As the COVID-19 pandemic continues to spread, informaticians and epidemiologists are leaping into action, doing what they can to attempt to shape the limited amount of data available into meaningful insights. One project that is doing this in a particularly compelling and meaningful way is Covid Act Now, a predictive statistical model that maps the likely spread of COVID-19 state-by-state across the U.S. By taking into account healthcare capacity data from each state, like hospital beds or ventilators per capita, Covid Act Now projects a timeline for when each state’s healthcare capacity is likely to be overwhelmed, and also offers a window of time indicating a possible “point of no return” for policy action.

Immunologist, physician and author Dr. Leo Nissola acts as a medical advisor for the Covid Act Now project, and he recently joined me on the podcast to share some more context for this statistical model, as well as some insights on the COVID-19 pandemic and what actions he believes are needed to stem the tide of the pandemic in the U.S.

If you haven’t subscribed to Inform Me, Informatics, you can do so on iTunesSoundcloud or most podcatchers. We’re also now on Google Play, Player.fm and Spotify. Like the podcast? Please consider rating us on iTunes! This will help other listeners find out about the show.

Dr. Leo Nissola of the Covid Act Now project

PIPER

You’re listening to another episode of Inform Me, Informatics. Once more, a new episode is hitting your podcast feed a little sooner than usual. But as the COVID-19 situation rapidly unfolds, we’re finding ourselves in a constantly changing public health landscape. And so I want to keep bringing you stories from the field as things are changing. Even in the couple of weeks or so since I’ve talked to Patrick O’Carroll for the last episode, everything has changed. In that time, we’ve moved from talking about public health education campaigns around hand washing, to sheltering in place in our homes, some of us under jurisdictional mandates and curfews.

Given that you’re the kind of person who seeks out a podcast on public health informatics and betting you can sympathize when I say that personally, I’ve been pretty obsessive about tracking numbers and statistics related to the pandemic. One resource I found particularly interesting is an innovative predictive model called COVID Act Now. It’s been widely shared, so many of you have probably already seen the COVID Act Now dashboard. It’s an interactive map of the U.S. that shows state by state predictive disease curves, based on what policy decisions each state is choosing to adopt, taking into account healthcare capacity data from each state like hospital beds per capita. It also offers a predictive window of a few days when each state can expect its healthcare capacity to be overwhelmed. It also offers a point of no return day for when action must be taken. You can see it for yourself at covidactnow.org.

So despite what it might sound like, the purpose of this site is not to make us all hyperventilate into paper bags together, but to provide a clear visualization of what may happen in the absence of highly targeted and decisive interventions. The large team of data scientists, engineers, designers, epidemiologists, public health officials, and political leaders who created and maintain this project did so to give policymakers the best information that data science can provide as they tackle some difficult decisions ahead. When I came across the COVID Act Now dashboard, I had to learn more about this project and its use of predictive statistical modeling to forecast a public health emergency.

I got in touch with Dr. Leo Nissola, an immunologist, medical doctor, and author. By day, Leo runs immunotherapy clinical trials for cancer patients in San Francisco. But he’s also a medical adviser on the COVID Act Now project. Leo wasn’t involved from the beginning, he actually found COVID Act Now the same way that I did. And after validating the model for himself, offered his services as a subject matter expert on the medical side of things, and became involved in the project that way. Much to my excitement, Leo agreed to come on the show. And I started the conversation by asking him to share a little bit about the model and what it can tell us about the spread of COVID-19 here in the United States.

LEO

I think this is a model that was first built on and based on information that was available about COVID-19 online. And the data that data scientists, engineers, public health officials, and epidemiologists came across was very striking. If you understand a little bit about epidemiology, you would understand that a critical premise of it is that a disease and other health events don’t occur randomly. So in a population, if it is likely to occur in some members, it might be likely to occur in other members too, if they have the same risk factors. And those factors are taken into account when you’re looking at information. So when you see data coming in from Wuhan, and then suddenly you see data coming in from other parts of the world, putting those pieces of information together can allow you to not only predict what can happen if you don’t take action, but also understand if you’re going to have the right resources in place for them.

So what these folks did is that they looked at the epidemiological background and the basis of epidemiology. And one important use of it is to identify the factors that some places are a greater risk than others. So one of the things that we could learn from this is that South Korea really handled it well. They were a little bit different from other places of the world like the United States, like Spain, like Italy. And if you don’t put those data points into a graph or into an epidemiological model, you can’t understand very well what is happening within the communities.

PIPER

So I suppose that the founders of this model, they were really looking around and seeing the situation heating up and just wanted to shed some light on what could happen in the future and offer that projection for people.

LEO

Yeah. Well, the pandemic spares no one. So I think it was an ongoing question to people who are not in the hot zones like San Francisco, Seattle, or New York being, “When will my region be impacted? If a lockdown is happening right now in Lombardi, when is this gonna come to the U.S.?” A doctor that is now working on a small hospital in Nashville would be wondering, “Will my hospital be able to take this much pressure, if this thing actually blows up over here?” And health officials were asking themselves if their counties would be able to offer patients the adequate care in case this disease is spread there?

