One thing that sometimes get lost in COVID stats: incidence is a function of confirmed cases per unit population. But it's related to testing rate, & population tested. It's not directly reflective of true prevalence - it also means you can't naïvely compare previous rates (1/n)

..for example, we're doing 2x tests now we were in May, so "highest daily total since May" headlines actually require some unpacking: it's likely prevalence was much higher then, we just weren't detecting it, nor asymptomatic cases - wrong to conclude we're "worse" now (2/n)

..and of course, deaths maybe a better proxy to true prevalence than detected cases. But even that must be with caveat that maybe different subpopulations getting infected, so direct comparison should be avoided. Not that we should be complacent at all, just wary (3/n)

On tangent, R0 is an average measure & varies hugely between subpopulations with different dynamics. Not always a useful metric, prone to be misunderstood. To conclude, testing good, but caution & nuance required if extrapolating actual prevalence etc (4/4)

Just to follow up on this, @bbchealth today have an article on why apparent rises in cases don't necessarily mean increased prevalence. How and who we test makes a huge difference https://www.bbc.co.uk/news/health-54064347