facebook-pixel

Coronavirus case counts are way off in Utah and elsewhere. What’s a more realistic estimate?

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.

Coronavirus testing has been pretty mangled from the beginning.

In the U.S., authorities rejected a working test created elsewhere in favor of having the Centers for Disease Control and Prevention establish its own. The CDC’s first test was immediately and obviously broken, delaying the ability of many Americans to get tested for weeks. When the test finally worked, the health care system did a poor job of distributing them. The nation is still struggling with this today.

Utah has been a relative bright spot. It’s the only state in the nation with a top-10 performance in tests per capita (seventh) and a bottom-10 ranking in deaths per capita (47th). Congratulations!

And yet, even here, significant problems remain. The big one right now is that doctors are not ordering tests for every symptomatic person despite state leaders imploring doctors to order the tests. Some testing centers were turning people away who should not be turned away.

It’s incredibly frustrating. And it’s incredibly clear: The coronavirus case count is way, way off. It is off in Utah, in this nation and in the world. But off by how much?

To answer that question, we need another kind of testing: antibody testing. This is a blood test that figures out whether you’ve had the disease by detecting the presence of the antibodies your body produces to fight it. This helps identify people who were asymptomatic, had mild cases or couldn’t get a traditional test for whatever reason.

Locally, ARUP Laboratories is making a test and wants to ramp up production significantly in the next couple of weeks.

There are some caveats. For one, these antibody tests have a relatively high false positive rate. ARUP’s test, for example, incorrectly tags a noninfected person as already having had the disease about 1.6% of the time. (Standard coronavirus swab testing has a much, much lower false-positive rate.) Right now, the official case count says that just 0.06% of Utah’s population has had COVID-19. It’s possible that the noise of false positives will overwhelm the signal of true positives at first. To prevent this, multiple tests may have to be done on individuals who test positive.

Still, once enough antibody tests are created, and with careful interpretation of the results, we should be able to figure out how much of the population has been infected by testing random samples. With this new data, we’ll be able to make significantly more informed policy decisions. This is pretty exciting.

With that in mind, let’s take a look at some of the prior experiments and studies that tried to get at this central question: “How many people have been infected by coronavirus, anyway?”

These aren’t all antibody studies — in fact, they’re mostly not — but all of these studies try to shed some light on that question. Shoutout to Economist writer Dan Rosenheck for putting many of these studies in one place in a Twitter thread earlier this week. Here are six of them:

1. One study looked at the number of patients across America who were reported to have a fever over 100 degrees and a cough or sore throat. Researchers then subtracted the number of people who had tested positive for flu, and the number of non-flu people who normally get a fever and cough. They figured the remaining cases were COVID-19. They estimate that, on March 15, only 1 out of every 100 to 1,000 cases of COVID-19 were being counted in the country. In Utah, they estimated it was about 1 out of every 300.

2. San Miguel County, Colo., population 8,200, is home to the ski town of Telluride. Residents were given free access to blood testing by a couple of rich locals and 4,400 tests were taken. So far, 1,631 tests have been completed, all done before March 31. Eight came back positive, 25 “borderline,” and the rest negative. Extrapolated out to the whole population, 40 to 165 infections were found. On that date, only six people had tested positive through standard procedures.

3. This study looked at the 21,000 tests that have occurred in the Icelandic population by April 2. Iceland has two testing programs: one that tests sick people in the normal way and one that tests random volunteers who have no symptoms. Researchers then estimated that 83% to 93% of infections were still going undetected.

4. One New York City hospital gave a COVID-19 test to the women who came in to deliver babies between March 22 and April 4. Of 215 women, four had symptoms. All four tested positive. That means 211 women were asymptomatic, but 29 of those women also tested positive. Of those 29 asymptomatic women, four later developed symptoms, but 25 never did. If you were a woman of childbearing age who was infected, this study would indicate that there was a 72% chance you would never show symptoms at all. Furthermore, 13.7% of this sample of New Yorkers tested positive, when 1.2% of all New Yorkers have received a positive test.

