The data keep confirming, Covid-19 kills about 1% of everyone who gets it
The death rate is not the only important statistic.
There’s a lot of argument about how many people this damn virus actually kills.
Case fatality rates (CFR), which are what we usually see reported, are in the 2-5% range. The CFR is clinically important. It tells us what happens to people we know are infected, people who doctors come into contact with. Doctors want to know what is likely to happen with the patient in front of them, and the CFR is highly useful to them.
That rate ignores anyone who has the virus and isn’t tested. For public health and epidemiology purposes, we want to know the infection fatality rate (IFR). Of everyone who ever had this virus, no matter whether we ever knew it or not, how many die? How many have other bad outcomes?
The infection fatality rate is a simple number. It’s the total number of Covid-19 deaths, divided by the total number of people who ever were infected with the virus.
IFR = total deaths / total infections.
The problem is that both of those numbers are hard to come by. Covid-19 deaths are almost certainly being undercounted. We know that actual numbers of Covid-19 infections are substantially higher than the official count based on positive test results, but we don’t know know many times higher.
That number is critically important to the epidemiology and public health management of this epidemic, and it is very difficult to determine. But we have several reasonably good data sources for this now, and they all converge on a value very close to 1 death for every 100 people who get infected.
We had a fairly good estimate of that number very early in the pandemic. In late February, the World Health Organization released the Report of the WHO-China Joint Mission on Coronavirus Disease 2019 (COVID-19). At the time China was already doing massive contact tracing, and about a million and a half tests every week. They were detecting close to every new case, and doing so efficiently enough to allow them to drive the transmission rate to very low levels — for a time they had a quite accurate count of total infections.
Based on that data, the WHO/China joint report estimated that in the outlying regions, where health care systems were not overwhelmed, the infection fatality rate was 0.9%. In the core regions where hospital care was compromised, that number went as high as 3–5%, but for basic epidemiology, policy making, and understanding the likely effects of this pandemic, we’re more interested in the baseline number, which was just under 1%.
We‘ve known since February that the baseline infection fatality rate for Covid-19 is near 1%.
There have been several recent studies claiming to show a much lower value for IFR. Many of these have been tied to arguments that we are overreacting to this epidemic, that it’s not as dangerous as we’ve been told.
Perhaps the most infamous of these was COVID-19 Antibody Seroprevalence in Santa Clara County, California, posted as an unreviewed preprint on MedRxiv on April 14. They claimed that in Santa Clara County, 14–35 times more people were infected than official counts showed, and that therefore the infection fatality rate was 0.17%. This was only 1/6 of the 1% IFR value most had been using in epidemiology analyses and policy making, based on the early WHO/China data.
This paper has been strongly crtiticised on several grounds. Perhaps the most serious, is that the total number of positive test results in their study, was within the number of expected potential false positives based on their test specificity — that every single case they found could have been a false positive.
In any case, the WHO/China data, and the additional data I’ll discuss below, shows that they can not be correct, and that the actual infection fatality rate must be close to 1%.
New York City has been the epicenter of the Covid-10 epidemic in the United States. As of May 1, 18,200 people have died just in the city. This tragic number allows us to do some simple calculations.
NYC has a population of 8.4 million people. 18,200 deaths divided by the entire population, give a population fatality rate of 0.2%. 2 of every 1,000 people in NYC have already died of this virus. One of every 500. It’s already true in NYC that nearly everyone knows someone or knows of someone who has died of this virus.
That places a constraint on possible values for the infection fatality rate. If every person in NYC has already been infected, then the infection fatality rate would be 0.2%, the same as the population rate. That means the actual IFR must be greater than 0.2%. Note that 0.2% is already greater than the claim from the Santa Clara study
In fact, survey data by the New York Antibody Testing Study, released on
April 23, found that about 1 in 5 people in New York City tested positive for Covid-19 infection. 1/5 of the people infected, means that the infection fatality rate is 5 times higher than 0.2% — which means the IFR in New York city is 1%.
Additional more recent data from the survey show that closer to 1/4 of NYC has been infected, which would drop the IFR to 0.8%. These results tell us that in NYC, Covid-19 is killing between 0.8–1% of everyone who gets the virus.
Similar calculations can be done for the entire state of New York, or for other region, and the resulting IFR values keep coming out between 0.7% and 1%. Preliminary data from Europe, time adjusted for lag between infection and death, arrives at similar values.
Remember that deaths are also hard to count. Official Covid-19 deaths don’t include anyone who dies while not under medical care. We know that cities that have been getting hard hit with coronavirus, have also been reporting very large increases in unattributed excess deaths, many of them from people dying at home.
It is almost certain that the actual number of deaths from this virus in the United States, and in New York, are much larger than the actual count. Including unattributed COVID deaths would raise the calculated infection fatality rate from Covid-19.
The US Centers for Disease Control attempts to estimate this, using excess mortality calculations. The estimates are that Covid is probably killing about 30% more people than official counts, with pretty wide variation in that estimate.
If we use the excess mortality data to correct the number of deaths by 30%, then the infection mortality numbers get adjusted upwards by about 30%. From 0.7–1%, we get adjusted values of 1–1.3%.
0.9% for China. 0.7–1.3% for New York. Values near 1% from Europe. The values for the infection fatality rate keep converging very close to 1%. Under current conditions, it seems reasonably certain now, this virus kills about 1 of every 100 people who get it.
This virus is highly infectious. In NYC now it has already infected almost one of every 4 people, and it’s still spreading. If it escapes the tenuous slowing we’ve imposed on it with our shutdowns, and infects 1 of 4 people in the United States at large over the course of this year, that’s 83,000,000 people infected, 830,000 deaths. If it grows and infects half of us, that 1,650,000 deaths.
Those are bleak numbers. We’re hoping for a vaccination, for effective therapies, for ways to slow the spread. Without those — these numbers are what this virus is capable of doing, and what should be driving our public policy decisions.
The focus is on death, perhaps justifiably so. But focusing on death blinds us to a lot of other effects of this virus. We know now that among those who get seriously ill and survive, including many people in younger age cohorts with lower death rates, there are other serious long term effects besides death: permanent lung scarring, heart damage, liver and kidney damage, nerve damage, alterations in blood chemistry and blood flow, higher rates of heart attack and stroke, and the list is still growing. Public policy that focuses only on death rates, misses the personal and economic costs of all these other kinds of damage. They too have to be as accurately estimated as possible, and included in our public heath policies around how to manage this virus.
