There Is More Than One COVID-19 Infection Fatality Rate

The wide range of estimates reflects real variations as well as methodological differences.


Two months ago, the U.S. Centers for Disease Control and Prevention (CDC) published "best estimates" implying that less than 0.3 percent of Americans infected by the COVID-19 virus will die from the disease. This month the CDC published a revised best estimate that puts the infection fatality rate (IFR) at 0.65 percent, more than twice as high. Meanwhile, the results of the CDC's seroprevalence studies in various parts of the country, which it first reported a month ago and updated this week, imply IFRs ranging from 0.1 percent in Utah to 1.4 percent in Connecticut.

The wide range of IFR estimates partly reflects uncertainty about exactly how deadly COVID-19 is. But it also reflects actual variations in how likely patients are to die in different places and at different times. Although people frequently talk about the infection fatality rate as if it were a single, constant number, it is bound to be affected by conditions, such as the patient mix and the quality and capacity of the local health care system, that vary from one place to another and that may change over time.

In May, the CDC's best estimate, included in its "COVID-19 Pandemic Planning Scenarios," was that 0.4 percent of Americans who develop symptoms of the disease will be killed by it. It also estimated that 35 percent of infections do not result in symptoms. Together, those estimates implied a national IFR of 0.26 percent. As critics noted, the CDC did not cite any sources for its numbers.

The new version of the planning scenarios includes a direct IFR estimate of 0.65 percent, based on preprint study that the Australian epidemiologists Gideon Meyerowitz-Katz and Lea Merone published on July 7. "Based on a systematic review and meta-analysis of published evidence on COVID-19 until May, 2020, the IFR of the disease across populations is 0.68%," they reported. "However, due to very high heterogeneity in the meta-analysis, it is difficult to know if this represents the true point estimate. It is likely that, due to age and perhaps underlying comorbidities in the population, different places will experience different IFRs due to the disease." They also cautioned that "given issues with mortality recording, it is also likely that this represents an underestimate of the true IFR figure."

The CDC has clarified its estimate by citing the source on which it is based, which is surely an improvement. But at the same time, the CDC has switched from an estimate of the symptomatic case fatality rate for Americans to an estimate of the infection fatality rate derived from research conducted around the world (which it revised downward slightly for reasons that are not explained in the planning scenario). Because of the high geographic variability noted by Meyerowitz-Katz and Merone, it is questionable whether the new estimate provides a more accurate picture of what is happening in the United States.

The CDC's seroprevalence studies seem more relevant to that question, although they have problems of their own. While virus testing in the United States has expanded dramatically since the early stage of the epidemic, it is still skewed toward people who have symptoms consistent with COVID-19 or who have reason to worry that they were exposed to the virus. The results of those tests are therefore a highly misleading indicator of virus prevalence, missing lots of people who have been infected but never sought testing because their symptoms were mild or nonexistent. According to the CDC's latest best estimate, 40 percent of COVID-19 infections are asymptomatic.

The CDC's antibody studies avoid the sampling bias of virus screening by using blood drawn from patients for routine tests unrelated to COVID-19. But it is still not clear that the patients included in the studies are representative of the local population, let alone the country as a whole. The CDC at one point was talking about conducting an antibody study with a random sample of the U.S. population, which would give us a clearer understanding of the national IFR, but I can no longer find any mention of that plan on the agency's website.

With that caveat in mind, the CDC's findings still reveal striking regional variations in both infection rates and implied IFRs.

The CDC estimates that nearly 7 percent of people in the New York City area had been infected as of April 1. That number had more than tripled by May 6, when the CDC estimates 23 percent of New Yorkers were infected. The estimated virus prevalence was almost 6 percent in Louisiana as of April 8 and about 5 percent in Connecticut as of May 3.

The CDC's prevalence estimates, when you take into account the dates when samples were drawn, indicate that the virus made much less progress elsewhere in the country during the spring. The estimated rate was less than 4 percent in Philadelphia at the end of May, about 3 percent in South Florida as of April 24, less than 3 percent in Missouri on both April 26 and May 30, about 2 percent in western Washington state on May 11, about the same in Minnesota on June 7, a bit more than 2 percent in Utah on May 3, and just 1 percent in the San Francisco Bay Area on April 27.

