China’s Fishy Numbers Drove Bad Modeling

By April 24, 2020 Commentary

China’s behavior in regard to the coronavirus epidemic has been atrocious, and no one should be making excuses for the country.  They failed to give early warning about the disease, which might have limited its spread.  And the number of cases the country reported and the descriptions of the course of those cases were inaccurate.  Early modelers relied upon those numbers to create incredibly exaggerated death rates.  What China did is inexcusable but the modelers should have known better as well.  But they couldn’t resist the headlines coming from spectacular claims of mass deaths.  We are finally beginning to get a little more reality on the Chinese experience with the virus.  A new study reported in Lancet finds far wider prevalence than was originally reported.   (Lancet Study)   The crux of the study revolves around the changing definitions of what a coronavirus disease case was, as issued by the Chinese Health Commission.  This definition changed seven times from January 15th to March 3rd.  If the last version had been used throughout the epidemic, the researchers found that on February 20th, 232,000 cases would have been reported instead of 55,500 (China currently says it had 88,000 cases).

Note also that this does not take into account persons who were infected but were asymptomatic or had mild illness.  Given that the virus ran unchecked for several weeks through a major metropolitan area that is densely populated, and given what we are seeing in terms of detection rates in the rest of the world, it is not unreasonable to suspect that there actually have been 5 to 10 times even the number of cases that these researchers estimate.  That would be 1 million to 2 million cases, which still seems impossibly low for a country of China’s size.  But consider the case fatality rate estimates with that number of cases, instead of what the Chinese government has reported. China claimed 4632 deaths (which also seems impossibly low, but that is close to what researchers were using) so on one million cases that is a .46 case fatality rate, and on two million it is .23.  A lot different than the Imperial College model that panicked policymakers and still higher than the final calculation will be.

Leave a Reply