Katriona Shea, an Alumni Professor in the biological sciences at Penn State, is co-leader of a national team that advises the Centers for Disease Control and Prevention. Known as the COVID-19 Scenario Modeling Hub Coordination Team, it brings together researchers from institutions across the United States to offer projections on the pandemic. They use about nine models for its projections, rather than relying on one or two. WPSU's Anne Danahy spoke with Shea about how scientific modeling works, how it can be used when dealing with pandemics and the Omicron variant.
Here's their conversation:
Anne Danahy
Welcome to Take Note on WPSU. I'm Anne Danahy. There had been some good news about the COVID 19 pandemic, including the availability of vaccines for children. But cases are climbing again in some places, including Pennsylvania. And a new variant has emerged, Omicron, which has been confirmed in a number of countries, including the United States. Katriona Shea, an Alumni Professor in the biological sciences at Penn State is co-leader of a national team that advises the Centers for Disease Control. Known as the COVID-19 Scenario Modeling Hub Coordination Team, it brings together researchers from institutions across the United States to offer projections on the pandemic. They use about nine models for their projections, rather than relying on one or two. Katriona Shea, thank you so much for talking with us.
Katriona Shea
It's a pleasure to be here.
Anne Danahy
You and I actually recorded an interview before Thanksgiving. But because of the new COVID variant Omicron, all the questions that it raises, we're talking again. Omicron has been confirmed in the United States. So let's start with that, broadly speaking, what do we know and what don't we know about the variant so far, and the impact it could have?
Katriona Shea
Well, it's incredibly early days, but it is a variant of concern for a reason. Which is that in South Africa, which does incredibly good genomic surveys, it has quite clearly risen rapidly and beaten up Delta, very quickly, where it's being surveyed. And South Africa warned the world. And then as people are looking elsewhere, we're finding that it has already spread around quite dramatically. So it's clearly got some advantages over Delta, which will mean that it will spread around. What we don't know is how bad it could be. It is possible that it could be less severe than Delta, for example, but it's also possible that it could be worse. And we do not know how it will respond to our vaccines or existing vaccines at this time. So there are some things that are so uncertain that it makes it very hard to see what will happen in the coming weeks and months.
Anne Danahy
So will it take a few weeks for us to know more about it? How will scientists come to some agreement about whether it's more or less serious?
Katriona Shea
We should know within a couple of weeks about the more or less serious because everyone who has it is being watched like a hawk. You know, so as soon as somebody is known to have it, then we're watching their outcomes and seeing — do they do better or worse. If they're vaccinated or not vaccinated. So some vaccinated people have caught Omicron. So breakthrough infections happen for all of these … for any disease. So we should have quite good information within a couple of weeks about how bad it could be. But it is spreading. So it's definitely going to be something to watch.
Anne Danahy
I don't know if you can answer this. But so for people who are thinking about getting their first vaccine shot, or the booster shot now, is it something they should say, "Oh, well, maybe I should wait and see if it's going to be effective," or no, you should just continue with it?
Katriona Shea
You should absolutely get your boosters or your first doses as soon as possible. Because even if the drug companies develop Omicron-specific vaccine versions, it's going to take a minimum of a couple of months to do that. And then they have to roll them out. So those would not be available. And the rate that Omicron is moving, it's likely to be all over the U.S. long before those vaccine variant versions would be available. So I would get your initial dose or your boosters as soon as you're eligible, which we are in this country, broadly speaking.
Anne Danahy
And there have been other variants too, of course, that have not ended up becoming the dominant one. So could that happen to this one? How do the variants kind of fight it out?
Katriona Shea
Well, in this case, because it's already been seen to take over from Delta in places like South Africa, I expect it will continue to do that. Delta also was very visibly taking over very quickly. Other variants of concern that were listed, like Lambda and Beta and Mu had reasons for people to worry. Maybe they were more severe or something else going on. But they just didn't take off. That something that's already spreading quite noticeably, I doubt is going to stop at this stage. So I would say that Omicron is likely to be here to stay, but I'm not sure. Right. It's very early days.
Anne Danahy
How do you incorporate all of this into your modeling?