PIPER

Leo wanted to emphasize that the COVID Act Now model, while useful, is a guiding tool based on the best information that we have. Not a crystal ball, and it does have some limitations.

LEO

From everything that I’ve seen and that I have studied, I think this model is pretty reasonable. Although there are no shortage of unknowns at this point, especially during times where resources are limited, this is a valuable resource, it does have some limitations. And because we are living in the present time where we are not analyzing the past, this model is based on information that is available right now. And as you know, information changes in a pandemic almost every hour. We know that this variables will change as new data comes in. So in the same way that Boris Johnson has imposed an unprecedented three-week nationwide, lockdown yesterday, the Brazilian president is saying that he’s willing to open up the economy in Brazil, and yet Germany’s saying that they have banned gatherings for more than two people. And we have some call out to open up the markets in the U.S. So you can see that there is conflicting information and conflicting attempts to tackle this problem based on different epidemiological models.

There is one way that one could potentially overcome all of this, and actually understand what the actual situation is without predicting the outcomes based on the data that we are seeing from other countries. But that would require massive screen testing for antibody in a large scale population. So I think one of the important things to point out is that hospital beds, ICU beds, ventilators, all of those assets, all of those resources are not spread across a country in a regular fashion. So you have some places like New York, like LA, like San Francisco where you have large hospitals or like Houston that you have a large cancer center, but there are places where the hospital beds are not available. We know right now that there are counties in the U.S. that have 60% of their hospital beds already utilized. And the crisis has not yet reached there.

We also know that there are places in America that there are no intensive care beds at all. And over seven million of people that are touched by this health care system are at the age of 60 and up, which accounts to the population at risk. So this model is based on different things. One of them is how quickly this virus gets spread. The other one is what kind of resources do we need to serve the community that may be affected by this virus? And the other is what kind of people actually will struggle more with this? And those stratifications, every time you stratify so much, you can incur an error either for overly aggressive strategy, or perhaps you’re missing a part of the population.

One thing that I want to point out to you though, is that we talk a lot about vulnerability. And we talk a lot about who are the Americans who would be vulnerable to this. And if you get what the WHO says, they said that people who are mostly vulnerable are people with diabetes or prediabetes, who have heart disease, or who have asthma. And just to give you a couple of numbers, the American Heart Association says that 121 million Americans have heart disease. Over 100 million U.S. adults have diabetes or prediabetes. And over 25 million Americans have asthma. So you see that this is not a small group of people that we actually thought about.

What I do see and hear from my colleagues who are on the front lines is that people are being diagnosed in San Francisco, for example, as a high probability of COVID-19 and are being sent home because they don’t need to be hospitalized. So what I’m afraid of on those numbers, is that the relatively low number of confirmed cases are just because they have to jump 100 hoops to get tested. For example, I was on a call with colleagues from Lombardi and they had to choose who they would give a mask ventilator to between two patients. And that’s pretty sad for a doctor. We study to save lives. We study to be there and to treat people and not to choose which one we’re going to treat first.

PIPER

So, speaking more generally about sort of data and epidemiology. Arguably the last time we faced a global pandemic on quite the scale was maybe the Spanish Flu in the early 20th century, when epidemiology was a much younger field. We had nothing approaching the computing and the statistical analysis capabilities that we have now. So what do you think the role of data in general is in an evolving global pandemic like this when we have so many more capabilities than we used to?

LEO

Well, the data that I’ve seen so far suggests that the new coronavirus is more infectious than the 1918 flu. And I think, you know, if you are estimating the R-naught value and measuring the infectiousness of this virus, it is far more infectious than the influenza. So I think the epidemiological models can provide us some sort of understanding of how far can this go. But at the same time, we see places that have been exposed to this virus and have completely shifted the pattern, like South Korea for example. And I think understanding what they’ve done in terms of measures and actions are a little bit more important than comparing it with a different strand of viruses.

PIPER

So what would you say is the key difference between the prevention and control efforts that South Korea has taken versus what the U.S. has done so far?

LEO

I think we should take a moment to acknowledge the need to do broad, fast testing. Testing should be easily available. If you can’t get tested to understand if you have been exposed to this disease or not, you don’t know if you can come out of your social confinement. You see, to resolve this, we would need random massive screen testing for a diverse set of people of age, gender, ancestry, geography, without symptoms. We need to understand how far has the virus spread within our communities, so we can make a more educated and perhaps informed decision of our actions as society. You see, the UK just announced about eight hours ago that they are going to do exactly what I’m telling you to their whole population.

PIPER

So when different local jurisdictions across the U.S. are thinking about how they want to respond to the outbreak, what advice would you give them? What guidance would you give them for how they should individually look at how to protect their populations?

