Recent reports have suggested that Covid-19 has become markedly less lethal in the United States. Our analysis of death rates and infection fatality rates from Arizona, the U.S. as a whole, and New York City shows it isn’t, indicating that public health measures to reduce infections should not be relaxed.

Determining the true fatality rate can also help identify why more people with Covid-19 are not benefiting from advances in care.

As of late July, the state of Arizona was reporting a fatality rate of 2.1% among people who contracted Covid-19. That is markedly lower than reports in the spring from areas such as New York City, which saw fatality rates as high as 10%.


The supposed reduction in lethality of Covid-19 has been largely attributed to improvements in treatments; earlier identification of Covid-19 infections, allowing for rapid medical intervention; and protection from infection of older higher-risk individuals. It also could be explained by increased testing, which would identify more mild and asymptomatic cases, leading to a lower reported fatality rate.

Knowing the true infection fatality rate (IFR), a key focus of research for one of us (D.L.R.), is essential in assessing whether regions are seeing improvements in the timeliness and effectiveness of Covid-19 treatment and, if they aren’t, rapidly identifying and implementing better practices.


The infection fatality rate represents the percentage of all people infected with Covid-19 who die from the disease. Getting a solid handle on the IFR is critical for accurate Covid-19 hospitalization and fatality projections, which are needed to guide public health measures.

To gauge possible changes in the lethality of Covid-19, we made a conservative estimate of the infection fatality rate in Arizona, which has seen the majority of its cases and fatalities since late June. We chose Arizona based on its excellent Department of Health Services data reporting site and its state-of-the-art hospital care system. We compared our estimate to the best estimate of the infection fatality rate across the United States by the Centers for Disease Control and Prevention (CDC) from data obtained earlier in the pandemic, primarily the spring of 2020.

To determine the infection fatality rate in Arizona, we divided the percentage of the state’s population who had died from Covid-19 as of July 30, 2020, by the 12.9% of the population that was infected based on antibody testing between July 20 and July 26, 2020. Antibody testing captures the total percentage of the population that had been infected with Covid-19 from the beginning of the outbreak. We then calculated and applied a standard correction factor for the delay between case diagnosis and death. This yielded an infection fatality rate of 0.63%, which is not significantly different from the CDC’s best estimate of 0.65% for the U.S. in the Spring of 2020.

Patrick Skerrett / STAT Source: Douglas L. Rothman

A similar value, 0.68%, was reported from an extensive meta-analysis of published reports from the United States and other developed countries through May 2020.

To independently assess if Covid-19 has become less lethal, we compared the ratio of deaths to hospitalizations reported by Arizona with the ratio reported by New York City. This comparison is independent of accurately knowing the percentage of the infected population. We chose New York City because it was among the earliest and hardest hit regions in the U.S., so improvements in the effectiveness and timeliness of treatments since then should be easily detected. There was surprisingly little difference in this ratio overall and within age groups, as shown in the chart below.

Patrick Skerrett / STAT Source: Douglas L. Rothman

It is unlikely that Arizona is an exceptional case: other states experiencing large increases in cases and fatalities since late June have reported similar fatality rates.

If Covid-19 is not becoming less deadly, what explains the five-fold lower reported fatality rate in Arizona now compared to New York City in the spring? The most likely explanation is that the large increase in testing since the spring has increased the number of diagnosed cases several fold. The reported fatality rate is calculated by dividing the number of deaths by the number of diagnosed cases. A larger number of diagnosed cases due to more testing would decrease the reported fatality rate. This conclusion is consistent with a CDC report that the true number of infections was underestimated in the U.S. during the March-to-May period by as much as 10 fold.

Given progress in the diagnosis and treatment of Covid-19, why has there been no apparent improvement in the infection fatality rate? The two main possibilities are that the improvements that have been made in treating Covid-19 are not enough to make a detectable difference in the infection fatality rate, or that a large fraction of those who die of Covid-19 do not get to the hospital in time for successful treatment. We were, unfortunately, unable to find the fatality rate of patients who were hospitalized. Such data would help distinguish between these possibilities.

We recommend that all states implement procedures to track the ratio of deaths to hospitalizations, as well as implement random testing studies to accurately track infection fatality rates. Without this vital information, our ability to improve the health infrastructure to treat Covid-19 is handcuffed and may well lead to deaths that could otherwise have been prevented.

Douglas L. Rothman is a professor of radiology and biomedical engineering at Yale School of Medicine. Jessica E. Rothman is a graduate student in biostatistics at Yale School of Public Health. Gerard Bossard is a freelance writer.

Editor’s note: The article was updated to clarify the difference between the true infection fatality rate and the reported fatality rate.

  • The infection fatality rate (IFR) is deaths/population while the case fatality rate (CFR) is deaths/confirmed cases. The IFR highly based on estimated while CFR is based on reported numbers. The IFR will always increase as more variables are introduced over time and the CFR can fluctuate due to outbreaks, healthcare, etc.

  • “The infection fatality rate, also known as the case fatality rate, represents the percentage of people with Covid-19 who die from the disease.”

    Ummm, yeah…not the same thing. Can’t really buy the premise of the article when you conflate CFR and IFR.

  • Infection Fatality Rate (IFR) and case fatality rate (CFR) are not the same thing! sheesh.

    This article is absolutely wrong and presents wrong data and comparisons!!!

  • It is impossiblle for you to prove anything in this article, and for the reader to understand what you are trying to say, because you are constantly ‘comparing apples to oranges. You use the same numbers of deaths (that’s good), but you use three different types of infections/cases/hospitalizations, mixing them up and misindentifying them repeatedly.
    You should define what each term is and then use those terms only as you defined them.
    Infections = number of all people who have been infected. (the largest number)
    Cases = number of infected people who see a doctor for treatment (a much smaller number, but a well known quantity)
    Hospitalizations = number of cases that required hospitalization (an even smaller number)

    And, of course, Infections fatality rate, Cases fatality rate, and Hospitalizations fatality rate are Fatalities divided by their respective portions of the population.
    One has to wonder whether the authors did this on purpose, to confuse the reader into believing something that is not true.

  • Hospitalization to death is interesting because I thought NY hospitals were especially inept, but apparently no better than Arizona. Comparing to flu is still tricky because they just estimate and extrapolate the IFR based on the CFR (where confirmed cases are basically all hospitalizations). They don’t go off flu antibody tests. And there was no mention of T-cell immunity, which would bring down the IFR considerably if it could be tested.

  • I don’t see how the difference in CFR between NYC and Arizona can simply be attributed to increased testing alone. How can you rule out improvement in treatments? In the absence of hard evidence, one must assume improved treatments is part of the reasons, perhaps even a major one.

  • If you are under 50 you have a better chance of dying from a lightning strike. We’ve quite literally destroyed the global economy over a flu like illness and propaganda things like this article are still trying to justify it. Disgusting.

  • CFR is the ratio of the number of deaths divided by the number of confirmed cases of disease. IFR is the ratio of deaths divided by the number of actual infections with SARS-CoV-2.

    If all cases are documented, the CFR and the IFR are the same. But this is generally not the case. While you did use Antibody testing to try to catch all cases, this is not always how the CFR is presented. The CFR and IFR are not always the same thing.

    • The Diamond Princess case was also a highly biased sample of people filled with retirees. That’s by no means representative to the general population.

Comments are closed.

Sign up to receive a free weekly opinions recap from our community of experts.
Privacy Policy