Hacker News Reader

Comment by nullc | Hacker News Reader

I'm confused about the Coronavirus mortality figures. The published numbers appear to be computed as deaths/cases. But cases is growing exponentially and deaths take time.

If you use deaths/(deaths+recoveries) you get a much higher figure, but presumably recoveries take longer than deaths.

View on HN

Another confounding factor is that you're more likely to identify severe cases and mild cases are less likely to get tested and confirmed.

This absolutely isn't accurate. The website promoting this garbage is dangerous. Nobody knows the total infected, nor the real fatality rate. But pay attention to how many on the cruise liners have died. 14% is a joke. My completely off the cuff, uneducated guess would be closer to 3%.

3% fatality rate with appropriate (if not the best available) medical attention and resources

I would expect a dismissal of a website as inaccurate garbage to be supported by a better counter-argument than a "completely off the cuff, uneducated guess."

What website are you referring to?

My comment was driven by my own experience with process estimation, though other commenters have pointed out that it's also supported by research on SARS: https://academic.oup.com/aje/article/162/5/479/82647

> If you use deaths/(deaths+recoveries) you get a much higher figure

You also have to take into account that mild cases are probably never detected and counted. So you are calculating the fraction of deaths among severe cases, not among all cases.

A blog article discussing exactly this. https://blog.zorinaq.com/case-fatality-ratio-ncov/

People have tried proper mathematical modelling but the data is fuzzy and it depends on how good the treatment is. Estimates seem to be about 2% of those getting it will die. I'm cautiously optimistic that chloroquine may drop the fatalities a fair bit. (http://www.china.org.cn/china/2020-02/22/content_75732846.ht...)

2-3% is the number you get from taking deaths divided by cases. This appears to be an obvious erroneous estimation technique.

The a parallel response links an article pointing to a paper evaluating this very question for the SARS outbreak. The paper shows that-- as I suggested-- dividing death by cases gives an extremely biased (low) estimate until the rate of new infections drops off. For SARS, deaths/(deaths+recoveries) was an accurate estimation.

:-/

Reply on news.ycombinator.com