| I hear a lot about how we may be seriously underestimating the CFR denominator, due to undiagnosed mild cases. Ok I get that. What I don’t hear much about is that the numerator is also probably understated at any point in time for several reasons. First I don’t trust the China data I think it’s understated and that’s the oldest, most mature data we have. Second, in an exponentially growing disease with something like a 6 week course from infection through to mortality/recovery, we will always have diagnosed cases that are 6 weeks, or whatever the true course is, ahead of the final death tally and as the diagnosed cases are rising so rapidly, including those weeks and weeks worth of diagnosed but unresolved cases could add up to a huge amount of error to a simple deaths/cases CFR analysis that would significantly understate what the final CFR will look like. Cohort analysis does not appear to have caught on in the CFR calculation world from what I’ve been reading. Third, I’ve read that there are likely a significant number deaths that are probably attributable to COVID-19 that are not being counted as Covid-19. The death rate in northern Italy over the last month, even when all the COVID-19 attributed deaths are removed, is significantly higher than it has been in similar periods in the past I have read. I believe the same is true in New York City. So if these stories are correct, there are likely more COVID-19 deaths than are being counted. So while it is probably true that the denominator is understated, it seems to me that it’s also very likely true that the numerator is also understated making it very difficult for me to believe any of these estimates are very accurate until both these issues are addressed. Has anyone seen anything that explains some of these issues and calculates a cohort-based death rate, which somehow estimates or adjusts were incorrect, time shifted or under counted Fatality data? |