LEO

This crisis is a crisis unlike anything we’ve seen so far. This is definitely the first pandemic I am living through. So this requires a new bold and aggressive policy solutions. I think people should be secured with their jobs. I think people should be able to be fed. I think people should be able to feel that the health officials are looking after them. And I would encourage all members of Congress to support an economic relief package that would give Americans recurring monthly payments and a support in order for them to get tested in case they feel that they’ve been exposed to this virus.

You see that there are some unsung heroes in this story. And those are the janitors, the cashier workers from our pharmacy, from our grocery stores who are out there right now in our community exposing themselves so we can continue to live. So I think we should look after these people and provide them whatever they need to continue living while we still try to grasp our way out of this crisis.

PIPER

So regarding the COVID Act Now projections, are there any key insights from this model that you’d like to point to? Key takeaways that people should really keep in mind as they struggle with how to wrap their mind around this pandemic?

LEO

Yeah. I think, again, models should not be static components. Epidemiological models should be smart and change as new data comes in. If you’re not changing your variables, you’re not perhaps capturing everything that has changed. And with this crisis, one thing that we know is that hour by hour we are learning something different. Even as physicians seeing patients, we learn that patients do progress very differently. Just yesterday, I learned from a colleague who was 34 years old, and he was an athlete at school, and he had literally not a single underlying condition. And unfortunately, he’s in a respirator right now. And we can’t understand why he didn’t make it through as easily as a 84-year-old who got cured and treated, and now he’s at home. So an epidemiological model should not be something that would scare people off. But an epidemiological model, in my understanding, should be something that would raise red flags, would point them out to government officials, and would perhaps give a little bit more of an understanding what the future could look like in case we are not prepared.

PIPER

Wow, I’m really sorry to hear about your colleague. That’s really unfortunate. I hope he pulls through.

LEO

Same. Thank you.

PIPER

Yeah. Well, that’s a great answer. Thank you for that insight. I also wanted to ask you, and you’ve touched on this in some of your previous answers, but how would you characterize the urgency of COVID-19 and the need for widespread public response?

LEO

So one of the things that we don’t know is how many people between January and February had pneumonia this year, and how many people in 2019 had pneumonia? And I would like to understand if maybe this virus was around us a little bit before we actually are thinking. Maybe we would see something there. So having answers to that question would make a very big difference in a policy level. Because if we were suddenly to see a surge in hidden pneumonia cases since mid-February or January, they’ll tell us that we are in a bigger trouble. One thing that we were seeing in New York is the rapid spread of new cases. And you see the curve is almost vertical.

So places like New York are very difficult to deal with a lockdown because of the high density. But lockdowns actually work. And we have seen how far the lockdowns have gone in other places of the world. But before we understand how long this virus has been around, we are flying blind. So if I could just say one thing, I think antibody tests can show us what the percentage of the population have been infected. And even if people aren’t currently infected, that would give us a more accurate fatality rate. So that would be my takeaway.

PIPER

Yeah, it all comes back to having the data there, right?

LEO

That’s correct.

PIPER

At this point, I asked Leo the question all the guests on this podcast have to answer sooner or later. How do you personally define public health informatics?

LEO

Public health informatics has used different tools in computer science and technology to help the practice, and research, and learning. In my field, I deal a lot with bioinformatics, in collection of public data, and the storage and analysis of public data. So that would be my definition.

PIPER

Normally, the show ends there. But I wanted to leave you with one last thought and an accompanying caution from Leo.

LEO

I just want to make sure that everyone is aware that this virus does spread very quickly. And we are now at a tipping point where our health care system may crash if we don’t take swift action. So if your health official is telling you to stay home and practice social distancing, just do that. I haven’t seen people in days, in perhaps weeks because I have been practicing social distancing in a very extreme way to make myself healthy to help others. But I would urge you to listen when we ask. Because well, some of us are just sitting at home and being bored. Some other of us are out there in the community unprotected to keep us fed and treated.

PIPER

A big thank you to Dr. Leo Nissola for taking the time to come on the show and share his insights around an evolving global pandemic. Again, you can find the COVID Act Now dashboard at covidactnow.org. The team has also made their model publicly available so that you can better dig around in all of its inner workings and sourced variables. This podcast is a project of the Public Health informatics Institute, which is a program of The Task Force for Global Health. Visit phii.org to learn more about all of our informatics work. You can also find us on Facebook and follow us on Twitter @PHInformatics. The music used in this episode was composed by Kevin MacLeod.

As always, if you have any thoughts or feedback on the show, or you want to send in a tip for an innovative or interesting public health informatics story that you think would be a good fit for the show, let us know on PHII’s social media or email us at podcast@phii.org. Thank you for listening and stay well out there. I’m Piper Hale and you’ve been informed.

BUTTON

LEO

I will definitely follow up and listen to your podcast now, and become a listener.

Share on facebook
Share on twitter
Copyright © 2021 Public Health Information Institute | All rights reserved