A claim has been brought to my attention that covid has cut life expectancy in England and Wales by a year. I think this is EXTRAORDINARILY misleading, so it deserves some clarification. •1/9 theguardian.com/world/2020/dec…
The thing is, life expectancy (at birth, or at any other age) is a bizarre notion. It is obtained by taking observed death rates by age at a given time (generally averaged over several years), assuming these rates stay constant, and computing expectancy from that. •2/9
So it's a concept which basically embeds the assumption that death rates hold constant over time. Which isn't true, of course. When they vary slowly (e.g., due to medical progress), it still makes sense. But in the face of an extraordinary even like covid? •3/9
Now what this preprint (medrxiv.org/content/10.110…) just does, if I understand it correctly, is take the mortality profile of 2020 and use it to compute a life expectancy (and compare it to a similar one without covid). •4/9 
In other words, what they are doing (by just using the 2020 death profile) amounts to ASSUMING THAT COVID-19 WILL CAUSE THE SAME DEATH PROFILE EVERY YEAR FROM NOW ON and computing the reduction in life expectancy from that assumption. •5/9
Now I do not believe that it any reasonable person would naturally interpret “covid-19 has cut life expectancy by one year” by understanding “if it is extended into perpetuity”. You tend to think of it as a one-off thing, as covid is believed to be. •6/9
So you might ask: then what IS the life expectancy loss for covid if it's just one-off? But the thing is, it's hard to define what the life expectancy loss of a one-off event means: “life expectancy” only has meaning in a stationary situation (see above). •7/9
What we CAN compute, instead, is the number of expected life years lost due to covid, and divide by the population. I didn't do the computation for the UK, but in France, the average life expectancy based on age&gender of covid deaths is about 11.1yr. •8/9
So if we multiply by the number of deaths and divide by the French population, we get… 3.4 days of life lost per person on average (so far!). I think this is far more meaningful than this “one year lost” computation which implicitly assumes covid will repeat every year. •9/9
• • •
Missing some Tweet in this thread? You can try to
force a refresh