The ratio of estimated total infections to known cases also varied widely. The CDC estimates that infections outnumbered confirmed cases by at least 24 to 1 in Missouri as of late April, while the ratio in Connecticut was more like 6 to 1 around the same time. During the first round of testing, the estimated ratio was 10 or more to 1 in seven of the 10 areas where samples were drawn. The wide range of ratios helps explain the wide range of implied infection fatality rates, since more undetected infections means a bigger denominator.

The current crude case fatality rate in the United States—reported deaths as a share of confirmed cases—is 3.6 percent. Assuming that 10 to 1 is a reasonable stab at the ratio of total infections to confirmed cases, the nationwide IFR would be less than 0.4 percent, substantially lower than the CDC's latest estimate. But that ratio is a moving target, since expanded virus testing can reduce the gap between total infections and confirmed cases. That seems to be what happened in some of the locations studied by the CDC, since their ratios fell between the first and second rounds of antibody testing. Unreported COVID-19 deaths, the extent of which is also a matter of much controversy, complicate the picture further.

Still, a few observations seem reasonable based on what we know at this point.

First, COVID-19 patients do worse in some parts of the country than others, a fact that is apparent from the crude case fatality rates for various states as well as the IFRs implied by the CDC's seroprevalence research. While Utah's IFR (about 0.1 percent in early May, based on the CDC's data) may rise as the epidemic progresses there, it is unlikely ever to approach the IFRs in Connecticut (1.4 percent around the same time) or New York City (0.6 percent or more, depending on whether "probable" cases are included, based on an antibody study conducted by the New York State Department of Health in April). Reasons why COVID-19 is especially deadly in some places may include age demographics, the prevalence of preexisting medical conditions, stress on local hospitals, and factors (such as population density and mass transit use) that not only promote virus transmission but may increase the typical viral load.

Second, the IFR can change over time in any given place. If the average age of people infected by the virus goes up, for instance, the IFR can be expected to rise, since older people face a much higher risk of death. Conversely, the IFR can be expected to decline when the average age of COVID-19 patients falls, as it has been in the Sunbelt states that are driving the recent surge in new infections. Other things being equal, the IFR also will fall as COVID-19 treatments improve, another factor that helps explain why those states have not seen increases in daily deaths commensurate with the rise in confirmed cases.

Third, COVID-19 is clearly deadlier than the seasonal flu, but it does not seem to be nearly as deadly as the "Spanish flu" of 1918, contrary to the worst-case scenarios that had a profound impact on policy makers early in the epidemic. Even the CDC's revised IFR estimate is substantially lower than the rates assumed in some of those projections.

Taking into account recent increases in infections, independent data scientist Youyang Gu, who has a good track record of predicting COVID-19 trends, is now projecting that the U.S. death toll as of November 1 will be about 218,000. That is hardly a happy prospect, but it is a far cry from the 2.2 million deaths the Trump administration was imagining in March.

The argument about the extent to which COVID-19 lockdowns can be credited for the gap between such nightmares and reality will no doubt continue. But at this point, the ultimate death toll will depend largely on the precautions that Americans voluntarily take to protect themselves and their neighbors. Barring the unexpectedly early deployment of an effective vaccine, our fate is in our hands.

[This post has been revised to correct Gu's death toll projection for November 1, which was about 218,000 when the post was published; that projection had been raised to about 222,000 as of July 28.]

NEXT: Should Congress Take Down Its Statue of Racist Chief Justice Roger Taney?

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  1. “our fate is in our hands”?

    Well, not really. Our fate is partially in our hands, and partially in that of those we come in contact with, and partially in the hands of the medical professionals we have access to. And then there’s, you know, chance. If I could add a cubit to my stature, I’d be taller.

    1. Yeah. The idea that everything is within our control, and we can stop anything bad from happening if we just try hard enough and listen to the right people is the kind of agoraphobic, control freak thinking that got us to where we are. Much of our destiny is in fact out of our hands, so the best we can do is make pragmatic risk management decisions for ourselves and live for the moment.

