From the Abstract — study preprint avalable here:
…seroprevalence study using self-administered lateral flow immunoassay (LFIA) test for IgG among a random population sample of 100,000 adults over 18 years in England, 20 June to 13July 2020… We estimate that 3.36 million (3.21, 3.51) people have been infected with SARS-CoV-2 in England to end June 2020, with an overall infection fatality ratio of 0.90% (0.86, 0.94).
The UK’s population of 66.65 million, so around 5 percent of the populace has been infected. On 1 July, the midpoint of data collection, 284K cases had been diagnosed, so the actual infection rate was 12 times as high as the case count.
Assuming a 20-day lag from infection to death, there were 41K covid-related deaths in the UK among the 3.36 million total infected. 41/3360 = 1.2 percent fatality rate: that’s quite a bit higher than the study’s estimate. The study didn’t sample from people living in residential care facilities, so nursing home deaths were excluded from the analyses. Nursing homes have accounted for around half of covid deaths in the UK, so that’s a sizable undercount.
I’ve been estimating a US infection fatality rate of 0.6 percent. The UK’s median age is 40 compared to the US’s 38, so I’d estimate a UK rate of 0.6 x 1.12 = 0.7 percent. That’s not far under the study’s estimate; however, the study greatly underestimated deaths by excluding nursing home data. The findings from this study support the tentative interpretation of the Italian serology study: my estimated age-adjusted mortality rate algorithm might need to be adjusted upward.