I was mocked at the time, but new data supports my thinking
For suggesting that the average IFR of Covid-19 might be between 0.01% and 0.05% during an early 2020 interview with UnHerd, I was mocked by large sections of the media and accused of minimising the harms of Covid-19 in service of a Right-wing libertarian agenda. My analysis was called “spurious” by the Sunday Times, and was quickly dismissed by senior MPs and the establishment press.
Data released this week by Denmark’s health registry however confirms, once again, that the age gradient in mortality from SARS-CoV-2 infection essentially resembles a cliff-face, accelerating sharply only in the ninth decade of life. The numbers come from an exercise undertaken to sift deaths from Covid-19 out of the deaths with Covid-19, which reveals an overcount of around 100% in the Danish data. This puts the average mortality risk over the last three years for those above 20 at 0.089%. Adding the under-20s reduces it further to 0.065%.
Like what you’re reading? Get the free UnHerd daily email
Already registered? Sign in
What can we usefully infer from this data about the lethality of SARS-CoV-2?
The first point to make is that it is unhelpful to talk about an ‘average IFR’ — and this is something I should certainly have avoided doing three years ago — as this is entirely sensitive to the age structure. Commonly, the risk to the older age groups is downplayed while the risk to the majority of the population is exaggerated. Policy decisions made on the basis of an average IFR are likely to be sub-optimal, and model projections based on an average IFR are bound to carry flaws.
It is worth noting in this context that the all-important supposed IFR of 0.9% in the Imperial College model of March 2020 was anchored to data — such as from the Diamond Princess cruise ship — that was skewed towards the high-risk groups. This was the main problem with their projection, rather than any errors in the methodology.
Some of us pointed out at the time that the UK data on deaths was equally compatible with a very low IFR, had the first wave of the epidemic occurred earlier than supposed by the Imperial model. The 0.065% mortality risk from Covid-19 across the past three years in Denmark is certainly closer to 0.05% than 0.9%. In fact, summing over several years can result in an overestimate of IFR, since the numerator (the number that die from the infection) will keep growing, while the denominator (the population size) remains static.
If you count for long enough, the IFR can exceed 100% — which is clearly nonsensical. That said, three years is probably a reasonable period over which to measure the overall IFR, as a significant proportion of people are likely to have been infected over that period, and so the total population can be safely used as a denominator.
However, we also need to bear in mind that vaccines have almost certainly reduced the total number of deaths. If, for example, vaccines reduced the expected number of deaths by a half, then the IFR is actually double what has been measured here. But to obtain an average IFR of 0.9% from this data, we would require over 75% of those aged over 90 to have died from Covid-19 in the absence of vaccination and, even so, the IFR for those under 60 would remain at 0.05%.
The main lesson to be learnt here is not to quibble about the true value of the IFR but, instead, to construct a strategy that is robust to these uncertainties. Focused protection of the vulnerable population, followed by targeted vaccination, remains the only reasonable solution given the very low IFR in the majority of the population. That’s not to mention the enormous costs of lockdown.