5. About 2,000 of the 6,000 residents in Robbio, Italy, paid to take an antibody test, which found that roughly 13% to 14% of them had the disease, or about 265 people. These tests took place in early April. On April 2, Robbio had 27 positive cases reported via traditional testing — in other words, only 1 of about every 30 people with the disease was reported in the case count.

6. The town of Gangelt, Germany, has about 12,000 people and 500 of them were antibody-tested. About 14% had COVID-19. In the Gangelt region, about 1% of people tested positive by traditional means.

What did we learn?

(Christopher Cherrington | The Salt Lake Tribune)

Every study found a ton of completely or mostly asymptomatic carriers.

In fact, to get the number of truly infected people, you have to multiply the number of positive cases by a significant amount. For the most recent studies, the estimated multiplier was somewhere from 6 to 30.

Experts polled on the issue on April 5 had a median multiplier of about 7.5 for American reporting. I personally believe that’s a little low given the huge problems with testing in the country and the studies listed above, but we’ll go with that number.

That has huge implications for policymakers in Utah and elsewhere. Right now, the U.S. statistics show about a 4% death rate: that’s just cases divided by deaths. Utah’s is about 1%. But if you consider the undercounting of cases, as well as the fact that it takes sick people some time to die from COVID-19, you’re really looking at a death rate of something like 0.1% to 0.5%. The death rate for regular influenza is about 0.1%. This is good news!

But it also means that we’ve been underestimating how contagious the disease is. And that’s bad news.

In a previous column, we looked at how quickly changes in R0, which is science-speak for how contagious a disease is, can lead to exponential increases in the number of sick people. Back then, we highlighted the difference between COVID-19 having a R0 of 2 vs. 3. It now appears that was quite optimistic: One CDC study published last week estimated the base coronavirus R0 at 5.7. In other words, every person who has coronavirus gives it on average to 5.7 people. I am stunned by that number.

What do you do with a disease that’s just a little more deadly than the regular flu, but wildly more contagious?

It’s tempting to say “if it’s like the flu, let’s get back to work,” and we are starting to hear from people nudging states, like Utah, to ease up on the restrictions. That’s especially true with declining case counts here.

But if anything, these numbers should indicate the opposite action is prudent, because any back-to-work action has significant potential to ignite the virus back to life more quickly than previously thought possible. Even at flu-like death rates, having double-digit percentages of the population all sick at one time will overwhelm the health care system. Just look at New York.

Some have examined this data and suggested herd immunity as the best immediate strategy.

As a country, we are closer to herd immunity than previously thought, but it is still a long way off. On April 13, the U.S. had about 587,000 reported cases. Multiplying that by 7.5 gets you to about 4 million Americans who have had the disease, only a little more than 1% of the country. If herd immunity were the goal, we’d need somewhere between 60% and 90% of the country to have recovered from the disease. Even in New York City, we’re not especially close.

As of Tuesday, Utah had 2,412 cases. Our death rate is unusually low and our testing rate is relatively high, but still: Our official number likely underestimates the number of people who have been infected. I’m relatively confident that the true total lies somewhere between 5,000 and 30,000. That’s between 0.15% and 0.93% of the state’s population. Herd immunity is a long ways off.

Unfortunately, even with the lower death rate in mind, a herd immunity strategy would have a significant cost on human lives — somewhere between 200,000 and 1 million deaths nationwide.

Even though the coronavirus is probably significantly more widespread than the official numbers indicate, and significantly less deadly, we’re still between a rock (shutting down the economy) and a hard place (killing huge numbers of people) in terms of dealing with it.

The way out still requires a vaccine, the development of a very effective treatment, or large-scale test and contain measures for which we’re not currently equipped. Right now, the last option looks closest to coming to fruition.

So if you have any of the symptoms, get tested. Doctors, get your patients tested. If the state sets up a random testing program and you are selected, participate. The first step to opening Utah back up is through testing, and a significantly better understanding of COVID-19′s spread.