      1. I read that sentence as meaning that we can choose whether or not to embrace the precautionary measures mentioned in the previous sentence. If we do so, the evidence suggests that we will reduce the death toll. If we don’t, it suggests that we won’t. This isn’t saying that we can stop everything bad from happening or that we shouldn’t make pragmatic risk management decisions. It sounds more like an assertion that, while we don’t have complete assurance that any given person won’t die in a car crash, we can reduce the number of fatal car crashes by implementing precautionary measures such as wearing seatbelts, having safety equipment, etc.

        1. But the risk of dying in a car crash can be reduced to zero by eliminating cars, which some people would do. Or slightly less ridiculous, the risk could be greatly reduced with a national 10 mph speed limit, which is more analogous to how we are now limiting other human behavior.

          1. It is almost like lie has risks.

            The think that Sullum also ignores is the increased risks from non covid patients not seeing specialists. Which is likely now greater than any benefit of mask usage against non mask usage.

            Hate to keep pointing it out but….

            The problem is, exclusively, racist institutions and policies.

            1. Whoops. Stupid clip board. This. Not the Tony quote.

              “Before this outbreak, the consensus literature was exactly what Dr. Fauci said it was originally, where there [did] not appear to be any studies that demonstrate the efficacy of masks in the transmission of what we call influenza-like illnesses, [which are] the types of viruses that are spread in respiratory droplets and cause respiratory illness,” Barbour said.

              “The concept of masks has been studied for a long time, especially in the dental community. They obviously work very closely with people, and a lot of literature has been written. There are some studies out there that show that dentists who wear masks and face shields have the same antibodies to the plethora of influenza and respiratory syncytial viruses, which sort of show that the masks were not effective in preventing the disease transmission in these types of professionals,” Barbour said.

              1. Google easily work and google pays me every hour and every week just $5K to $8K for doing online work from home. I am a universty student and I work on my part time just 2 to 3 hours a day easily from home.JEd Now every one can earn extra cash for doing online home system and make a good life by just open this website and Follow instructions on this page…

                ►══════► Online Income Websites

          2. Argumentum ad absurdum is a logical fallacy, dangerous and misguided. If we killed everyone, there would be no murders…so what is the point of laws against murders, since those who murder will disobey the law. Nor do speeders obey speed laws…so let’s eliminate them as ineffective.

            The flaw is in the assumption that if you can’t reduce all murders or accidents, that it’s useless to try. The actual truth is that we CAN reduce the bad shit, regardless of being unable to totally eliminate it. Nations with effective gun laws have reduced gun violence by 99% (Japan) tho there are still a few gun murders each year. Cars are safer than ever tho there are still thousands of fatalities each year.

            1. Wouldn’t use Japan as an example, yes they have low gun deaths but they also have very low crime in general. This is a cultural peculiarity, not because of harsh gun laws. You can give Britain as an example but they still have quite a few gun related crime even with draconian laws, not to mention any object they can get their hands on. They also have an almost blanket prohibition on self-defense to the point that a man who used a knife to thwart a man who was attempting to rape his wife was arrested for murder. I also don’t believe our gun murders are that high, the control advocates always quote the high number of deaths, but if you take out the suicides ( I thought we had as a country agreed that people are allowed to end their own life if they wished so this is more about the method than the result ), criminals killing each other ( are we supposed to believe those that killed Albert Anastasia would not have found a way to kill him much less your average Crip and Blood, again it is about the method and not the crime ) that leaves accidental deaths, murders by criminal of innocent victims, crimes of passion and or jealousy and finally the overly hyped but fairly rare mass shootings.

            2. Gun laws in Japan are not the reason for fewer murders in Japan.

            3. Nations with effective gun laws have reduced gun violence by 99%

              No, they haven’t. It’s funny to watch someone lecture others about informal logic fallacies, then in the next virtual breath just start pulling baseless assertions out of their ass.

              1. It’s a matter of definitions, I guess? If they haven’t reduced gun violence by 99%, clearly the laws were not effective. If they did, then they were effective.

      2. agoraphobic, control freak thinking

        ie the mind of a politician

        1. And Karen. Especially Karen.

    2. Or maybe out fate is in the hands of the idiots promulgating bad policy just because they have a new name for the flu this year.
      I almost, repeat almost, wish the press would go back to fixating on Russia/Trump.
      At least they left the businesses open then.

