Case Fatality Rate – What does it tell us?

By Foss Tighe

During the last presidential debate, to counter Biden’s criticism of the President’s handling of the Coronavirus pandemic, President Trump referred to declines in the US Case Fatality Rate (CFR). For months now, Trump has be bragging about the decline in the CFR. What exactly is a Case Fatality Rate, what is the COVID19 CFR? And what do changes in the CFR mean?

In the abstract a fatality rate is simply the relationship between the number of people having a condition and the number of people dying from that condition. But a “Case Fatality Rate” is a little more specific, it measures deaths caused by a disease as a percentage of those who have been identified as having a case of the disease. For some diseases this distinction would not be all that important, but with the coronavirus, where many people have no symptoms, or may have mild symptoms easily confused with other milder illnesses, the measurement and meaning of Case Fatality Rates gets trickier.

So lets talk real numbers. One way to calculate the coronavirus CFR in the US is to take the total number of deaths and divide by the total number of cases. According to as of October 23, 2020 those figures were:

Deaths/Cases: 229,284/2,819,508 = 0.08132
Expressed as a percent: 0.08132 * 100 = 8.132%

This is a very scary number. It says that over 8% of the people in whom Coronavirus is detected will die. If this accurately represented the current CFR, we would expect far more deaths than we currently see. For the last two months we have averaged about 50,000 cases per day, 8.132% of 50,000 is a little over 4000 people. So if our CFR were 8.132% we would expect to see an average of 4000 deaths per day in the US. But instead we see an average of around 700. It turns out the CFR was very high during the early months of the pandemic and has declined steadily since then. Lets look at cases and deaths since August 1, 2020, again using data:

Deaths/Cases: 71,031/1,978,699 = 0.03590
Expressed as a percent: 0.03590 * 100 = 3.59%

Why has the Case Fatality Rate declined?

  1. Have our treatments improved?
  2. Has the infection become less deadly?
  3. Has our ability to detect cases changed?
Has Treatment Improved?

The ability of our hospitals to treat severely ill COVID-19 patients has improved a small amount. Physicians have learned how to better manage the severely ill. Dispite a lot of aspirational thinking by the President regarding the benefits of the malaria drug Hydroxychloroquine and the Anti-viral developed for Ebola remdesivir, neither have been shown to significantly reduction in mortality. However, a treatment regime studied in the UK using a relatively old steroid called Dexamethasone, has been shown to reduce mortality in severely ill patients. Dexamethasone and improved management techniques of the severely ill only account for a part of the reduced CFR. In recent weeks there have been a couple studies showing the decline in mortality among hospitalized patients. Figure 1 below shows a chart from the pre-print version of one of these studies.

Figure 1: Change in mortality in a single health system in NYC:

Figure 1 has a lot of stuff. Lets just look at a couple things. The gray bars are the number COVID-19 admissions, with the scale shown on the right side, with the time scale in 2 week increments at the bottom. So we see the really high number of admissions in April when New York was the center of the US outbreak. The blue and orange lines represent two different measurements of the death rate for those patients admitted, using the percent scale on the left over the same time period. The blue line is “unadjusted” mortality where no attempt is made to take into account any changes in the patients over time. This unadjusted mortality shows a huge drop for 30% mortality at the start of the outbreak, declining to just 3% by late June. This is represents a 90% reduction in the number of people dying after hospitalization between March 2020 and June 2020. However, in the adjusted mortality, which makes adjustment for change in the patients being admitted, the changes were not as great. The adjustment took into account age, underlying conditions, race, severity of symptoms at admissions and level of virus found in the patient. With these adjustments the blue line shows mortality starting at 23% and falling to 8%, showing a 65% decline in the number of COVID hospitalized patients dying in this Hospital system in New York. The triangles indicate data points that are statistically significant, that is data points that are highly unlikely to represent just random variability in the data. Only one of the data points for the adjusted mortality falls in the category of statistically significant, somewhat weakening the strength of the evidence here.

The authors also point out that during the start of the pandemic, the hospital system was under intense pressure due to capacity limitations. Though not specifically measured in this study, it is a well documented finding that hospital outcomes worsen when the staff and facility in general is under stress. So some portion of the improvement after April might be explained by reduce stress on the health care system itself. Nonetheless, it is clear that at least some portion of the decline seen in the Case Fatality Rate is due to improvement in treatment among hospitalized patients.

Has the Virus Become Less Deadly?

Though viruses evolve quite quickly compared to human beings, there is no indication that the ability of the virus to kill has changed radically since its discovery in China. Given the reduction in travel around the world, we would expect any significant evolution in the viruses ability to kill to first show up in one part of the world and gradually spread. The fact that much of the world seems to experiencing death rates based on age and underlying illness in a more or less uniform manner makes it un-likely that there as been a significant evolution in the virus because if an evolution occurred, it would not occur simultaneously around the world.

Has Our Ability to Detect Cases Changed?

The ability of the health care system to detect cases of coronavirus infections has changed dramatically since the start of the pandemic. At the start of the pandemic in the United States our ability to test for the virus was very limited. Supplies of test kits were so limited that there were severe restrictions of who could even be tested. Even after community spread had been detected in the United States, CDC recommendations remained that only people with recent travel to international hot spots, or people with contact to someone already infected were to be tested. Most testing was occurring at hospitals as severely ill patients were arriving in ambulances. People with mild symptoms were unlikely to get tested, and testing for asymptomatic cases was virtually non-existent. Figure 2 shows the growth in Coronavirus testing in the United States from March 6th 2020, through October 23, 2020. On March 6th there were only 263 tests performed, on October 23, 2020 there were over 1.2 million.

