like is the idea I was lining outlining in the the first couple of slides right which is this idea that you're protected indirectly by the people around you who are protected and um you know one of the great tragedies as far as myself and other people who have come back seem preventable diseases has been in the context of this outbreak is that people are now using herd immunity like it's a bad thing it's a wonderful thing it's an incredible social good um it's the but the way you want to get there is by vaccination uh the um the way people offer and interpret it is as being the point at which the number of cases starts falling and we know exactly when that will be which is um it's defined as one over our naught so it would be 50 of the population if our naught was two okay and and you know slightly higher if our naught is slightly higher the um the issue is that uh so so I just want to also emphasize that you know the cases start falling but they can fall for a very long time I mean the cases in Wuhan have been falling for you know there's still cases happening you still get sort of dribbles of trails of of contact tracing so sorry of chains of transmission are still occurring so it doesn't mean that once you've achieved that number you're out of the woods either um it just means that the outbreak is forwarding I think that every uh so there's been a couple in Germany which suggested that maybe um 10 to 15 percent of the population had been infected in them in very severely affected locations I don't think that's going to be enough to start counting on for indirect protection of individuals um but I think we need more serological surveys you know it's we're still in a very uncertain part we're still we still don't really know um then another question is about the second wave that you showed the one of your simulations I believe you showed one and a half years from now would be the second wave so not this but next winter is this more or less your best estimate of when you expect the the next second wave um so I that's a very hard thing to say I wouldn't I wouldn't put money on it so the the nuances of how that plays out will depend on those models that I showed have this assumption in them that the populations are what we call well mixed which means that every individual has an equal probability of encountering every other individual that's clearly not true um and so getting at the exact timing is likely to be quite nuanced and it will also depend on those simulations I showed were completely uncontrolled right and so if you have a completely uncontrolled outbreak that means it tears through the population and affects just about everybody which means that you actually are in a situation where there is it's going to take a long time before there's sufficient build-up of susceptible individuals new birth-centering population before you have enough susceptibles to allow a new outbreak to happen so um I think it's a tricky question I would not put my money on two years necessarily I think Mark Lipsitch's group recently um had some estimates which I can't remember off the top of my head but I think there's uncertain there's a lot of uncertainty around that I think so this is actually one more thing is that you know what the the there will be a second wave if people just relax um the interventions that are in place you get that's a very fast way to get a second wave right that's um okay um well Jessica I most of the other comments I can see are great talk nice survey so all compliments um thank you very much for doing this for us and um we reached sort of the end of the talk now I wanted to just thank you and all the colleagues with whom you have worked stay healthy you see yeah um so we will be sharing the presentation as usual but everybody we will upload it internally on the website um and then we take it from there so all the best thank you thanks very much