      1. I’d take Russia?Trump over this shit in a second. That was just annoying. This is scary.

    3. Our fate is not in our hands, because we are not *free to choose* the medicine and medical care we receive, but must get permission from the rent seeking medical mafia to have the legal right to purchase medicine, and the medical #DeepState prefers to protect their power and privilege over our lives.

      There’s much to condemn in the faux libertarianism of Reason, but the most glaring is their refusal to support our right to heal ourselves. They support our right to take heroin, but not hydroxychloroquine. There’s not the tiniest shred of intellectual excuse for this.

    4. Getting paid every month easily more than $15k just by doing simple job online. Last month i have exactly received $16839 from this online job just by giving this 2 hrs a day online. Now everybody on this earth can get this job and start earning more cash online just by follow instructions here……..for more info visit any tab this site Thanks a lotCopy Here………>>  Click for full details 

    5. My last month’s on-line earning was $17930 simply by doing a straightforward job on-line. best home primarily based on-line job to earn additional greenbacks monthly simply by doing work for optimum two to three hrs daily. I actually have joined this job concerning three months past and in my 1st month I actually have created $12k+ simply with none special on-line expertise. everyone on this earth will get this job nowadays and begin creating money on-line by simply follow details on this web site……… HERE══════►►►Only Here Click for full Details

    6. Start earning today from $600 to $754 easily by working online from home. Last month i have generated and received $19663 from this job by giving this only maximum 2 hours a day of my life. Easiest job in the world and earning from this job are just awesome. Everybody can now get this job and start earning cash online right now by just follow instructions click on this link and vist tabs( Home, Media, Tech ) for more details thanks…………..Join Here  

  2. As I understand it, the CDC’s claim that there are around ten people who have been exposed to/infected by the virus for every confirmed case is based on antibody testing. However, I’ve heard (and it even just showed up in a video playing to the right of this comment box) that antibodies may fall off to an undetectable level some weeks after infection. If that’s the case, then doesn’t that indicate that even more people may have been exposed than ten times the confirmed case rate, since they might test negative due to the reduced antibody levels?

    If we are lucky, even those whose antibodies drop to low levels may still have t-cell immunity and immune systems that will respond to the virus and generate new antibodies if re-exposed.

    1. BillE,

      The Ab tests for this virus are unlikely to give a positive for people who had a Corona virus in the last few years and have a head start against the virus that causes Covid-19. They will have other Abs that may not show up as Antibodies for this strain and may also have T-cell immunity against COvid-19 due to having a corona virus similar in 2018. This would mean that many more had been exposed in previous years to a similar virus (hmm, maybe those with kids in daycare). So, we may be closer to herd immunity than we thought.

    2. What % of people who died from the COVID-19 virus (confirmed by their overt symptoms) will test negative for antibodies? We have been told it takes two weeks to become positive. Once a person dies, the virus cannot reproduce and the immune system stops producing antibodies.

  3. if all the numbers matter, no number matters.

  4. Anyone looked at TFR or CFR compared to population density? Seems like there might be a correlation except as habits change with more or less mask wearing and keeping distance.

    Anyway just more evidence that lockdowns don’t work and there is no reason to destroy an economy just for an election. Oops sorry I forget who I’m dealing with here chaotic crazy nasty bitches in the Democrat party. Here’s mud in your eye you squad of harpies.

  5. I read that among dead people the rate is as high as 100%.

    1. Especially motorcycle riders in Florida.

    2. Heck, if dead people can vote in many places, why can’t they get ChiComVirus too?

  6. What factors with measurable ranges contribute to IFR?
    general state of health?
    number and severity of other existing health problems
    number of people in contact
    time indoors with other people
    density of people indoors
    age range of people in contact?
    time with symptomatic people
    density of symptomatic people

  7. “Third, COVID-19 is clearly deadlier than the seasonal flu, but it does not seem to be nearly as deadly as the “Spanish flu” of 1918, contrary to the worst-case scenarios that had a profound impact on policy makers”

    Yeh well, too late. Mask up!

    They will always act as though 2.2 million is possible. Even though it stands at around 600k for the entire globe. To the extent you can trust these numbers.

    I bet it’s half of what ever they claim. The co-morbidity aspect alone leaves me suspect.

