Coronavirus (COVID-19): Press Conference with Marc Lipsitch, 03/04/20


You’re listening to a press conference from the Harvard T.H. Chan School of Public Health featuring Marc Lipsitch, professor of epidemiology and director of the School’s Center for Communicable Disease Dynamics. This call was recorded at 11:30 am Eastern Time on Wednesday, March 4. 

MARC LIPSITCH: Good day, everybody. My name is Marc Lipsitch. I’m a professor of epidemiology and the director of the Center for Communicable Disease Dynamics at the Harvard Chan School of Public Health. I have been following this coronavirus very closely, drawing on a background of having worked on SARS in 2003, pandemic flu in 2009, and other outbreaks.

Our center is engaged in about a dozen research streams right now on this topic. And joining a group of colleagues around the world that are working very hard to try to understand both the situation and the possible responses to it.

I’m going to — I’ve been getting on the order of 100 media requests a day, I think, recently, and thought this would be a way to try to answer some of the questions that different people have, because there are overlaps. So we’ll see if this is a good forum. I hope it is.

I’m going to use most of the time to take questions, but I thought maybe two topics of particular interest today. One is the WHO briefing from Dr. Tedros, in which he used the 3.4% case fatality ratio number, which has been disturbing a lot of people. And I’ll address that in a second.

And then the second is sort of where we are in the United States in terms of the transition from isolated cases being known about to more widespread cases being known about. So on the severity aspect, this number has prompted a lot of discussion.

It’s simply dividing the number of confirmed deaths by the number of confirmed cases. We and others have done a lot of methodological work over the years in trying to understand how the estimate of the case fatality ratio changes over the course of an epidemic.

At the early part, there’s almost always a bias. And I think this is reflected in that number, that preferentially, severe cases are detected. And so when you divide cases, deaths by detected cases, you get an unfairly high estimate of the severity and of the case fatality risk.

On the other hand, in a growing epidemic, which it is in some places and is not in others, but in the places where the epidemic is growing, there is a lag which causes one to underestimate the risk of dying. And that lag is the fact that, if you look at the cases today and the deaths today, some of the people who are cases today will, in the future, die. And they’re not being included in the numerator of the fraction.

So those are the countervailing biases. That’s why numbers — those are two of the main reasons why numbers bounce around, and it’s not necessarily a reflection of any change in the reality. It might be, but it’s also very often a change in the way that people are calculating and trying to account for those biases.

And there are various approaches to doing so, which different groups have done. But the raw number is only one possible number. The last thing I want to say about that is that people — here’s a case where the very, very precise terminology matters a lot. And I know it’s hard to use jargon in articles.

But mortality rate is a really vague term that nobody knows the meaning of, and nobody uses precisely in technical discussions. So the number that is of most interest, but right now, quite uncertain, is the infection fatality rate, or the risk of dying if you’re infected with this virus.

The reason we don’t know that is that we essentially have no good handle yet on the proportion of cases that get symptomatic, much less symptomatic enough to be detected. So what we’re, in fact, able to calculate right now is a symptomatic case fatality risk, or rate.

And that 3.4% is perhaps even an overestimate of that, because it’s not even detecting all the symptomatic people, since most of the denominator is in a place with an overwhelmed health care system. The infection fatality rate will become clearer, as Dr. Tedros said, when we have serologic studies that can detect who’s been infected in a way that was not noticed at the time.

And will almost undoubtedly be considerably smaller than those numbers that you’re seeing. But how much smaller is unclear, because we don’t know the proportion of very, very mild or asymptomatic cases. The second broad area, I think, of interest is trying to understand what’s going on in the United States right now.

I’ve compared it in a few places to sort of raising the lights on a scene that’s already in progress. I think what’s confusing is that we’re used to news events being actual things that just happened. But in fact, a lot of what we’re uncovering as we begin to do a little more testing in the United States is that there has been transmission in the past.

There is ongoing transmission. And we’re getting, still, a very dim view of what that is. But it’s — as we get more numbers, a lot of that is just uncovering an existing situation in progress. So there’s been a lot of discussion [about] testing in the United States that’s been quite inadequate to define the extent of transmission.