Katriona Shea
Well, we actually in a weird way already did, in the sense that some of the work we've already done. We looked at prior variants that emerged so an Alpha came on top of the original strain and then six months or so later Delta came in. So we'd already included that in modeling that we did. In September, we'd included the idea of a more transmissible variant showing up in the middle of November, which makes us look eerily prescient. And we did that in the context of what it would mean for childhood vaccination. So with and without a new variant, with and without childhood vaccination, what could the outcomes be? And that's how we explored different scenarios. And we did that work, as I said in September, and initially, it was intended to inform the decision about vaccinating 5 to 11 year old children, which has already now been made, using in part our contributions. But now, obviously, the teams that we asked to help model that are in part set up to help modeling this new variant as it arises. But every time we get more information, we can tweak our models, put that new information into the models and update them.
Anne Danahy
And when we spoke, there was a pretty positive outlook for the new year. Is that still on the horizon that hopefully things will be simmering down by February into March?
Katriona Shea
Yeah, our projections for no new variants, so with only Delta, we're fairly positive. Some peaks in, you know, October, November, but then a generic trend down nationally, at least in the country through February and March. With pockets, of course of outbreaks where there's low vaccine uptake. Or low background in immunity levels because there's been no disease. So pockets that have not been hit by COVID previously, and where there's low vaccine uptake, are expected to surge anyway.
But now with this new variant, all bets are off I'm afraid. And information will be coming in. But based on the past, I think there's a chance, a fairly good chance, now of a surge. So the more that people get vaccinated, and get boosted, the better. But given that there is some resistance to doing that, please just be careful, continue to mask, just be a bit cognizant of not, you know, walking into big crowded rooms when nobody is wearing a mask, and you don't know the vaccine status of the other people there. And just try to protect yourselves and your families from these risks.
Anne Danahy
Right. I think in another interview that you had done earlier in the year, you were talking about projections that they were positive, but you had also said that a variant could be like a slap to the system. So there is always kind of that footnote, right?
Katriona Shea
Always. Because, you know, every time someone doesn't get vaccinated and gets sick, we give this disease, you know, a ticket to the lottery, to mutate, and to come up with some sort of variant that may throw a spanner in the works, so to speak.
Anne Danahy
And do you have any other advice on how average people, non-scientists, should react to this new development? I mean you want to take the precautions like the ones you were talking about? But you don't want to have unneeded alarm? Are we just kind of going to pause and wait and see?
Katriona Shea
There's a bit of wait and see, there's a bit of please be sensible. You know, you don't want to take unnecessary risks. But you know, until we get a bit more information about how severe the infection is in people that are unvaccinated or vaccinated. We don't know how bad this could get. I mean, it is possible that it could just be less horrible when you get it or that it doesn't get around our existing vaccines, right? If our vaccines are just as good against Omicron as they are against Delta and Alpha, for example. It it will be kind of OK, you know, it's just easier to catch it, it looks like. But if there's some sort of immune escape, which is what they're very worried about, because of the number of mutations in Omicron, compared to the existing ones. There's just more chances that it's different in a way that gets around vaccines, but not completely, There's probably still a lot of protection. So again, I'll say get vaccinated get boosted. It's definitely your best line of defense at this point, as well as sort of careful masking and other sorts of social distancing,
Anne Danahy
The new variant and the possibility that this could continue to mutate and the idea that it has even more mutations, it could potentially spread more easily, you can't help but think like "OK, then what comes after Omicron?"
Katriona Shea
There's absolutely the possibility that we could see additional future variants, as long as people are unvaccinated and catching it. There's the opportunity for this to change every year and we see that with flu.
Anne Danahy
If you're just joining us, this is Take Note on WPSU. I'm Anne Danahy. We're talking with Kat Shea, a professor of biology and alumni professor in the biological sciences at Penn State. She's one of the leaders of the national team that brings together researchers and models from different institutions to offer projections on COVID-19.
You are a lead author of a paper published in Science last year and it outlined the process for merging scientific models that offer scenarios about what to expect from a pandemic, rather than relying on just one model. Can you tell us about that approach and why it's better?