    “our fate is in our hands.”

    It’s in the fate of incompetent buffoons actually.

    1. The real metric is “excess deaths”, to flush out the people who would have died of something else anyhow. It is always the case that confirmed deaths greatly exceeds the actual excess deaths.

      To be coldly analytical (see Ezekiel Emmanuel), the real metric to care about is years of life lost. While an 87 year old Alzheimer’s patient in a nursing home dying a year earlier than they might have is unfortunate, it is much less significant than a 25 year old nurse with a family to raise and 40 more years to work.

      1. I went to a funeral this week. He didn’t die of covid; he had cancer. But he may well have not died so soon if the medical industry in my state hadn’t decided to shut everything down for a few months. So excess deaths is not always a reliable indicator of the deadliness of the virus itself rather than our panicked response to it.

  8. Today Canada had under 300 cases. No deaths. Zero.

    Quebec mandates masks.


    1. and for Pete’s sake what of the Blue Jays!!!!!

  9. BillE,

    The Ab tests for this virus are unlikely to give a positive for people who had a Corona virus in the last few years and have a head start against the virus that causes Covid-19. They will have other Abs that may not show up as Antibodies for this strain and may also have T-cell immunity against COvid-19 due to having a corona virus similar in 2018. This would mean that many more had been exposed in previous years to a similar virus (hmm, maybe those with kids in daycare). So, we may be closer to herd immunity than we thought.

  10. “There Is More Than One COVID-19 Infection Fatality Rate”

    Well, of course; just like there is more than one truth.

  11. After containment failed the choices were, 1) isolate everyone to protect the vulnerable at all costs. 2) isolate the vulnerable at all costs. #1 was chosen and humanity will be paying the costs for one hundred years.

    1. Or at least until November – – – – – –

  12. When calculated out based on age, the results are that you will die at a rate equal to your current age + 2 to 4 years. Interestingly, your chance of being hospitalized for COVID is equal to current age + 2 to 4 years as well. People are going bat shit crazy worrying about both dying and being hospitalized with a long term illness. Going to the hospital because of COVID for a week isn’t that different than being hospitalized for a week (due to surgery or whatever). Now if you are in the hospital longer for COVID, well – I can’t find the numbers for extended hospital stays for non-COVID problems. So that could be much worse than current age + 2 to 4 years. But overall in ballpark of deaths so might make you feel like current age + 3 to 5 years instead. Not great if you are 80+. Not horrible if you 50. Laughable at 30.

  13. covid-19 patient increase day by day and death ratio is increase. if you want to safe your self do not move out from home without any work. wash your hand after some time . dont touch anything. use face mask .

  14. now in these days covid-19 patient are double within 3 days. Death ratio are also increase day by day . our medical team team is working to make coviud-19 waxin.

  15. corona virus patient death ratio increase day by day. if you want to save your life use sanitizer after some time. use face mask and avoid to touch anything.

  16. New York state’s Covid fatality rate is 168 per 100,000 people. Andrew Cuomo is the pandemic hero and is doing a great job.

    Florida’s Covid fatality rate is 26 per 100,000 people. Ron DeSantis is being labeled “DeathSantis”.

    Apparently success in getting the Covid death rate down depends on political party. Who knew?

  17. CDC’s use of antibody tests has a lot of false negatives: Other studies show that (1) many who have had (especially the mild cases) of the virus show up negative on antibody tests but do show specific T-cell reactivity to it; and (2) random samples from blood donors indicate that T-cell specific reactivity without antibodies is about as prevalent as the antibodies. Note that among the several studies of this, there is considerable variability in the ratios between T-cell specific reactivity and antibodies (probably due to geographic and other differences)…

    So the CDC’s measure of total cases based upon its antibody sampling may be understating the total-cases number by as much as a factor of 2.

    1. Yup. All the t-cell studies seem to be telling the same story: A significant portion of the country is already immune to this. The northeast is closed for no reason, their curve is done.