And there are various reasons for that. I don’t really want to go into those. That’s the past, and we need to look to the future. I think it is going to get much better. Whether we get enough testing capacity is still an open question, but we will certainly be doing better and be getting a better sense of what’s going on. I think I’ll stop there and use the rest of the time for questions.

Q: Thanks so much for setting up this call. I really appreciate it. Two questions for you. One on the global numbers that we’re seeing, and then one more specifically on the U.S. In terms of the global numbers, specifically in mainland China, we’re wondering if you think that the vast reduction in case numbers that we’re seeing there is plausible, or if you think that we’re still missing quite a number of the asymptomatic or mildly symptomatic patients.

So if you trust that drop, and then in the U.S., I wonder if you’ve done any modeling in terms of trying to determine how broadly it might have already spread in the U.S., and what our underlying risk might be. Thanks.

MARC LIPSITCH: Sure. So take the first question about the drops in China and whether they’re real or whether asymptomatic or mildly symptomatic cases are being missed. I think both are true. And what I mean by that is, I think that the decline in case numbers is confirmed by multiple lines of evidence.

And by my — certainly by my discussions with colleagues who have been there, including on the WHO mission. And nevertheless, I think there are a lot of cases that are being missed. So, I don’t — it’s not that I believe the absolute numbers are correct, but I believe that the trend is downward.

And, for example, during Guangdong fever clinic data where they tested over 300,000 fever clinic patients and saw a downward trend, quite consistent downward trend over the period from late January to mid-February, in the proportion of those cases that were COVID-19, that seems, to me, a very good indication that it’s not just inadequate testing and the like.

I was skeptical before I saw those data and talked to people and got a better sense. So, I think we should all be skeptical. But I think my current view of what I know of the Chinese data is that it is — that the trend is real, and there are still many cases being missed.

In terms of the second question about the United States, we have done some modeling. But it is — there are so many uncertainties that I don’t know which version of the modeling to believe. So, the rough idea is that we’ve had cases in the United States that were probably missed as they came into the United States sometime around about a month to a month and 1/2 ago.

And our calculations that are in a pre-print and that are almost the exact same result as Imperial College’s results about missed cases in travelers, suggests that about 2/3 of cases in travelers were missed worldwide when they came into the country, into other countries from China.

So, if we knew about on the order of a dozen imported cases and, I’m going to say lots of approximate things. And all of these are very approximate. If we miss — if we knew about a dozen, then there might have been another 20 or 25 that were missed.

And those missed ones, some of them, and we don’t know exactly what proportion of them will start to spark the chains of transmission. And depending on your assumptions about the size of those chains of transmission and which ones did and didn’t, I think you will find — I mean, you could have certainly hundreds if not more cases around the United States by this point.

The problem is with giving a firm number. And the reason that I’m being vague is that there’s a lot of randomness in whether a particular case does or doesn’t sort of spark a chain of transmission, meaning have secondary and tertiary cases and so on.

For example, in 2003 with SARS, there were approximately two dozen imported cases of SARS in the United States, and no known secondary cases. And probably we would have known, because they would have started chains of transmission. That was partly due to good containment, and it was also probably due to luck.

So it may be, and this is speculation, but based on what our models suggest, it may be that, for example, Seattle got unlucky and had an early introduction that did take off into a chain of transmission. And other places that did nothing different might have had better luck.

And so I think it’s quite possible that we’ll see some places with lots of cases, once we start testing at other places that you would have expected to have similar risks maybe having fewer cases. But it’s really early. And the only way to move this from speculation to fact is the fact.

Q: I don’t know if other people are going to jump on, so I’ll just my follow up question, which is, when do you think that we are going to have a good idea of what’s happening in the U.S. Like, the only way we know is to test, but eventually, time will catch up with us.

So if you’re somebody who’s worrying about what this is going to mean in the future, when do you think we’ll really know in the U.S.?

