2 ms·
There is one major bias in the sort of studies you just linked to that can't really be smarted away. When looking at things like a mortality rate, you need to s
by somenameforme 5y ago
There is one major bias in the sort of studies you just linked to that can't really be smarted away. When looking at things like a mortality rate, you need to somehow determine how many people are infected. Imagine there were some weird disease where it was completely asymptomatic in 99% of cases and fatal in 1%. The vast majority of contemporary COVID related studies would claim this disease would have an extremely high mortality rate, far higher than 1%.
The reason is because diagnosis is almost never done randomly, but instead relies on different avenues like hospitalization data. So you tend to already be biasing yourself to severe outcomes because milder cases are not going to end up getting diagnosed. In omicron's case this effect has been extreme as things like sewage samples showed dramatically higher rates of of the disease than were being officially reported. In the study you mentioned it determined diagnosis using:
"CDC used data from three surveillance systems to assess U.S. disease related to COVID-19 during December 1, 2020–January 15, 2022. COVID-19 aggregate cases and deaths reported to CDC by state and territorial health departments¶ were tabulated by report date.* ED visits with COVID-19 diagnosis codes were obtained from the National Syndromic Surveillance Program (NSSP).†† Hospital admissions and inpatient and ICU bed use among patients with laboratory-confirmed COVID-19 were obtained from the Unified Hospital Data Surveillance System."
So your numbers are biased to heavily rely on things like hospitalization data, which is going to make the numbers borderline useless for trying to evaluate the overall mortality rate. To be fair I'm not really attacking the study either. Like I said this is a problem that really can't be solved in any way other than an involuntary lottery with mandatory testing + reporting + profiling/classification, which is something I'd expect to see in China, but not the US - for better and for worse.