Katriona Shea
Yeah, I'd be happy to. So modeling is an art as well as the science. Because what you're doing is basically a simple representation of reality. And the problem is, when something happens like a pandemic, there's a ton of uncertainty. And so nobody actually knows what's going on. So really expert scientists might write models that summarize what's happening in their opinion. But because we know so little, it's really hard to be sure that they're all going to be on the same page. And even a year and a half into the pandemic, there's still a lot we don't know, about COVID-19. Right. And so, rather than relying on one model, what we try to do is get a bunch of experts together. And that paper was using ideas from expert judgment where people talk to rather than just one expert, they talk to multiple experts and get lots of opinions. And we thought we would do the same thing with models.
Anne Danahy
I think that might be surprising to some people. It was a little bit surprising to me to hear you talk about it as an art form, rather than just purely science. Like, here's the numbers, they don't lie.
Katriona Shea
Well, you know, for example, right at the beginning, we didn't know whether there were asymptomatic carriers of COVID or not. So some models included asymptomatic carriers, and others didn't. We genuinely didn't know. And of course, then the models disagree. Because models with asymptomatic people who are walking around infecting other people without even knowing they're sick, of course, then the outbreaks going to be much bigger in models that include that idea, compared to models, but don't. So that's why we really like to use a lot of different models. I guess the analogy would be is like, if you've got a little problem, you might talk to a couple of your friends and get their opinions, I would reckon. But if you've got a really big problem, like should I quit my job? Or how do I deal with this horrible diagnosis of a disease? You're going to talk to a lot of people and get a lot of advice. And try and catch the sort of weird things that could trip you up in making a decision or moving forward.
Anne Danahy
Interesting. So as you and other scientists know more about the disease and you have more information about when it's transmissible, for example, then you can feed that into the models, and they become more accurate?
Katriona Shea
So all the models are definitely getting better. But they're all quite different. Nearly all of them track, you know who's sick, who's recovered, who's died, who might get sick again. But some of them do it by talking about, say the proportion of humans in the population that are in each of those different groups. Others get really down into the details and track individual people and how they move, and where they go when, and who they interact with. And so they can come up with some quite different answers about what might happen in the future based on some of those differences. But also on things like including asymptomatic or not, carriers in different ways. And there's still a lot we don't know. We don't know about waning. We don't know how boosters will work. There's so much uncertainty. And the reason we use lots of models is because it helps us to deal with that uncertainty, or to warn us about weird outcomes that we might want to guard against if we do know what sort of things could happen.
Anne Danahy
So in general, is this how disease modeling was done in the past? Did it bring different models together? Or is that a new approach?
Katriona Shea
So people have done forecasts in groups before. So for Ebola, and certainly flu has an approach like that in the USA. What we really wanted to do was think about how we aggregate those different groups and how you take the multiple opinions. And partly because we want to get an idea of what we think could happen. But we do something a bit different than a lot of that earlier work, which is we ask about what ifs. So a forecast says what we think will happen just like the weather forecast. But we can't change the weather. So what we do is with a disease, we can do action, take actions that change what happens. And so we say what if we do this? What do we think would happen? What if we do that? What would happen? Is there something we can do that's not too burdensome, but might have a big impact? Obviously, that would be ideal.
Anne Danahy
So is that the difference between a forecast and a projection?
Katriona Shea
Absolutely. Yeah, absolutely. Yeah.
Anne Danahy
So after the paper was published, the National Science Foundation provided funding to help implement what was outlined. And is that what was used to create this COVID-19 scenario modeling team?
Katriona Shea
What we did first was this was an untried, unproven method. We were pretty confident it would be useful, but we'd never really demonstrated it. So we did a case study or a pilot study in collaboration with the CDC to look at reopening strategies. And this was in summer 2020, long before we had vaccination available. So what we wanted to do was ask questions about if we open up like this, or if we open up more, how bad could it get? What could the adverse, you know, results be? And what can we guard against? And we did that case study in collaboration with the CDC. And they saw that it was useful. And then that in part led to us setting up something that regularly, about once a month, run scenarios for the CDC. And that was set up in December 2020.
Anne Danahy
When you run the scenarios, you provide that information, it obviously goes out to the public and the media enjoys reading about it. But does it go to the Centers for Disease Control too and to the federal government to help them?
Katriona Shea
Yes, so the reports are really detailed. One hundred eighty-page reports about every possible thing, every model for every possible state. That goes to the CDC. And it also goes to the White House data team. And we now have started posting some of those reports online. But initially, those were just sent to the federal government. But then we put the aggregate results on a website that you can reach online.