MARC LIPSITCH: I don’t have a good answer for you on that. I think it really depends on how fast the testing is rolled out, and on how many cases there actually are, and on whether we start to see situations like Italy or Iran in parts of the US. And I have no way of — I could not have forecast that Italy or Iran would have the situations they did.

And so as much as I want to say what I think, I believe in humility. There are certain things that you couldn’t have forecast in the past, I don’t think it maybe is responsible to try to do that in the present. So we don’t know anything more than we knew then.

Q: Hi, thanks for taking my question. You have been quoted widely with an estimate that 40% to 70% of Americans, of humanity, rather, will be infected with coronavirus. And then I gather you walked it back a little bit.

But I’m just wondering, what is your current prediction, what is it based on, and what do you think needs to happen in response? And also the follow up on testing, I’m curious what you think, how testing should proceed. Who should be tested? The Vice President last night said, any American who wants a test should — can get it if their doctor orders it.

And I’m wondering if that’s feasible, if that’s advisable. I mean, how should testing occur? 

MARC LIPSITCH: So the first question is about the ultimate size. And my original statement was that, based on the data and estimates that were around two weeks ago, I said, I thought 40% to 70% of adults, because we don’t know what’s going on in children.

40% to 70% of adults might get infected. Since then, the estimates of transmissibility have been lower than the previous estimates. And so the latest numbers that I gave are 20% to 60%. I’m trying to here to balance careful analysis with trying to answer a question that nobody has a good answer to.

So the preface to this is, projections are only right once they’re — once you have the data. And projections are projections until they come true. But if we don’t have any ideas about what size of epidemic we might be dealing with, then we can’t make rational plans.

So we have to have uncertain projections. And indeed, 20% to 60% is a pretty wide range. The basis of that is really a combination of two complementary types of thinking. One is the historical experience with pandemic influenza, where we have had four pandemics in the last 100 years or so, 102 years.

The proportion of people who got ill in those epidemics was ranged from under 20% in 2009, on the low end, or probably about 20% in the United States, for example. To almost 40% in 1968. So that’s ill, that’s not infected.

In flu, as in this coronavirus, there are some number of subclinical or even very mild or asymptomatic infections. So we don’t know how many people in most of those pre-2009 pandemics got infected, but I think it’s safe to say that over 40% got infected in probably most of those, if not all of those. Except for 2009, where it might have been a little bit lower.

So that’s one set of data. It’s experience from flu. And then that’s where the second type of reasoning comes in, which is that we do mathematical modeling in our group and many other groups around the world. And what the mathematical models of any infectious disease say is that the ultimate number who get infected depends, very importantly, on something called the basic reproductive number, sometimes called “r naught” or r 0, which is when a disease first starts out in a population, the number of secondary cases that each primary case infects. The estimates, as I say, of that number had been bouncing around, and mostly trending somewhat downward. I think not because of changes, but because of changing analytic approaches.

And so when I said 40% to 70%, that was a period when it looked like most of the estimates of contagiousness or reproductive number of COVID-19, of SARS CoV-2 were higher than those from pandemic flu. And now it seems that the estimates are more in the same range or a little bit lower than pandemic flu. So that’s why I thought it was necessary to revise downwards.

And then the last thing to say about that is that the number infected is something that you can predict with more certainty than the number who get sick, or the number who die. Because each virus has its own particular profile of symptomatic, and very symptomatic, and critical, and fatal.

But what limits the spread of viruses, and therefore, what you can sort of predict quantitatively better, is the number who get infected. So that’s why I say that the infection fatality rate, in an answer to a previous question, the infection fatality rate is the number that we would really like to know.

Because if it’s 20% to 60% of adults getting infected, we would like to know what number to multiply by that to estimate impact in terms of deaths. And we don’t have that number yet, because we don’t know the proportion symptomatic, which is a crucial ingredient.

So unfortunately, we have two numbers that are perhaps useful. But what we would like to do is have a third, and those are the other numbers that might get infected and the numbers of symptomatic people that end up dying. But we need a third number in between them, which is the proportion symptomatic. And that’s the yet uncertain number. So we can’t do that calculation yet reliably.

