When the head of the World Health Organization said this week that the new coronavirus’s death rate was an estimated 3.4 percent, the figure seemed to shock both experts and President Trump.
What a 3.4 percent death rate looks like:
“I think the 3.4 percent number is really a false number,” Mr. Trump said in a Fox News interview. “Now, this is just my hunch, but based on a lot of conversations,” he added, “I’d say the number is way under 1 percent.”
By definition, the case fatality rate is the number of deaths divided by the total number of confirmed cases, which appears to be what the W.H.O. did to arrive at its rate.
Is 3.4 percent a misleading number? We spoke to a number of experts in epidemiology, and they all agreed that 1 percent was probably more realistic (the W.H.O. has also said the number would probably fall). But they also said evidence about the spread and severity of the disease was still too new and spotty to know for sure.
The fatality rate is a key figure that public health officials use to respond to disease outbreaks. The more deadly a disease, the more aggressive they’re willing to be in disrupting normal life. But current data allows scientists to measure only a crude statistic called the case fatality rate, which is based on reported cases of an illness. Eventually, they hope to have a more comprehensive number called the infection fatality rate, which includes everyone who is infected with the virus.
“It’s essential for understanding how big our response should be,” said Marc Lipsitch, a professor of epidemiology at Harvard. “All responses have costs. If we think the risk is higher, then we should be willing to tolerate bigger costs, more inconvenience and the mental health loss from social distancing.”
There are several reasons we still don’t know the right number. Insufficient testing, for example, may be making the fatality rate look larger than it actually is — but deaths where a coronavirus infection was never diagnosed could make it look smaller. These are the key biases that epidemiologists and public health officials think about when looking at the case fatality estimates so far, and how they might change in coming weeks and months.
Not enough people have been tested
The fewer people you test for a disease, the fewer infections you are going to measure. In the United States, until this week, the only people being tested for the disease were those who had traveled to China or were known to have had contact with other ill people. Those strict standards were driven in part by a shortage of reliable tests. But we now know that there were many infected people in the country who weren’t being counted.
An example of how the death rate could go down
Think about that problem on a much larger scale. If there were a magical way to test everyone in the world for the disease, we would know exactly how many people have the infection. Discovering every case would tend to drive down the fatality rate, since the number of deaths would be divided over a much larger number of living infected people. There is increasing evidence that some people infected with coronavirus have few or no symptoms. Those people are the least likely to seek or receive tests.
Limited testing in many countries means that the reported death rates probably skew high.
“Since most cases are mild, and testing has not been universal, almost by definition we are failing to detect and therefore count all of the cases,” said Mark Lurie, an associate professor of epidemiology at Brown University.
Over the long term, epidemiologists often do a kind of blood testing of large numbers of people in a given community. By testing their immune systems, they can measure how many people have been exposed to a disease. That type of research is often the gold standard for getting a real infection rate and a better fatality rate, called the infection fatality rate. The infection fatality rate for the flu, for example, is about one tenth to two tenths of 1 percent — far lower than any of the estimates for the coronavirus. But that measurement technique is most useful after a disease has already spread widely, so it can’t be easily used now.
The number of deaths could be wrong
Compared with infections, deaths are relatively easy to count, especially now that we know that this disease exists and what its symptoms look like. But public health experts say we still may not have a complete count of all coronavirus deaths. In some countries, frail people have died of pneumonia and weren’t tested, including an elderly Spanish patient who was tested for the coronavirus only after his death. If sick people are dying without going to a hospital, they could be missed.
An example of how the death rate could go up
But the biggest challenge for measuring deaths right now is that people can be infected with coronavirus for a long time before becoming sick enough to be at risk of death. Currently, we are counting everyone who tests positive for the virus as infected and alive. But, in the future, some of those people will die of Covid-19, the illness caused by the virus.
Justin Lessler, an epidemiologist at Johns Hopkins, was part of a team of scientists who studied a group of Covid-19 cases in Shenzhen, China. He found that most people who died had been sick for longer than 30 days.“Think of when all the cases outside of Hubei have occurred,” he said of the province whose capital is Wuhan. “If it’s 30 days or even two weeks, we’re really at the tip of the iceberg.”
Generally, epidemiologists like to measure the fatality rate for a disease over a set period. They look at everyone who gets sick and see how many are still alive over weeks, months or years, depending on the disease. So far, scientists have been unable to do those kinds of studies for the novel coronavirus.
Conditions in countries vary
Right now, the global estimates are combining deaths and cases from countries around the world with very different populations and different health systems. But experts say differences between populations in each country and in the nations’ health systems may make death rates higher in some places than in others.
Examples of how death rates can vary between people …
The risk factors for death or severe illness from coronavirus are still being studied, but there is strong evidence that older people are at a higher risk of dying. There are very few documented cases of children who have developed serious illness. A disproportionate number of deaths have been among patients older than 65. The share of people over 65 in China is 11 percent, and in Italy it’s 23 percent.
In the United States, it’s 16 percent. Countries like Italy, with more older people, may end up with a higher rate of death.
Smoking may also play a role, evidence suggests, and the smoking rates in different countries vary considerably. Smoking among men in China is common. In the United States, smoking rates are substantially lower. Other health problems, like diabetes, cardiovascular disease and lung ailments like asthma, may also predispose people to a greater chance of severe illness, though the effects are still being studied.
The sophistication and capacity of the health care system most likely matters a lot, too. Patients with severe Covid-19 often need complex care for pneumonia and respiratory failure, sometimes including mechanical ventilation. The quality of that care will probably depend on the availability of ventilators and trained staff to monitor them. “When facilities got overwhelmed, there were more deaths,” said Dr. Thomas Frieden of the experience in China. Dr. Frieden, who was the director of the Centers for Disease Control and Prevention in the Obama administration, said that when he was in government, he worked to expand the country’s strategic reserve of ventilator machines. Whether there will ultimately be enough hospital capacity for everyone with serious illness in the United States depends on how quickly and broadly the virus spreads.
Researchers are racing to develop treatments for the disease, as well as a vaccine. Once there are better ways to help people who are infected, the fatality rate may go down for everyone.
Eventually, scientists should be able to offer still more granular estimates of risk. This would allow people of different ages and health histories, in different countries, to estimate their risk of serious illness or death.
“When I looked at the 3.4 percent number and where they got it, I thought this is both wrong and irrelevant,” said Dr. Ashish Jha, the director of the Harvard Global Health Institute. “It’s not relevant to nearly any single person. This is a worldwide average.”
As Dr. Jha noted, most people want to know their personal risk, not the risk for the average person worldwide. Developing estimates with that level of nuance will take even longer than building a more reliable infection fatality rate.