Did the coronavirus get more deadly? The death rate, explained

Trump says Covid-19 isn’t that deadly. Here’s what we know.

Eight weeks into an epidemic that has killed more than 4,000 people, the World Health Organization — for the first time — released a Covid-19 case fatality rate (CFR) on March 3.

It was not a modest estimate.

“Globally, about 3.4 percent of reported COVID-19 cases have died,” said WHO Director-General Tedros Adhanom Ghebreyesus at a press briefing last Tuesday. The rate describes the proportion of deaths among confirmed cases. And it was greater than the previous coronavirus CFR estimate (2 percent in China), far higher than the seasonal flu (which kills 0.1 percent of those infected on average), and even worse than the Spanish flu pandemic (which killed an estimated 2 to 3 percent of those infected).

It’s no surprise that, very quickly, media around the world interpreted the news as evidence the disease is deadlier than feared, and Twitter was buzzing with speculation about Covid-19’s real death rate. The number spread so far so fast, even President Trump took to Fox News on March 5 to try and tamp down the worry: “I think the 3.4 percent number is really a false number. Now, this is just my hunch, but based on a lot of conversations … personally, I’d say the number is way under 1 percent.”

In the end, the 3.4 percent figure appears to have sowed more confusion than clarity. While a WHO spokesperson called the number “the current global ‘snapshot,’” it misled some: it was delivered without context, omitting key factors that determine a disease’s severity. In reality, we still don’t know Covid-19’s precise death rate — but we do have some guesses based on models that are more sophisticated than just dividing the deaths by cases.

Epidemiologists and disease modelers studying Covid-19 told Vox a more reliable global case fatality rate is about 1 percent — but there’s still a lot we have to learn about the disease. And even when we better understand how deadly this virus is, it’s likely to look pretty different country to country.

What we know about how deadly the coronavirus is

One key piece of information we still need is how many people in a population have the virus. That’s the denominator in the CFR equation.

When Tedros made his remarks, WHO had recorded 3,112 deaths and 90,869 cases. Dividing the deaths by the cases, you get 3.4 percent. But that denominator is actually pretty fuzzy. So far, many countries — even those with sophisticated health systems, like the US — have struggled to get adequate testing up and running. And, to date, much of the focus has been on checking people who traveled to affected countries, not the broader population.

That means, in many places, health authorities are only picking up a subset of the sick — and perhaps the most severe cases, given that roughly 80 percent of people with Covid-19 have a mild disease. (Remember: The sickest are usually the ones who show up at doctor’s offices and in hospitals, while there may be hundreds or thousands of others with the virus who never show symptoms or don’t bother going to see a doctor. That’s why the CFR can often look much worse in the early days of an outbreak.)

Some of the countries that have done more widespread testing appear to have relatively fewer deaths in the case mix. In South Korea, for example, where they are testing thousands of people every day, they’ve picked up more than 7,500 people with the virus. Among those, 54 have died. If we use WHO’s method of calculating the CFR — and don’t take into account the potential problem of under-estimating the number of mild cases out there — a crude case fatality estimate hovers under 1 percent.

Contrast that with the US: 26 people have died and there are currently 754 known cases. That’s a crude CFR of 3.4 percent. This doesn’t mean the virus is deadlier in the US; it’s more likely, once again, an artifact of undertesting or some other factor simply dividing deaths by cases doesn’t account for. (We know that, before CDC removed its testing numbers from its website on March 2, fewer than 500 Americans had been checked for the virus.)

“When we see systems that are able to account for not only the really sick, but also those who have infection and are outpatients, the CFR is more accurate and drops considerably,” said Isaac Bogoch, a professor at the University of Toronto.

There are other problems with the 3.4 percent CFR: Global estimates hide regional variation that’ll likely emerge because of variations in health system capacity. Consider the death rate of the 2014-’16 Ebola epidemic. The disease was far deadlier in West Africa, where it hit some of the poorest countries on earth, than it was in America, where nearly every person who fell ill had access to state-of-the-art treatment.

As of February 20, the WHO also reported lots of regional disparities in the disease’s deadliness even within China: 3.8 percent on average, but 5.8 percent in Wuhan, Hubei province (where the virus first emerged) and 0.7 percent in other areas in China. (A WHO spokesperson acknowledged these regional differences and said we’re likely to see something similar “in different countries and outbreaks.”)

One final consideration: CFRs do change over time. That’s exactly what happened in China, as you can see in this figure from the WHO. Even the first and hardest-hit province, Hubei, saw its death rate tumble as public health measures were strengthened and clinicians got better at identifying and treating people with the disease:

WHO

As this pandemic threat wears on, we may also see more deaths — which could cause the CFR to increase if those mild cases don’t appear. “It takes a long time to die” from Covid-19, said University of Toronto epidemiologist David Fisman. “We estimate the length of stay in cases who died in ICU in China to be 28 days on average — three days in hospital and 25 days in ICU.”

But most experts anticipate the more we test for Covid-19, the the more mild cases we’ll find. “To get a [true] rate, you need the deaths — which you can more reliably pick up — and you need cases, which you can less reliably pick up,” Lawrence Gostin, a global health law professor at Georgetown University, summed up. “It’s not irresponsible to come out with that [3.4 percent] number, but it should have been more clearly interpreted as not being reliable, or at least mention it’ll vary in regions.”

A better way to calculate the Covid-19 death rate

NEJM

The ideal way to arrive at an accurate CFR involves a survey of the populations affected to find out who has antibodies for the virus, including the folks who didn’t even know they had it, said Maia Majumder of Boston Children’s Hospital. That’ll give experts the denominator — the real case toll — in the CFR equation. “Until we’ve done [that] — and I’m sure it’ll happen sometime in the future — there are going to be some people that have mild infections or are asymptomatic infections that we’re not picking up.”

As it stands, no such serology test has been completed — though China has approved at least two, and Singapore says it’s already using an experimental test to check for antibodies in patients there.

In the meantime, we have disease modelers who can give us a provisional portrait of Covid-19’s deadliness accounting for things like the delay in the reporting of deaths and potential uncounted cases.

In a new — but not-yet-peer-reviewed — pre-print out of Switzerland’s University of Bern, researchers did just that, using data on the outbreak from China’s CDC. They came up with an overall CFR of 1.6 percent — suggesting Covid-19 is less deadly than SARS, which had a CFR of around 10 percent. It’s also less than the WHO’s estimate.

Importantly, the researchers also found a wide variation across the age spectrum: less than one percent of people in their 20s die from the virus, they estimated, while the rate shoots up to 18 percent for 80-year-olds.

Christina Animashaun/Vox

“This 1.6 percent [figure] is based on a lot of assumptions we feel comfortable with but it’s not hard data,” Julien Riou, a Bern research fellow and author on the paper, emphasized. “It’s more [an attempt] to correct the biases we know about.” It’s also based solely on China’s data — and, again, the outbreaks in other countries will likely play out differently.

So these figures are not only likely to change within China and outside over time — once we have better data, they’re going to look different depending on how strong a country’s health systems are. But it’ll be a while before we know the true number of cases and have a better understanding of the deaths. Until then, when you see case fatality rates floating around, ask what they’re based on.