Now I’ve already forgotten your second question.

Q: OK, so actually, it’s part of the first question. What is — so knowing this, that it might be, that it looks like it might be 20% to 60%, what does that mean in terms of the response that should happen? Are we at the point where we should be talking about canceling mass gatherings? When do you get to the point where you decide you need to close schools? I mean, what should be happening?

MARC LIPSITCH: I think that’s a very complicated question. I think that, from the epidemic control perspective, it’s pretty clear, based on historical studies that we did and others did of the 1918 flu pandemic, which was a long time ago. But diseases are diseases, and cities are cities.

The cities that implemented those kinds of control measures, canceling mass gatherings, closing the churches and saloons and other places of public gathering, and other types of interventions early, had smaller epidemics and flatter epidemics, or less peak, lower peak, than those that waited longer.

There are papers in 2005 that came out from our group in the proceedings of the National Academy of Sciences and from a historian at the University of Michigan named Markel, and the Journal of the American Medical Association. We both looked at the same question and came to the same result.

Which was that, the other result that we came to was, and this is why it’s such a complicated question, is that if you let off those interventions, then the virus comes back. And the reason why that makes it more complicated, well, the reason why that happens is just that the virus is still around. You don’t remove all the cases. And even if it’s not around in your location, it can be reintroduced by a traveler.

So when you let up on those interventions, then transmission resumes, because the virus doesn’t know that people who were social distancing last week. It just knows that it can get from one person to another. And so the dilemma, which I don’t have a good answer for, is, it’s clearly best to start those interventions early and to leave them in place from a disease control perspective.

From a human perspective, disease control is very important. Human contact is very important for social well-being and mental health. It’s important for education. It’s important for commerce. It’s important for keeping the economy going and for keeping people fed, and other basic services.

And so what’s right from the disease control perspective has to be balanced against what’s right from keeping society functioning perspective. And I have expertise in one of those, but not in the other. So I don’t have a really strong answer for you, except to say that 1/2 of that part, we sort of know what to do.

But the challenge is that it could take quite a long time of doing it. And if we start when there are no cases or very few cases, my concern is that people will not understand the gravity of why it’s necessary, and will have a hard time maintaining it.

So I don’t have a very good answer for you about what should be done. I think, as cases begin to appear in individual communities, and we get a better sense, it’s going to become more important to just start taking those measures. But exactly how to do that is a political and policy decision.

The last thing I’ll say is that, if this were flu, then everyone would agree that we should close schools as part of the response. And some of you will remember in 2009, that that happened, at least in Boston, and certainly in other places as well. Closing schools was one of the main counter measures for, in that case, a much milder pandemic of a different virus.

This virus clearly does not make children as sick as it makes adults. And we don’t know whether that means that it’s sparing the infection of children or if it’s just making them very mildly ill. And we don’t know whether children are important in transmission. So for flu, they’re important in transmission. Closing schools is an essential control measure.

For this virus, we don’t know that that’s the case. And I think probably there will be some places that don’t close schools, and we’ll be able to observe in those schools whether, if there’s enough testing, whether there are significant transmissions in the school. And then we will get educated about this question.

But at the moment, a lot of places are cracking down in every possible way, like Italy, on public gatherings, schools, and other ways of trying to stop transmission. And so the role of school closure is hard to disentangle from the others. Same in China.

Q: Hi. Again, my thanks for doing this. I wanted to clarify when I thought you said, Dr. Lipsitch, that modeling suggests that — did I hear you say about 2/3 of cases in travelers you think were missed worldwide when they came into this country, the United States, and other countries? 2/3 of travelers?

MARC LIPSITCH: Into — so, yes. So what we did, and it’s on MedRxiv, Pablo Salazar is the first author, I think, or Rene Niehus. I can’t remember who was the first author. The analysis we did was to compare everybody to Singapore, because Singapore has historically, and apparently in this epidemic also, among the most effective methods for detecting cases in travelers and dealing with them.

