SA, some questions: what is your source for the figures, and what is the formula for the O rate?
Perceiving trends in this pandemic has so many problems of variability of accuracy of data; some counties have tested a much higher proportion of the population than others, and most countries’ rate of testing has presumably increased; testing is targetted in various ways, very little of it is random and thus a representative sample of the population; attribution of cause of death probably varies a lot; we don’t have any metrics for either imposition of social restrictions or actual public behaviour in social distancing.
Maybe proxy data would serve us better. For instance, when considering diverse countries, the overall excess death rate is probably more reliable than the covid-19 death rate.
Population density is an interesting one because the overall population density of a country tells us hardly anything about, for instance, how often people inhale each other’s breath. England has a population density of 430 per square kilometre, but obviously the potential for cross-infection varies hugely between 430 people out looking for mushrooms compared with 430 people attending a meeting in the village hall.
What would serve as a good proxy for social proximity distribution? And what about social mixing, ie. unusual meetings between people as opposed to repeated ones? I suppose transport use might help – a distribution curve of number of journeys and their distance.
It’s frustrating. The things to be considered are so mundane, yet quantifying them is so elusive. I can certainly see the temptation to modelling – just write a program, let the machine crunch the numbers, and ponder upon the output at leisure. But I expect that hammering out good parameters and directly measurable proxies for them would provide a better route to being able to think about the problem directly.