Going against Nature

I have on my desk a paper from Nature , the most influential and highly ranked multidisciplinary science journal. This paper addresses the reduction in life expectancy due to COVID-19, a very important topic. It makes the following claim: The average years of life lost per death is 16 years.

I have addressed this problem on this blog and in "COVID-19: Observations from a world upside down" in March 2020. My conclusion was that the loss of life expectancy for each Corona death was about 2 years. In later discussions, I rounded this number to one, as it's easier to manipulate.

So, how can Nature claim 16?

Here I am revising the argument I put forward in March last year with some updated numbers. Before I looked at the average age at death of Italian Coronavirus victims and found it to be 81. The life expectancy in Italy is 83

Thus, my natural conclusion is that the people who died of COVID-19 lost, on average 2 years. Clinically, they don't seem to be too different from the cohort of people dying from other causes -- old, with preexisting conditions, etc.

The thesis that the life expectancy of COVID-19 victims is very short and, thus, the YLL (Years of Life Lost) is of order 1 and close to my estimate of 2 years is supported by the fact that the excess mortality drops below zero (mortality drops below average) in the middle of the Corona pandemic.

For example, in April 2021, Belgian mortality from all causes dropped to 20% below average, despite the Corona virus killing some people (few, as, in my view, the Belgian population was thoroughly infected before and had good herd immunity by April). Still, cases of long Covid and the few new cases of Covid should have increase mortality to something above average.

If COVID kills people with long life expectancy (16 years, according to the Nature paper), these deaths should affect the mortality in April. If, however, COVID kills people with short life expectancy, many of the COVID victims from the winter would have naturally died in April, and, as they are dead already, they don't die again and don't show up in the statistic.

To their credit, the authors do acknowledge that their YLL of 16 years may be an overestimate, due to Covid selectively killing people who are already sick and have low life expectancy. I very much believe it is indeed -- and very much do.

'However, our key results are not the total YLL but YLL ratios and YLL distributions which are relatively robust to the co-morbidity bias.'

These alternative claim in that, in the most affected countries, COVID-19 kills between 2 and 9 times more than the average flu. In Figure 1 in the paper, we see that, even if we restrict ourselves to the last relatively mild flu years, there is considerable variation in the severity of flu in different seasons. In severely affected countries, like Italy, US, Spain and Belgium, the most severe recent flu comes fairly close to Covid.

This is more consistent with my claim that COVID-19 is part of the natural dying process in humans. It is like a bad flu -- perhaps comparable to 1958 or 1967, and worse than the typical flu from the past few years. The impact on life expectancy on most victims is low, with most of the life lost being end-of-life care often in specialized institution.

An Economist article stated that in about half of the US states, over half of COVID deaths occurred in care homes and other kinds of end-of-life institutions. In April, the NY times counts the deaths from nursing homes to be about a third of the total COVID-19 deaths.

I thank Anna Bojds for useful discussion.