  18. Very efficiently written information. It will be beneficial to anybody who utilizes it, including me. Keep up the good work. For sure i will check out more posts. This site seems to get a good amount of visitors

  19. When you read this in a study “meta-analysis”, you should probably ignore it.

  20. There Is More Than One COVID-19 Infection Fatality Rate

    Third, COVID-19 is clearly deadlier than the seasonal flu


  21. The article suggests that the ONLY way we have to improve our odds of not dying from Covid-19 is to wear masks, avoid contact and sequester ourselves. What is NOT published is more important that what is. We see this in alarmist Media all the time. The CORRECT answer is that there is more that you can do. Here’s what we do:

    Take Quercetin 500 mgs once/day
    Take 500 mgs. Vit C once/day
    Take 11 mg zinc lozenges twice/day
    Take Mag 70 mgs once/day
    Take Vit D3 2,000 IU once/day
    Try to get some sunlight without sun block 10 minutes/day in the afternoon on your chest and arms. Don’t overdo it.
    I can go over the data and science on this regimen, but it would make the post too long and few will read it. This is what some prominent doctors who SEE ICU Covid patients do themselves.

    You CAN do more. You should do more.

    Sanjosemike (no longer in CA)
    Retired surgeon and prior owner of Medicare Certified Surgical Center

  22. If you can avoid the Cytokine/RANTES storm of enormous damage to your vasculature and lungs, it is VERY unlikely you will DIE from the virus. The key is avoiding the hyper-immune storm.

    Dr. Zev Zelenko has a regimen that is enormously effective, even on elderly and frail populations. It uses Hydroxychloroquine, high doses of zinc sulfate and Azithromycin or Doxycycline. He has excellent research data on this. He prevents the storm from developing. He starts this as SOON as a person gets infected.

    HCQ has been politicized. That will result in probably another 100,000 deaths. You can blame this on whom you wish. But I know where to accurately place the blame.

    Sanjosemike (no longer in CA)
    Retired surgeon and prior owner of Medicare Certified Surgical Center

    1. The key is avoiding the hyper-immune storm.

      The cytokine storm hypothesis is bunk (cytokine storm diseases are notorious for killing largely regardless of age and would run counter to the fact that the immune compromised are exceptionally susceptible to COVID) and, even if it weren’t, this doesn’t jive with the smorgasbord of… therapeutics you’ve recommended.

      I’m not in favor of masks and one-size-fits-everyone-indoors solution, but “Take Vitamins and get sunlight to support your immune system, but avoid the hyper-immune storm.” isn’t helping dispel that solution.

  23. Did you know there’s a “deep detox” you can do first thing in the morning to burn more fat? And the good news is It only takes 13-seconds! Here it is—>>Click For Full Detail Here.

  24. Is the CDC still conflating the COVID-19 Viral test results with the COVID-19 Antibody test results?

    I do not see any reporting of two distinct results, but instead just a single result which makes it appear that the CDC is rendering the data as worthless.

    There should be graphs of the COVID-19 Viral test results and graphs of the COVID-19 Antibody test results. If the data is kept separate you can still combine them into a totals graphs, but if the data is comingled into a single data set you can’t extract the differences.

    This is yet another example of where big government CDC has been a colossal failure and needs to be either eliminated or needs to return to its original mandate.

    We have a required point of failure at the federal level, but if left at the state level at least some percentage of the states would get it right. Instead here we are with the 100% failure of the CDC and FDA during the pandemic.

    Kudos to both the self-serving Democrat and Republican parties for building such a huge and essentially worthless bureaucracy.

  25. This is yet more evidence that we really don’t know shit. It is also more evidence that humanity’s attempt to control and tame nature in this case is futile. Every day there are new numbers, new estimates, and millions of self-appointed experts on social media. Whatever you believe about this virus, you’ll be able to find a scientific study that supports your position.

    We are tying ourselves into knots thinking that we can defeat a friggin virus. We can’t. We can’t even decide how deadly it is, and we can’t get any clear information.

  26. I have earned $ 18394 last month by W0rking Online from home. I am a full time college student and just doing this Job in my part time for maximum 2 hrs a day using my laptop. This Job is just awesome and regular earning from this easy home Job is much times better than other regular 9 to 5 office Jobs. I suggest you all to join this right now and start earning easily by just follow details on the given WebSite……..
    here……. ⇢⇢⇢⇢⇢⇢⇢⇢click for jobs its a limited offER.

  27. This is great article and give good knowledge about this topic satta king

Please to post comments