Anne Danahy
Do you think that this approach for bringing all these institutions and researchers and models together will be something that's used in the future? Hopefully, we won't have another situation. Another COVID again. But if there is something like that, do you see this being a useful tool?
Katriona Shea
I absolutely do. Because imagine another Ebola outbreak. Or something like Ebola that was threatening to come into the U.S. Or threatening anyone anywhere on the planet. Clearly, a lot of people would want to get together and work out what could happen under different circumstances. What could we do best to protect people both where the outbreak is happening, and also to prevent it from spreading around the world the way COVID did. So I absolutely think this method would be useful. But generally, for very important, really urgent problems. If you've got a lot of time, if it's not urgent, you can sit and run a lot of models yourself. But it just takes such a long time to do this, that having multiple teams doing it at the same time really saves a lot in terms of being able to get results quickly enough to actually help.
Anne Danahy
For non-scientists, can you describe how this contagious disease modeling is done?
Katriona Shea
What is a model? I never really understood it for quite a long time, even when I was doing it myself. I didn't really get it. It's a simplification of the real world. And if you think about something like a toy dog. That is a model of a real dog. But it's a real simplification. And so for some purposes, it can be really useful. And for others, it might be a disaster. So for example, if you give a toy, a stuffed animal, to a young child, it's very good for teaching that child what a dog is compared to a giraffe or a shark, for example. But it'd be absolutely awful for teaching a young child how much work it would be to look after a real dog to feed it, take it for walks, clean up its messes, etc. So the purpose of the model also drives what you would include. And so the models we're talking about are mathematical representations rather than physical representations. But just like you'd build a model airplane, before you flew a real one to test that it works aerodynamically. We can play around with these mathematical models that are just little simple representations of the real world. And it's just useful to keep track of things. As I said earlier, if you're trying to keep track of individual people, you can't do that in your head. And think it's hard enough to keep track of one person's schedule much less a million people's schedule. And so you have a computer that will do that. And so all of this is written down mathematically and then encoded in computer programs that you can run on computers or supercomputers. And some of the models that we use run very quickly. Others take a really long time to do any projection over something like six months.
Anne Danahy
And you talked about the differences in models and why some of them are done differently, and they might get different information or emphasize it differently. How do you bring that together? How do you combine them?
Katriona Shea
Well, we want all the models to do a better job through time. So all of the models, modeling teams meet. We meet regularly, once a week, and we talk through how people are approaching it. But we don't expect everyone to do the same thing. In fact, scientific disagreement is something we encourage. If someone really strongly thinks no, that's not how this works. We want to know that they've included that in their models, because then we can guard against dangerous outcomes. So one example would be you know, if on average, I'm running a hospital and someone tells me I've got all these models and they say well On average, there's going to be a couple 100 people sick and I have 300 beds. I might go "Great, I'm really happy, I'm never going to get near there." But if the models, some of them say, "Oh, but there's this risk, that actually we might get 500 sick people." And a few of the models say that. You want to know that because then you can say, "Oh, I should be good on average, but there's this chance a bad thing would happen." And so I could put in place a plan to either send patients to other institutions, or if they are going to be overflowing too, I can plan for a field hospital. I can sort of plan or set up preventative or protective measures, because I know there's a chance of a bad outcome. And so if the models disagree, that's actually really useful to know. Whereas if they all just say the same thing, that's very reassuring. But if they say the same thing, because you didn't let them say different things, then you're just not asking for good advice. Just like you'd want to know if you would going to do, make a big decision in your own personal life. You know, a bunch of your friends might say the same thing. But if one of them goes, "Hey, but did you think of this?" And you go, "Oh, my gosh, that would be a disaster!" You know, at least then you can guard against it or plan for it. So you'd rather be informed.
Anne Danahy
You've studied and written about other disease outbreaks as well, including Ebola. In that case, you were part of the team that researched adaptive management, being able to get key information and then change the response quickly. So does that same idea apply here as well?