So there’s no perfect gold standard of what detection means. But compared to them, compared to Singapore, we estimate that the average, that on average, a case had a 1/3 chance of being detected in some other country compared to its chance in Singapore.

Q: Including the U.S.?

MARC LIPSITCH: [INAUDIBLE] Including the U.S., yes.

Q: OK, thank you.

MARC LIPSITCH: We did stratify a little bit based on some characteristics of country. It didn’t matter very much. Most countries were sort of around that same — the U.S. is — the group of countries, including the U.S., was a little bit better, but not much.

Q: And again, you’re saying that was because of lack of testing in the early days and relying on symptom screening?

MARC LIPSITCH: I think it’s a number of things. It’s lack of testing. It was a lack of symptom screening. It’s also just the impossibility of symptom screening to get all cases. Because if someone isn’t sick right at the moment that they’re screened, you don’t find them. So it’s a number of different factors.

Q: Thanks very much.

Q: Hi. Thank you for taking my question and for doing this. So it seems like, I guess like, what data is available suggests that this is not as infectious as other viruses. In part because pre-symptomatic people don’t seem to be driving a lot of spread, and then it seems that transmission is largely among close household contact, especially just like, kind of out in the community.

And I just wanted to see if that’s kind of how you interpreted it. And if that is the case, how that compares to other viruses, and then what that means about epidemiology and response strategies. Thanks.

MARC LIPSITCH: Yeah. I have reached out to the World Health Organization to try to understand the basis of some of those statements. My perception is that there is significant community transmission, especially when you aren’t aware that someone’s sick, because there’s not enough testing.

And that pre-symptomatic transmission has unquestionably been documented, so it does exist. The only question is, how often it happens. And my interpretation of the evidence that’s out there is that it probably happens quite a bit.

And that’s an inference from data on the timing of symptoms and the timing of transmission, both of which are uncertain. And so I’m not fully sure that that’s the case. But my best assessment from the evidence is that there is substantial pre-symptomatic transmission in the 10s of percents.

And I have, as I said, reached out to try to understand the basis of that statement because I found it surprising. And I think a number of other modelers who do the same sort of work that I do have made the same inference about pre-symptomatic transmission.

Q: OK, thanks. That’s helpful.

Q: Marc, I wanted to ask you to clarify something you said earlier. When you said the number of people infected is something you can predict with more certainty than the number who will get sick or die, are you referring to the R naught of the virus? And if so, can you tell us currently what you believe that R naught to be for COVID-19?

MARC LIPSITCH: No. What I was referring to was that number, the 20% to 60%, as I updated to, or the sort of proportion of the population that will ultimately get infected, which is itself determined importantly by the R naught, but is not the R naught. The R naught is a number greater than one, and this is a proportion that’s less than one. So it’s two different things.

I don’t think anybody has a certain view of what the R naught is. I think the work that we’ve done led by colleagues in Hong Kong, Gabriel Leung’s group, suggests it was around 1.9 in the early days of Wuhan. And some of the other estimates that have been coming out are also on the sort of just under two range, with some uncertainty.

But I would caution — I mean, the reason that the picture is changing is again, not that the data are necessarily different, but that different people have tried different ways to compensate for the limitations of the data. And so I think we have to be a little humble about what the truth is, and say that our model suggests it’s around 1.9 with maybe plus or minus 0.2.

But our model is one model, and other models suggest other numbers. So I think if I had to guess, I would put money somewhere between 1 1/2 and 2. Well, maybe 1 1/2 and 2 1/4. But again, no data set that I’m aware of is totally reliable. And so we’re all making different approaches of attempts at trying to understand what’s going on from limited data.

Q: OK. Thank you.

Q: Hi, Professor Lipsitch, thank you for taking this question. I’m wondering about Singapore as a model for epidemiological data. When do you think you’ll have good enough information from a well-documented place like that in order to calculate both R naught serial interval and then the number of infected asymptomatic people?

MARC LIPSITCH: So that is also a very complicated question. The places that are doing the best job of control and documentation are also those in which R naught is not manifesting itself, because they have the epidemic under control. So I don’t think Singapore is going to be a source of R naught data anytime soon, because it just isn’t possible.

