The Corona Conundrum: How deadly is it – really?

There is a titanic struggle between two forces at the moment – concern over the spread of the COVID-19 pandemic and concern for the economic effects of the continued global lock-down. Both sides are basing their arguments on the same basic statistics, but there is really no way to accurately analyze Case Fatality Rates without testing large random samples of people in an antibody study. .

The CFR is the percentage of cases that ultimately prove deadly, and it has driven policy far more than any other statistic, because our models could be far more accurate if we could properly project this pandemic. The anger over our continued lock-down is fueled by the notion that the CFR will greatly decline once asymptomatic people are properly measured, and two recent studies have hinted that this is the case; one from Santa Clara County, California and one from England (since i first started typing this, more have been released). In Santa Clara, they tested 3,000 residents for COVID-19 antibodies and found far more than would explain the 1,000 reported cases; home to 2 million people, Santa Clara includes San Jose, and based on their results they estimated that between 48-81K people had actually been infected and mostly been asymptomatic.

That is encouraging news, but the implied CFR would be in the neighborhood of an average flu strain, close to the .0193% of H1N1 during the 2009 pandemic (in which 57 million Americans and 1 billion people worldwide contracted that flu strain). That’s a stark contrast with the 5.42% current rate for the US and 12% in the UK, and the difference can be partially explained by the invisible cases described above.

But there’s still a piece of evidence to the contrary that’s being ignored, and I decided to go through the CDC’s historic influenza records to dig deeper. The Santa Clara study led to an estimate of approximately 1.5-5% rate of current infection, and so I decided to calculate the impacts of infection rates at 2.8%, 5%, 10.8% and 17.43% (pulled from the Santa Clara study, my butt, a 35 million guess and the infection rate for H1N1, respectively). My personal, unsophisticated guesstimate would be a CFR of .7-.9%, with a low around .4% and a high around 1.5%.

Why am I skeptical? Because the number of fatalities from confirmed cases is 39,090 – only 9,000 less than the most deadly flu season in the CDC’s memory (since 1918, of course), in 2004. The A(H3N2) B strain has consistently produced strong flu seasons in the last 20 years, but the timescale is very different; the average 23,000 death-season picks up in October and declines in March, a six-month window. When I landed on 3/7 after traveling back from Prague through Madrid, there were 20 confirmed deaths in the United States. In all likelihood, COVID-19 will pass 2003-04’s flu record around a week from today, 50 days from 3/7.

It’s possible that COVID-19 spreads at record speed, in the middle of the most drastic lock-down imposed in modern history (and by modern I mean a few centuries), but to yield 40,000 deaths in 43 days would mean that either a) somewhere between 120-200 million Americans have already been infected (per swine flu mortality), or b) COVID-19 is deadly at a historic rate.

You can probably guess my pick. I didn’t just guess, though – i took the state numbers of cases and calculated the correlations between 18 variables, including population density (after I finish cleaning up the density for every city above 50,000 people, in the US (about 750) I’ll exchange urban density for each state instead).

Total Cases (by state standard deviations from the US mean) against typical demographics

There are far too many partisan assumptions that could be drawn incorrectly from this data, but 65+ state proportions of pop. are far weaker as a correlation than income, foreign birth, or total population – Texas, with two of the five largest metropolitan areas in the US, goes against each of these correlations and Florida isn’t far behind (data in my more expansive chart to be published later today). I’m eager to return to work, but whether I go solo or join some consulting firm, I’ve prepared myself for the rise in deaths and infections that could follow.

It’s far too late/early to dig deeper at the moment, but I want to emphasize that the only interest I have is in providing unbiased personal observations based on factual evidence.

– Rixey

On the estimation of Case Fatality Rates for COVID-19 (3/22):

Clinical-Course-and-outcomes-of-critically-ill-patients-with-SARS-CoV-2-in-Wuhan-China-1

Their overall estimate for SARS-nCoV-2’s case fatality rate (CFR) in China was 2.4% (crude), 3.0% (among all infected) and 3.6% among symptomatic patients.

Their overall estimate for SARS-nCoV-2’s case fatality rate (CFR) in Italy was 1.3% (crude) and 3.3% (overall). Additional estimates were provided for northern Italy specifically.

The rapid increase in reported deaths from Italy after 3 March lends weight to our estimates, but indicates that the number of deaths will continue to increase for some weeks, despite strict social distancing measures.

Additionally, dichotomization into asymptomatic and symptomatic is a simplification; SARS-CoV-2 causes a spectrum of symptoms, likely depending on age, sex and comorbidities. Serological surveys will be needed to better characterize asymptomatic infections [25]

Author: Prometheus