Katriona Shea
Yeah, the whole point is that a problem will come up and you need to do something about it. But you don't have enough information to make a good decision. So you do the best you can until you can learn something more to help you. And that's fine. And it happens passively like every time we see a patient or a new outbreak or something, we learn something, and we can improve what's going on. Adaptive management, the idea is that you actually go, "What would help me make a decision better, right from the get go? What is the thing that's making it hard to decide?" And you target that information right from the start. So you plan to learn. I guess an analogy might be, say, you've got a job offer in another city with a pay raise. And you go, "Yay, we're going to earn more money!" And then you go, "Well, I don't know very much. And I need to make a decision by tomorrow morning." You know, the absolute first thing you're going to plan to learn about is what's the cost of living in my new city. Because if I'm moving from Arkansas to New York City, clearly the cost of living is maybe going to outweigh the benefit of an increased salary. Whereas there might be other information that would be nice to know. How good is the school district? Is there a nice music scene? But it wouldn't be the first thing you'd target. So what we're trying to do then with adaptive management is say, what is the thing that's really going to help us very, very quickly to do a better job. And to target it and actually plan to learn, right from the get go.
Anne Danahy
What are you and the other scientists who are part of this team learning now that could be applied to future diseases and modeling?
Katriona Shea
Oh, my gosh, so many things. I mean, first of all, how to make something like this work. There are, you know, multiple teams. We have to meet them all. We have to get everybody to contribute. They're all amazing. But nobody is really being paid for this, right? Everyone is just trying to save the world, and leveraging what they already have in place to do it. So sometimes it's a bit like herding cats, but very nice, friendly, helpful cats. But still like herding cats. So we've learned a lot about the process of doing it. But we're also learning about how the the sort of mathematical and the modeling parts can move more smoothly and go better. For example, how do you average models? How do you average a device? There's actually quite a lot of work that, for example, my group has a graduate student who's working on that right now. What's the best way, what's the most informative way to average across lots of different models, but still express that uncertainty. The risk part as well as the average expectation part. And then thinking about how this would work for other diseases. You know, it's a ton of work. So we wouldn't do it for a minor disease outbreak that wasn't going to threaten a lot of people because it involves a huge amount of logistics and effort and time for people. But if there was a big crisis, again, like another Ebola outbreak threatening to spread more widely, for example, then everyone could, you know jump into the mix and contribute. But you'd need to think about when is that worthwhile. When is it just too much effort for the result that you'd get. There's a lot to think about and a lot of that has to do with how urgent the problem is. So there's just tons and tons of questions.
And then how would this work for other diseases and other places? We rely on a ton of different model groups, you know, across the country. Some countries don't have that luxury. You know, so how can we build capacity in countries where there isn't that sort of a resource that they can draw on? So there's a lot of different ways we could to move this in the future.
Anne Danahy
This is more of a personal question. Did the work you do on the pandemic end up taking much more time than you expected? I think you already answered that question to some degree. But what are you going to do for research once this pandemic slows down? It sounds like you've probably already have other research kind of in the backburner waiting?
Katriona Shea
Yeah, it really did take over everything. But, you know, it's been really exciting. And it is exciting to contribute to public health decisions. I mean, we actually were part of the public record for the decision about vaccinating 5 to 11 year olds. And earlier last year, we had, earlier this year, I beg your pardon, we had some of our earlier recommendations to keep masking were presented by the White House, press releases and so forth. It was really exciting to actually have a contribution. But making it sustainable so that, you know, for example, if I got run over by a bus, that someone else could do this for future outbreaks. So that it's not reliant on one person or one group of people, would be really exciting. And so part of what I'm hoping to do is build it up so that there are several people who could do work like this. And there are. I mean, and it's getting better all the time. But also, then I guess I'm excited to … I study outbreaks and other systems as well, not just diseases. So I also study plant invasions and things like that. And you know, how we make those sorts of decisions. It's a lot calmer working on plants. Because things happen more slowly and they don't threaten people's lives. They maybe threaten livelihoods, but not lives. So it's a little calmer. And I'm really excited about making sure that that work keeps going too. It's going on in the background.
Anne Danahy
Kat Shea, thank you so much for talking with us and explaining the science in a way that's understandable.
Katriona Shea
Well, it's been a great pleasure to be here. And thank you very much for having me.
Anne Danahy
We've been talking with Kat Shea, a professor of biology and alumni professor in the biological sciences at Penn State. She's part of a national team that brings together researchers and models from different institutions to offer projections on COVID-19. To listen to this and other episodes of Take Note, go to wpsu.org/takenote. I'm Anne Danahy, WPSU.
