1) Antibody studies
The covid-19 infection fatality rate (IFR) depends on demographics (age and risk structure), public policies (e.g. protection of nursing homes), and medical treatment quality.
Covid-19 IFRs are strongly age-dependent, with a steep increase above the age of 70. The median age of covid-related deaths in most Western countries is 80 to 86 years (see section 3 below). In most Western countries, about half of all deaths occurred in nursing homes (see section 5).
In terms of covid-19 IFRs, an important difference exists between places with and without a partial or total collapse of local health and elderly care, and between the early and late pandemic phase.
A. Places without a collapse of health and elderly care
|Global||Oct. 7||51 locations
Below 70 years
|Japan||Sept. 23||Tokyo <70y||0.01||Hibino|
|Iceland||Sept. 1||General population
Below 70 years
|Austria||June 25||Ischgl hotspot||0.26||von Laer|
|Sweden||June 16||Stockholm NNH
|Slovenia||May 6||General population||0.16||GSI|
City of Munich
|Iran||May 1||Guilan province||0.12||Shakiba|
|Santa Clara County
Los Angeles County
|Denmark||April 28||Blood donors (<70y)||0.08||Erikstrup|
1) 0.64% and 0.60% including nursing homes; 2) 0.14% and 0.23% assuming 40% missing fatalities (more); 3) median values; 4) the unadjusted IFR is 0.28% (page 9); 5) general population (excl. nursing homes); 6) 0.58% including Stockholm nursing homes (about 40% of deaths, see page 23); 7) These US studies may underestimate the true IFR, as they were done early during a locally accelerating pandemic; 8) 0.76% including nursing homes (36% of deaths).
Note: The much-cited Meyerowitz-Katz meta-study claiming a global Covid-19 IFR of 0.68% is misleading because it mixes modelling studies and antibody studies, nursing homes and the general population, early and late phase IFRs, and commits several methodological mistakes.
B. Places with a partial or total collapse of health and elderly care
Places with a partial or total collapse of local health and eldery care experienced significantly higher and very strongly age-dependent IFR values, especially during the early phase of the pandemic.
However, IgG antibody tests may underestimate the true prevalence of coronavirus infections and may thus overestimate the IFR by a factor of two to five (see section 2 below).
|Spain||August 7||Covid confirmed
Below 50 years
Below 40 years
A Spanish seroprevalence study found an overall IFR between 0.82% (based on confirmed Covid-19 deaths) and 1.07% (based on excess all-cause deaths). The study didn’t include nursing homes, which accounted for about 50% of all deaths. The IFR was strongly age-dependent, with values below 0.03% until 40 and below 0.1% until 50 but reaching very high levels above 70 years.
The study found a country-wide IgG antibody seroprevalence of just 4.9% (about 12% in Madrid). However, less than 20% of symptomatic people (3+ symptoms or anosmia) had IgG antibodies. This may indicate that infections were up to five times more widespread than detected by IgG antibody tests (see section 2 below on this topic). If so, Spanish IFR values might drop below 0.5%.
Above 60 years, there was a significant difference in lethality between men and women. This might be due to e.g. genetic reasons, cardiovascular health, or certain habits like smoking.
2) Northern Italy
|Northern Italy||August 6||Above 70 years
Below 70 years
Below 50 years
80+, first phase
80+, second ph.
An Italian study considered contacts of confirmed Covid-19 cases in the Lombardy region, which includes hotspots like Bergamo and Cremona, to determine their fatality risk and their comorbidities. They found that the overall IFR was 62% lower in the second phase of the pandemic (after March 16) compared to the first, cataclysmic phase (up to March 15).
This was particularly evident in people above 80, where the IFR dropped from 30% in the early phase to 8% in the later phase (4% for women, 16% for men). Below 50 years, IFRs were near 0%; below 70 years, IFRs were 0.43% (both phases combined). More than 80% of deaths occurred in patients with cardiovascular diseases, which are known to be an important risk factor.
Of note, among Italian people with anosmia (temporary loss of the sense of smell or taste), a very typical Covid symptom, only about 25% were found to have IgG antibodies. This could indicate that coronavirus infections are more widespread, and IFRs lower, than assumed.
3) New York City
|New York City||June 29||Overall||0.70||Stadlbauer|
|New York City||June 29||Confirmed
25 to 44 y.
Until May 2020, New York City counted about 20,000 confirmed and probable Covid-19 deaths among its 8.4 million citizens and registered an antibody prevalence of about 20%. Studies estimating the infection fatality rate (IFR) for New York City found values between 0.7% and 1.1% based on confirmed deaths and up to 1.45% based on confirmed and probable deaths.
About 52% of Covid deaths in NYC occurred in the 75+ age group. This value is lower than in Europe, where about 90% of deaths were 70+. In all of New York State, about 6,300 patients were sent from hospitals into nursing homes, which ultimately registered between 6,600 and 13,000 deaths.
As in Italy and other hard-hit places, the IFR for age groups above 65 dropped by about 50% during the course of the pandemic, possibly due to better medical preparedness and treatment strategies.
Assuming that serological IgG antibody tests do not capture the full extent of coronavirus infections (e.g. due to mild cases without IgG antibodies), the overall IFR in New York City might drop to about 0.50% or below, and the actual spread of the coronavirus might be above 50%.