It’s not happening there. It’s not an uncontrolled epidemic. Serial interval, this gives me a flashback to SARS, because that was — Singapore was where we got the data to estimate the serial interval. And one of the striking things there was that the serial interval in the initially uncontrolled epidemic of SARS in 2003 was declining week by week, because the more rapidly they found contacts and isolated or quarantined the contacts and isolated the cases, the more the end of the transmission time was cut off.

And so the serial interval was itself a moving target, and that was one of the findings of our paper, actually, in SARS. So I think the serial interval is going to be harder and harder to nail down in the places that can measure it best, because it’s going to be a moving target.

I think the pre-symptomatic transmission question, which is, in a way, more important than the serial interval, well, they’re all important. But the pre-symptomatic transmission question, I think if there are places that have enough cases to really do multiple contact investigations, but not so many cases that they are overwhelmed, those are going to be the places that can estimate the pre-symptomatic transmission and the serial interval, because there will be people who had contact only with a pre-symptomatic person, not a symptomatic person.

And which countries, country or countries or regions that sort of remains to be seen. It’s a funny combination of having good enough resources to gather data and enough cases to do analysis on, but still not have it totally under control. So it’s an unusual balance.

The proportion symptomatic, I think, data sets like the Diamond Princess data set and the data that have been reported from by journalists, at least from Korea, where there are large numbers of people who are exposed and they can be tested heavily, because the testing capacity is enough. I think those are going to be the most informative data sets for defining this sort of spectrum of severity.

Q: Great. Thank you very much.

Q: Hi Professor. Thank you very much for doing this. You had said the main thing that we don’t know right now is the infection fatality rate. Can you sort of give us a sense of how far we are from knowing that, and what other information we’re going to need to get a better handle on it?

And then my other question is that you mentioned something about that 2/3 of the travelers we missed, that there was a large number of travelers who were missed in the modeling that you have. Does that suggest that we should be doing something more to restrict travel until we have a better handle on how this is spreading and how it can be contained?

MARC LIPSITCH: OK. For the infection fatality rate, what will ultimately answer the question is when we have serologic studies, studies of blood samples from exposed people, where we can find out how many people have been exposed and developed an antibody response, which indicates infection.

That will be the gold standard, then what proportion of them were symptomatic or died. So there will be some inference needed, because they’re not going to give blood if they’ve already died. But combining data and serology with data on other sorts can give a sort of final estimate of that.

In the meanwhile, I think these studies of large, exposed populations and just whether they had symptoms or not, among those who were detected to be positive for virus will give a bound on the number of exposed, will give a lower bound on the number. Sorry. It will give a lower bound on the number infected, because some people may be infected and not even detected by those assays.

Serology is the final answer, and that’s the intermediate step. In terms of travel restriction, I think pretty soon, the travel restrictions are going to be far less relevant. Because, to the extent that the virus is many places and is transmitted locally, that’s an exponentially growing process.

And adding a case here and a case there through travel is not necessarily the most high priority. So there’s a really nice post on the “Virology Down Under” blog from Australia, guest written by Jody Lanard and Peter Sandman, who are risk communications experts.

And they have a phrase that we’re moving now from the phase of keeping them from infecting us to the phase of keeping us from infecting each other. That’s a really nice summary of what we’re in the middle of right now. So shutting the borders if we have massive transmission ongoing, or even significant transmission ongoing inside, is a somewhat ineffectual approach.

I think in the very short term, while we still don’t know if there’s ongoing transmission in, say, some parts of the United States, it may make sense to keep restrictions in place for places like South Korea that we know have a lot of transmission. But for all sorts of reasons, restricting travel is not a great situation.

And so I hope that there will be a rational letup of those kinds of restrictions as it becomes that this is a global problem.

Q: Thank you.

Q: Hi, thanks for taking my question. I’m wondering if we know anything more about the role or contribution of super spreaders in terms of how this virus works as it continues to spread. Is it more evenly split between steady spread, between a lot of infected people, or are these super spreaders more important? Will the contact tracing data from China help us understand that better? So that’s my first question.

