COVID-19 case fatality rates

Why do fatality proportions differ between countries?

If 1,000 people became infected with the novel Coronavirus, how many people may die?

This article looks at the COVID-19 case fatality rate. I also look at why this proportion differs between countries.

Challenges and caution

Estimating the deadliness of a new virus is challenging.

In opening remarks on 3rd March, Dr Tedros Adhanom Ghebreyesus (WHO Director-General) said:

Globally, about 3.4% of reported COVID-19 cases have died. By comparison, seasonal flu generally kills far fewer than 1% of those infected.

At that time, dividing total deaths (3,110) by total number of confirmed cases (90,893) gives 3.4%. On 29th January, that proportion was around 2%.

As of 9th March 2020, Johns Hopkins University states there are over 113,000 COVID-19 cases. The total number of deaths was about 4,000.

We should treat the case fatality rate with caution during an epidemic. Two competing biases affect this proportion: undetected cases and future deaths.

Undetected cases

The standard calculation uses the number of infected people. Usually, this is the recorded number of patients who have the virus.

More serious cases may need urgent healthcare. Rest at home could treat milder cases. If mild cases go undetected, those cases are missing from total confirmed patients.

Healthcare systems could mistake COVID-19 cases for influenza. We could be understating the true total cases. There is uncertainty about how many people have the COVID-19 virus.

Based on confirmed cases, the case fatality rate could overestimate COVID-19’s virulence.

Confirmed cases could be the ‘tip of the iceberg’. (Image: Swiss Medical Weekly)


The current case fatality rate could be an underestimate.

People receive diagnoses near the start of their illness. Patients recover or die away at the end. A new case will not become either of these outcomes for several days or weeks. The outcomes of latest cases are not captured when looking at the snapshot.

True fatality may be higher. (Image: New York Times)

Attribution is also difficult, if a patient was suffering from many ailments. Healthcare workers could diagnose a COVID-19 death as something else. Due to overwork, those services could delay reporting deaths too.

Differences between countries

As of 9th March, the case fatality rate in Hubei was 4.4%, compared to 0.9% elsewhere in mainland China. In the United States, there were 607 cases of the Coronavirus, and 22 deaths. That is a case fatality rate of 3.6%. In South Korea, there are 53 recorded deaths — 0.7% of over 7,400 cases of Coronavirus in that country.

Case under-reporting and death delays affect the estimates in countries. These competing biases will also differ between and within countries.

Different countries may have different standards for confirmed cases. Usually, we need positive test results. Clinical diagnoses using scans were used in Hubei for a short period.

Some countries have large testing programmes to identify patients. South Korea has tested over 66,000 people for COVID-19. The United Kingdom has tested in excess of 24,000 people.

There is a relationship between age and COVID-19’s fatality. A study by China’s Center for Disease Control and Prevention analysed over 72,000 COVID-19 cases. The central estimate of the virus’s case fatality rate was 2.3%. Among those aged 80 or over, that rate was 14.8%. The demographic make-up of a country affects its case fatality rate.

Supportive care is important for severe diseases. The capacity of healthcare systems will influence a country’s case fatality rate.


Researchers can use statistical modelling to overcome these issues. Riou, Hauser, Counette and Althaus (University of Bern) fitted a transmission model. They fitted the model to cases and deaths in Hubei, before 11th February 2020. The model then produced estimates of an adjusted COVID-19 case fatality rate of 1.6%. The credible interval is between 1.4% and 1.8%.

This rate is age-specific: for those aged 80 or over, the plausible interval ranges from 14% to 22%.

Researchers can model deaths after the reporting period. (Image: medRxiv)

There are limitations to this model. It assumes there is not a lower risk of infection among younger people. There is uncertainty on how many cases without symptoms there are. This figure depends on the spread of an infection on the Diamond Princess cruise ship.

These estimates are for Hubei, between 1st January and 11th February 2020. Different modelling assumptions can generate distinct estimates.

Competing biases

Estimating the fatality rate of a virus during an epidemic is challenging. The standard case fatality rate is: the number of deaths divided by the number of confirmed cases. We should treat this figure with caution in an ongoing epidemic. Two competing biases affect this proportion:

  • Undetected cases: mild cases may go undetected. Those cases are missing from the confirmed total.
  • Future deaths: it takes time for patients to recover or die due to the virus. Future deaths are not included in a particular snapshot.

Countries may have different case fatality rates. Coverage of virus cases may differ between and within countries. Other factors include: delays, demography, testing programmes, and healthcare capacity. Statistical modelling can account for issues with the standard calculation.

This blog looks at the use of statistics in Britain and beyond. It is written by RSS Statistical Ambassador and Chartered Statistician @anthonybmasters.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store