Read on Twitter
This article raises some interesting points. Have death rates from COVID-19 gone done since March? https://www.npr.org/sections/health-shots/2020/10/20/925441975/studies-point-to-big-drop-in-covid-19-death-rates">https://www.npr.org/sections/...
Now, I think it& #39;s reasonable to assume that the proportion of people who die after being infected by COVID-19 will fall over time. It& #39;s probably true that being infected today is less deadly than earlier this year
That being said, I& #39;m quite skeptical of the evidence presented. This story seems to claim that death rates have dropped ENORMOUSLY, while I would expect something more in the range of a relative 10-20% drop (say from 25% to 20%)
The study that& #39;s linked to to verify the statement above is this one, and it& #39;s an interesting piece of research looking at some hospitals in NYC and the people they admitted for COVID-19 over time https://www.medrxiv.org/content/10.1101/2020.08.11.20172775v1.full.pdf">https://www.medrxiv.org/content/1...
Looking at the results table, something immediately springs out. The denominator here - the number of patients being admitted by week - has changed DRAMATICALLY over time
Perhaps more worryingly, that precipitous drop from 25% death rate to 7.6% cited in the news article appears to be based on...a single week of data. If you move back two weeks, there& #39;s no drop in death rates at all!
There also isn& #39;t really any trend here - a nonsignificant drop from 23% to 21% over two months, and then a sudden halving of the death rate in just one week, which corresponds to a sudden decrease in the average age
Moving on, we have the second paper cited in the NPR article, which looks at a very large sample of people who were admitted to ICU in English hospitals for COVID-19 https://www.medrxiv.org/content/10.1101/2020.07.30.20165134v2">https://www.medrxiv.org/content/1...
This study found that if you take the average death rate of these patients from the & #39;peak& #39;, it decreased in a linear fashion each week, and thus COVID-19 is getting less lethal
But why did they choose the & #39;peak& #39; week? It wasn& #39;t the first week of data collection, and it& #39;s a bit arbitrary to only start counting from the worst week in the whole dataset
Indeed, if we extend the analysis back even just 3 weeks, the relationship disappears almost entirely and suddenly there& #39;s been very little difference since March!
I would go so far as to say that this study implies that any differences in ICU mortality in England due to COVID-19 are ~almost certainly~ down to admission criteria (which were strained during the peak in early April)