MARC LIPSITCH: I think — I have not seen good data on the role of super spreaders. Which is not say it’s important or unimportant. I just, I don’t think I have a good sense of that yet. Super spreaders are very important at the beginning of an epidemic in a place, because they can start multiple chains of transmission with a single introduction.

Where, as the epidemic progresses, from a sort of population level perspective, they matter less because they average out with everybody else. And so what matters is the average rather than the variation.

Q: I was also hoping you could clarify with your estimate of R naught currently being around maybe 1.9 or between 1.5 and 2. But then also this idea that transmissibility is fairly high. So could you clarify what you mean by those two things?

MARC LIPSITCH: Yeah. So the transmissibility being fairly high is to say that, relative to say, seasonal flu, which is kind of the 1.5 range. It’s relatively high compared to pandemic flu, which is in the, I think 1918 is the only one for which we have a really good estimate, which was around 2 on average in the United States in cities.

It’s sort of in the same range. So it’s high enough. It’s not high compared to measles or compared to a lot of other things. But it is high compared to — and it’s not even high compared to SARS, where it seemed to be about 3. But it’s high enough, and it’s a difficult to control type of transmission in the sense that whether pre-symptomatic transmission is 1% or 40%, which I think is probably the reasonable range of estimates based on very, very uncertain data.

Whatever it is, there certainly seems to be a lot of transmission right around the time people get sick. And that’s bad enough, even if it’s not exactly pre-symptomatic.

Q: I see. Thank you for clarifying that.

Q: Hi Marc. Just wanted to ask a question related to what we’ve been seeing out in Washington state and, I gather, in China as well. But what do we know about the vulnerability of older adults to get symptomatic disease?

MARC LIPSITCH: It’s clear that, from the Wuhan data, at least, and from the sort of height of the epidemic, that older adults are at greater risk of getting symptomatic disease compared to the proportion of the population, and are at greater risk of dying if they get symptomatic disease. So it’s a kind of a double hit.

And it goes up sharply with age. So the 70 to 79 group is worse than the younger groups. So that’s what we know based on those data. And I think — I haven’t kept up as much as I’d like with all the other analyses, but they basically seem to be similar, as far as I can tell.

Q: Has that been quantified at all? Are they twice as susceptible as someone, you know, in their 20s or anything like that?

MARC LIPSITCH: We have a paper coming out that I don’t know if I can yet talk about, because I’m not exactly sure at what stage it is. It’s the one with Gabriel Leung’s group, and they’re the leader. So I don’t want to speak about it yet. But I think I would refer you to the Chinese CDC report, which I think has some estimates of the relative risks of different groups.

Q: All right. Thank you.

Q: Hi. Thanks for letting me back on. I really appreciate it. I wanted to address the testing question. You talked about the importance of more testing. But, I mean, who should be tested? Everyone who has a sniffle and fears they have coronavirus virus?

Or how do you — and also, what is the capacity for testing right now? How fast can it be rolled out, and how should it be rolled out?

MARC LIPSITCH: Yeah. I’m glad we got back to that. I’m sorry to hear the Vice President saying anybody can be tested if their doctor orders it. That might be — yeah, that doesn’t seem like a helpful way to phrase it. There are really two purposes for testing that are distinct.

One is to improve clinical care so that people who are significantly ill can find out whether this is what they have, and that is important for knowing how to treat them, and also for knowing how to protect health care workers. So that’s one purpose, and that would argue for particularly testing the most severely ill.

And then the second purpose, which is distinct, is to do surveillance in epidemiologic studies to help us understand how many cases there are, what the spectrum of severity is, and those kinds of questions. And those are — and in those cases, there are different kinds of studies you can set up, and surveillance systems that answer different questions.

So the fever clinics in Guangdong are a good example. As I mentioned, they tested 300 something thousand fever clinic attendees over several weeks, and were able to get a sense of the trajectory of those disease in Guangdong because they could see the proportion of those people who were positive for this virus.