Figure 2: US Coronavirus Tests Administered Per Day –

This significant increase in testing has greatly increased the number of cases detected, particularly among those with mild and asymptomatic cases of infection. This increased the denominator in the Deaths/Cases calculation, thereby leading to a systematic lowering the CFR over time.

There is a certain irony in President Trump pointing to the decline in CFR as proof of his adept handling of the pandemic. In June, when cases were rising in Florida, Texas and Arizona, the President was quick to blame increased testing as the cause. Even suggesting that perhaps we should test less so that things would look better. However, in part, a direct result of that increased testing that he can now brag about how low our Case Fatality Rate has become.

What about Death Counts?

We have discussed how changes in treatment and case detection can impact the CFR, it is also important to remember that CFR is the relationship between cases and deaths. How deaths are counted can also impact CFR. You might expect deaths to be a less ambiguous measure than cases but death count methodologies can also change. Figure 3 shows daily deaths in New York, based on

Figure 3: Daily Deaths in New York:

On May 6th, New York reported a spike of nearly 500 additional deaths. It turned out this reflected a change in how New York counted deaths. They started including deaths from people who had not died in a hospital and where COVID was merely presumed to be the cause death. Many states now include both confirmed and presumed cases in their death counts.

Even with the inclusion of presumptive cases, there is a general belief that death counts represent an undercount. CDC presents data showing that during the months of the pandemic, the number of deaths reported were higher than what would have been expected based on historical data. These “excess” deaths figures are consistently above expected deaths even when taking into account reported COVID deaths. The assumption here is not all COVID related deaths are being reported as such. This is the same approach used to estimate the number of deaths caused by the 1918 flu. Using this notion of excess deaths, the CDC data shows that an additional 10 to 30% of deaths above those reported as COVID deaths, may be attributable to COVID. Not surprisingly the highest rates of excess deaths were during the early months of the pandemic. Most calculations of CFR do not use this excess death method for counting COVID deaths, but instead use counts based on probable and confirmed COVID on death certificates.

Are there others more useful measures?

The biggest problem with Coronavirus Case Fatality Rate is the “Case” part. What people really want to know is “If someone gets infected, what is the chance that they will die?”. But that is a different measure, that is the Infection Fatality Rate (IFR). The COVID-19 IFR has been intensely debated. Based on data from China there was fear that the IFR might have been as high at 3%. The problem with calculation IFR is that you need to know all the people who have been infected, not just those that were detected as infected through testing. According to some experts, the IFR for COVID is around 0.6%. This is still 6 times as deadly as the annual flu.

As we have seen the main problem with CFR is measuring mild and asymptomatic infections. I look to outbreaks where the detection of cases was as close to complete as possible to get an estimate of the high end of what true infection fatality rate might be.

Outbreak LocationFatality Rate
Diamond Princess1.8%
New Zealand1.3%
* only included cases and deaths prior to 9/1/2020 – Iceland has a current a new outbreak for which deaths may be lagging
Figure 4: Fatality Rates for selected outbreaks where case detection extensive – 10/24/2020

In the above 4 outbreaks officials aggressively tested for cases far more extensively than was done in most locations. But even in these locations it is likely that some mild and asymptomatic cases were not detected. The Iceland figure is very close to the public health consensus of 0.6%

Things CFR can tell you

So Case Fatality Rates may not be consistent and comparable with each other over time during the pandemic, but in the short term they do provide the ability to predict the near future. For example, nationally the CFR for the last few weeks has been running hovered around 1.75%, with around 40,000 cases per day and about 700 deaths per day (700/40,000 = 0.0175 or 0.175%). What will happen now that daily cases have risen to 60 or 70,000? Well, for me its a pretty good bet that deaths will soon rise to over 1200 (70,000 x 0.075 = 1225). Exactly how soon will that happen? Estimating the lag between changes in daily cases and daily deaths will be the subject of a future post. But my guess, based on the recent increases is case counts is that we will have over 1000 daily deaths by the election and over 1200 hundred deaths per day by mid-November.

The Case Fatality Rate can also tell shed light on the thoroughness and adequacy of the testing being done in any given area. Though there is a slight variation in the age distribution and risk factor distribution across states and there is also some variation in the quality of health care across states, for the most part we would expect state level CFRs to be relatively similar to the overall U.S. State level CFRs that vary significantly from the national average, for the same period of time, would likely indicate that the state is either doing better or worse than the country as a whole on testing. A future post will look at this issue.


So the Case Fatality Rate in the United States is declining. During the early days of the pandemic in the US CFR was over 8%. Looking only at cases since August 1st, the CFR drops to 3.5%. Looking at only the last few weeks the rate appears to be about 1.75%. The decline in Case Fatality Rates is likely due to improvement in hospital care for the severely ill, and a significant expansion of testing which means more mild and asymptomatic cases are being detected. The CFR can be used to make near term mortality predictions based on case counts as well as make broad assessments of testing quality and adequacy.

4 thoughts on “Case Fatality Rate – What does it tell us?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Create your website with
Get started
%d bloggers like this: