Looking purely at that paper’s methodology - by meta-analysing each hospital’s odds ratio for a time effect, they all end up equally weighted. Surely higher volume centres should count for more. They also report a significant date of admission x mortality effect for Nottingham, but looking at the numbers they had a pretty consistent 25-30% mortality when there were 100-200 admissions, and then it jumps all over the place when the numbers are smaller.
It’s possible that temperature has an effect, but I’d be interested to see some analysis of air quality in these cities over the course of the pandemic as lockdowns take effect, and some kind of measure of “overwhelmedness” eg analysis of total daily hospital admissions vs normal number of beds & covid mortality.
I don’t think it’s totally implausible that treatment did massively improve. Difficult to overstate how quickly the understanding of the disease changed - even from March to April the way we were treating the patients changed significantly.