And that also gives you some sense of the number of people who have it, because you’re testing a sample that’s enriched for sick people, but is not just the most severely ill. Another type of testing is in outbreaks and also in households. And there, the goal is to try to characterize who gets it, what the age-related risks are, what the risks are from contact with different sorts.

And that’s been done, for example, in Shenzhen in that same province of China, where they’ve done, Justin Lessler’s group at Johns Hopkins, along with a number of collaborators from Shenzhen CDC, have a preprint out that — well, I don’t know if it’s out. It’s been submitted. That tries to define some of those questions.

And so it’s really two parallel purposes. One is to try to help patients, and the other is to try to help reduce the number of patients that ever get it by informing public health. And there are different types of surveillance that are appropriate, or different types of testing that are appropriate.

Q: So outside of a study, or any kind of epidemiological surveillance effort, for patients, you would recommend just testing those who are severely ill and in the hospital?

MARC LIPSITCH: I think until the testing capacity is considerably bigger than it is now, there’s not a choice.

Q: OK.

MARC LIPSITCH: Because it’s still flu season. It’s still cold season. There are going to be a lot of respiratory infections, and we can’t test everyone. And it’s not going to help anyone if people go to their doctors and say, I want a COVID-19 test because I have the sniffles. That’s going to just bring the health system crashing down if that happens on a large scale, winter.

Q: OK. OK. Great. Thank you.

Q: Hi, Marc. I wanted to come back to the 2/3 of missed cases one more time. But with the symptomatic cases only, that estimate does not include asymptomatic cases. Is that correct?

MARC LIPSITCH: Correct. Yeah. So really, the definition, and I realize this is kind of — sounds like a technicality, but there’s no — since there’s no way to know the number of cases, the best we can do is a Singapore detectable case.

Q: Got it. OK.

MARC LIPSITCH: Which is a funny kind of ad hoc approach, but it gives us a bound on the total cases. So no, it doesn’t include the asymptomatics.

Q: Thank you.

Q: Hi. Yeah, I cover K-12 education, and we’re hearing a lot about like, deep cleaning in schools and disinfecting surfaces. And I’m just wondering what’s known about how long this virus can survive on hard surfaces, and whether that’s even a way that it’s being passed.

MARC LIPSITCH: I think that that is important. I am not expert in that, and I need to get up to speed on it. But it’s not something that I have had time to get up to speed on in the last three weeks. So I am going to take a pass on that, but I think the answer is that on hard surfaces, like door knobs and those kinds of things, there is significant survival.

What I would say, and we’re actually trying to get the word out more broadly, but please help us. I’ve been in two school districts with either my wife as a teacher or my children as students, which I won’t name. You can maybe guess one of them.

In none of the schools that they worked in/attended was there routinely soap in the bathrooms. And then none of the — and in several of them, there was no hot water. So if there’s one thing that schools could do, I would not necessarily — I mean, deep cleaning may be a good idea. But a more fundamental thing is to make sure that there is soap in every soap dispenser, hot water in every sink.

And then make sure the kids know that. Because if they’ve gone to that school for a long time, they would not even think to wash their hands because they can’t. I know that sounds kind of out there, but it is. It is a fact in a number of schools that I know about.

Q: Thank you.

MARC LIPSITCH: Thank you for a lot of good questions. I hope that they’re useful, not only for the questioner but for others.

This concludes the Wednesday, March 4 press conference. 

Visit the Harvard Chan School website for the latest newspress releases, and multimedia offerings.


Harvard T.H. Chan School of Public Health brings together dedicated experts from many disciplines to educate new generations of global health leaders and produce powerful ideas that improve the lives and health of people everywhere. As a community of leading scientists, educators, and students, we work together to take innovative ideas from the laboratory to people’s lives—not only making scientific breakthroughs, but also working to change individual behaviors, public policies, and health care practices. Each year, more than 400 faculty members at Harvard Chan School teach 1,000-plus full-time students from around the world and train thousands more through online and executive education courses. Founded in 1913 as the Harvard-MIT School of Health Officers, the School is recognized as America’s oldest professional training program in public health.