Estimating coronavirus deaths in Utah and elsewhere? Here’s what experts say and why they may be off.

Editor’s note: The Salt Lake Tribune is providing readers free access to critical local stories about the coronavirus during this time of heightened concern. See more coverage here. To support journalism like this, please consider donating or become a subscriber.

When trying to wrap your head around the coronavirus and how it spreads, it’s helpful to reimagine it as a secret so explosive that you can’t keep it to yourself.

Let’s say you know something scandalous, so you tell your three best friends asking them to keep it to themselves. But let’s say your three best friends aren’t particularly trustworthy, and they each tell three other friends the next day. And let’s say that pattern of secret-telling continues for a couple of weeks.

(Christopher Cherrington | The Salt Lake Tribune)

Holy cow, over 7 million people know the secret! This is what’s called “exponential growth,” something you’ve probably heard a lot about recently.

But let’s imagine that your community isn’t so full of gossipers, that at least some in the group grew up adhering to the adage “secrets, secrets, are no fun, secrets, secrets hurt someone.” Now, the secret is only told to two people the first time, and on average, each of them tell two people every day after that. Here’s what that graph looks like.

(Christopher Cherrington | The Salt Lake Tribune)

So after two weeks, 28,672 people know your secret. While that is a lot, it’s a heck of a lot less than 7 million.

Sicknesses, like COVID-19, spread in much the same way: there’s an initial carrier, who gives it to a certain number of people, who themselves give it to a certain number of people, and so on. Time moves forward, and a lot of people catch it. The average number of people who get the disease from one individual carrier is called R0, which stands for the basic reproduction number. It’s the same idea as the scenario we played out above.

R0 varies based on the characteristics of the disease and the population that carries it. For example, the R0 of measles ranges from 12 to 18 — it’s wildly contagious. For unchecked coronavirus, different studies of different populations put R0 between 1.5 and 3.5. Most studies seem to cluster between two and three.

For government officials calculating how many masks, beds, and ventilators we’ll need, there’s a huge difference in preparing for an R0 of 2 (28,000 sick people) versus a 3 (7 million sick).

That’s especially true when there are other variables at play. The whole point of these social-distancing measures is to reduce R0, ideally to something less than one — if most people don’t tell your secret to anyone, your secret is going to fade away eventually.

So each additional measure reduces R0 by a few tenths of a percentage point. Tell people to wash their hands? It helps. Stopping people from eating at restaurants is worth a little bit. Closing schools? Lop some more off. Telling people to stay home is worth a little bit more. If we adopted more extreme measures, like some other countries, and, say, arrest people who didn’t stay home, that might help the average R0 too.

The problem is that nobody really knows the value of each additional restriction. We can look at China, Italy or Iran and try to figure it out, but there are so many differences between each of those places and the ones we live in. To a huge degree, estimating R0 means a huge amount of conjecture.

And to be frank, there are reasons to doubt the data coming from each of those countries, ranging from data reporting that can’t keep up in a crisis to governments that might be incentivized to make the picture seem rosier than it is.

One model that’s been making the rounds recently is one created by the Institute for Health Metrics and Evaluation (IHME) at the University of Washington. It’s a good one. They created a mathematical formula based on solid research of R0, then adapted it for all 50 U.S. states. Then, they compare the expected peak of the disease to the hospital capacities in each location, to find out how well prepared each state is. A separate Salt Lake Tribune review shows that Utah, for example, looks potentially poised to run out of ventilators.

The IHME model estimates 582 deaths in Utah from the disease by August, but their model says the final tally will likely fall between 255 and 1,266. This model is updated daily, so these were the projections as of Tuesday.

That’s a big range in itself. In the United States, the IHME model projects deaths to fall between 36,000 and 150,000.

But once you look at the specifics of the model, there are reasons to question the ranges. They created a math function that is in large part designed to match the data from Wuhan, China, which may or may not reflect what actually happened there.

Furthermore, they made a key assumption. They assumed every state would, within seven days, enact four restrictions on movement: a stay-at-home order, closing schools, closing nonessential businesses, and heavy restrictions on travel. They released the model on March 25; it’s nearly a week later and those measures aren’t particularly close to universal.

As I said: I think the IHME model is well-made. But the truth is that, given so many political and scientific unknowns, their study was forced to make assumptions. And those assumptions fall on the optimistic side of things, even with the error bars they’ve put into their work.

But that’s been a fairly common mistake among experts to this point. Take a study from the University of Massachusetts-Amherst, which on March 16 and 17 asked a group of 18 professionals in disease modeling and policy this question: “How many cumulative positive tests of coronavirus will the U.S. have by March 29?” The average guess was 19,000.

The experts knew that the disease was difficult to model, so the study asked them to create an uncertainty range in their estimates as well. The 80% range from that data projected anywhere from 10,500 to 81,500 cases.

The real number was 140,904 cases.

What’s a little bit remarkable is that the experts were asked the same question again a week after their first attempt, and they still underestimated by a significant margin. On March 23 and 24, knowing that their initial estimates were low, they still projected only 117,000 cases. They were asked to make a projection for just five days later, and were off by over 20,000 cases. So far, they’ve estimated on the low side of things pretty consistently.

I don’t blame the experts one bit, either: I think they’re very good at their jobs. Nevertheless, forget what will happen a month from now — there’s real uncertainty about what will happen this weekend with this disease.

Those experts, by the way, predicted an average of 200,000 deaths in the U.S from the coronavirus in 2020. Their 80% range of estimates fell from 36,000 to 1.1 million, a huge spread in the impact on our country.

When dealing with a virus that we didn’t know existed four months ago, unpredictable governments making decisions in hodgepodge fashion, and the mathematical curse of exponential growth, knowing what to expect moving forward is nearly impossible.

Until we learn more about this disease and our impact on its spread, consider the future a well-kept secret. Planning for an incredibly wide range of scenarios is probably best.