I& #39;d like to clarify, in this thread Lewis talks about % increases and decreases in different causes of death at home vs in hospital, and suggests that the % rise in home deaths being larger than % drop in hospital deaths means more deaths of that cause overall. Not necessarily. https://twitter.com/lewis_goodall/status/1318180176480653312">https://twitter.com/lewis_goo...

We can see that the *total* deaths in private homes versus hospitals have changed by almost identical amounts, suggesting across all causes a 1 for 1 swap. But if we imagine a disease that is largely hospitalised at end of life, say, 80-20 in favour of hospitalisation, then
>

if the were was a 25% drop in hospitalised deaths, and a 50% gain in home deaths, that actually doesn& #39;t mean more deaths overall. 80 becomes 60, but 20 only becomes 30 - the total is now 90 instead of 100. Less deaths total, not more!
>

Now, I& #39;m not saying this is definitely happening, but it would surprise me to learn that each individual cause of death splits 50/50 between home and hospital. Percentage differences here are less important than the raw numbers, like we have for total deaths.
/end

ADDENDUM: At some point during me writing this thread, Lewis deleted the tweet where he suggested higher overall heart disease deaths based on their percentages (I pointed this out there too). That& #39;s good, actually; far better than leaving potentially misleading suggestions up!

I don& #39;t think he meant anything nefarious by it - it& #39;s really easy to just think & #39;this number went up more than this went down, must be more overall& #39;, but you just can& #39;t do that with percentages alone, you need the raw numbers.