4) United Kingdom
|England||August 21||July 28 (MCT)
July 28 (ONS)
|England||August 14||General population
Incl. care homes
45 to 64 years
Below 44 years
Until July 2020, England counted about 30,000 Covid deaths in the general population and about 20,000 Covid-related deaths in nursing homes (which had to receive patients). According to the Oxford Centre for Evidence-Based Medicine, the Covid IFR fell by 50% to 80% during the epidemic and reached a value between 0.3% and 0.5% by the end of July.
A study by Imperial College London estimated an IgG antibody seroprevalence of 6% overall and 13% in London by mid-July. However, according to Public Health England, London blood donors had an antibody seroprevalence of 17.5% already in May.
Of note, only about 50% of people with anosmia (temporary loss of the sense of smell or taste), a very typical Covid symptom, had IgG antibodies. Only 35% of people who were suspected to be Covid cases by a doctor, had IgG antibodies. And only 28% of people who self-reported “severe symptoms” had detectable IgG antibodies against SARS-CoV-2.
If some of these people were indeed Covid cases (without detectable antibodies at the time of testing), the overall IFR value in the general population may drop to about 0.50% or below. The overall mortality of 2020 is comparable to the strong flu season of 1999/2000 (see below).
|Belgium||June 20||General population
Incl. care homes
45 to 64 years
Below 44 years
Belgium reported one of the highest covid death rates in Europe, in part because it always included confirmed and probable covid deaths. 50% of excess deaths in Belgium occurred in nursing homes. Of these, only about a third were confirmed by a PCR test. It is possible that some of the non-confirmed nursing home deaths were not due to covid, but due to the extreme circumstances.
Due to the high proportion of nursing home deaths, IFRs differ markedly between the general population and the nursing home population. The IFR for the general population is estimated between 0.30% to 0.62%, while the IFR for the nursing home population is estimated between 28% and 45%. For people aged 45 to 64, the IFR is 0.21, and for people aged 25 to 44, the IFR is 0.02%.
Even without age-adjustment, the peak monthly mortality due to Covid in April 2020 was equal to the 1989 seasonal flu wave and lower than the 1951, 1960 and 1970 flu waves (see chart below). The median age of covid deaths in Belgium was about 86 years. However, Belgian antibody seroprevalence until June was only about 7% at the national level and 12% in Brussels.
2) Immunological studies
Immunological research indicates that serological antibody studies, which measure antibodies in the blood (mostly IgG), may detect only about 20% to 80% of all coronavirus infections, depending on the sensitivity of the assay, the timing of the test, and the population tested.
This is because up to 80% of people develop only mild symptoms or no symptoms if infected, as they neutralize the coronavirus with their mucosal (IgA) or cellular (T-cells) immune system. These people may develop no measurable IgG antibodies or may show them only for a few weeks.
Most global Covid-19 hotspots peaked at about 20% IgG antibody prevalence (e.g. New York City, London, Stockholm, Madrid, Bergamo). Moreover, among people with anosmia (temporary loss of the sense of taste or smell) – a very typical Covid-19 symptom – only about 20% to 50% had detectable IgG antibodies, according to surveys in several countries (see below).
See also: Are we underestimating seroprevalence of SARS-CoV-2? (BMJ, 09/2020) and Coronavirus up to five times more common and less deadly than assumed (SZ, 06/2020).
1) Only 16% of likely infected HCW had IgG; 2) People with anosmia but without IgG antibodies.
3) Median age of Covid-19 deaths per country
Half of all deaths were below, half were above the median age.
4) Hospitalization rate
Initial estimates based on Chinese data assumed a very high 20% hospitalization rate, which led to the strategy of ‘flattening the curve’ to avoid overburdening hospitals. However, population-based antibody studies (see above) have since shown that actual hospitalization rates are close to 1%, which is within the range of hospitalization rates for influenza (1 to 2%).
The US CDC found that Covid-19 hospitalization rates for people aged 65 and over are “within ranges of influenza hospitalization rates”, with rates slightly higher for people aged 18 to 64 and “much lower” (compared to influenza) for people under 18.
5) Percentage of Covid-19 deaths in care homes
In many countries, deaths in care homes account for 30 to 60% of all additional deaths. In Canada and some US states, care homes account for up to 80% of all “Covid19-related” deaths. In Sweden, deaths in nursing homes plus nursing apartments account for 75% of all deaths.
Source: Mortality associated with COVID-19 outbreaks in care homes (LTC Covid, May 21, 2020)
Source: The Covid-19 Nursing Home Crisis by The Numbers (Freopp, June 19, 2020)
The following chart shows the very significant difference between IFRs for the entire population (including nursing homes, top row) and for the non-nursing home population (bottom row) at the example of Belgium. IFRs for nursing home patients (ca. 30%) are about 10 times higher than IFRs for people of the same age (75+) outside of nursing homes (ca. 3%).
When calculating and communicating IFRs, it is therefore important to distinguish between nursing homes (which require focused and humane protection) and the general population.
6) Overall mortality
In countries like the UK (lockdown) and Sweden (no lockdown), overall mortality since the beginning of the year is in the range of a strong influenza season; in the US, mortality is in the range of the 1957 and 1968 influenza pandemics; in countries like Germany, Austria and Switzerland, overall mortality is in the range of a mild influenza season (but antibody levels are still low).
A comparison between the number of coronavirus deaths predicted by the influential model of Imperial College London (no measures or moderate measures) and the actual number of deaths in Sweden indicates that the model significantly overestimated the impact of the epidemic:
7) Antibody seroprevalence per country
Percentage of people with measurable antibodies to the new coronavirus.