One of the great unknowns of the Covid-19 crisis is just how deadly the disease is. Much of the panic dates from the moment, in early March, when the World Health Organisation (WHO) published a mortality rate of 3.2 per cent – which turned out to be a crude ‘case fatality rate’ dividing the number of deaths by the number of recorded cases, ignoring the large number of cases which are asymptomatic or otherwise go unrecorded.
The Imperial College modelling, which has been so influential on the government, assumed an infection fatality rate (IFR) of 0.9 per cent. This was used to compute the infamous prediction that 250,000 Britons would die unless the government abandoned its mitigation strategy and adopted instead a policy of suppressing the virus through lockdown. Imperial later revised its estimate of the IFR down to 0.66 per cent – although the 16 March paper which predicted 250,000 deaths was not updated.
In the past few weeks, a slew of serological studies estimating the prevalence of infection in the general population has become available. This has allowed professor John Ioannidis of Stanford university to work out the IFR in 12 different locations.
They range between 0.02 per cent and 0.5 per cent – although Ioannidis has corrected those raw figures to take account of demographic balance and come up with estimates between 0.02 per cent and 0.4 per cent. The lowest estimates came from Kobe, Japan, found to have an IFR of 0.02 per cent and Oise in northern France, with an IFR of 0.04 per cent. The highest were in Geneva (a raw figure of 0.5 per cent) and Gangelt in Germany (0.28 per cent).
The usual caveats apply: most studies to detect the prevalence of the SARS-CoV-2 virus in the general population remain unpublished, and have not yet been peer-reviewed. Some are likely to be unrepresentative of the general population. The Oise study, in particular, was based on pupils, teachers and parents in a single high school which was known to be a hotspot on Covid-19 infection. At the other end of the table, Geneva has a relatively high age profile, which is likely to skew its death rate upwards.
But it is noticeable how all these estimates for IFR are markedly lower than the figures thrown about a couple of months ago, when it was widely asserted that Covid-19 was a whole magnitude worse than flu. Seasonal influenza is often quoted as having an IFR of 0.1 to 0.2 per cent. The Stanford study suggests that Covid-19 might not, after all, be more deadly than flu – although, as Ioannidis notes, the profile is very different: seasonal flu has a higher IFR in developing countries, where vaccination is rare, while Covid-19 has a higher death rate in the developed world, thanks in part of more elderly populations.
The Stanford study, however, does not include the largest antibody study to date: that involving a randomised sample of 70,000 Spanish residents, whose preliminary results were published by the Carlos III Institute of Health two weeks ago. That suggested that five per cent of the Spanish population had been infected with the virus. With 27,000 deaths in the country, that would convert to an IFR of 1.1 per cent.