Corona virus

Here are the fits from data of the John Hoskins CSSE site.

See here my current fits from WHO-data.

Data show a jump to +15,000 cases in one day because CT scans were used as well to find infected people. Therefore data show a bit inconsistency and the thurn of the now 7 data points from this is stronger than expected. We need to wait a few more days to follow that trend. If it continues, we might overestimate the numbers.

Today 20th February, I have modified the fit routines, in order to fit the new trend of recovered people. Before I had set their Nmax number to the average recovered that are shown in a graph below. This was necessary because the Gompertz function cannot fit Nmax well, when it is too far from the turning point, resulting in unreasonable numbers. The same applied for the number of deaths. As a solution, I have used now the second function in the graph to limit Nmax to the maximal number of infected people and fit the fraction of it as sin(m). As the deaths are near or at the turning point, I substract their maximum number from the maximum number of infected as a theoretical maximum number of recovered, of that I fit the fraction.

As a result, I can first time estimate the death rate from deaths-Nmax and infected-Nmax. Interestingly this matches nearly the same value I guessed from my approach by accounting the delay between diagnosis of infection and death, see link or last section.

Please be aware that this number may not be the final number, which we can only count, when the epidemic has ended, because during the epidemic news methods of treatment will be develop and help more people to survive. I hope to expect this number to be smaller.

Three data sections are availble: World, China only, and World without China.

World
China only
World without China
Death rate as function of average delay between infection diagnosis and death
Compare with Ebola data 2014/2015


If you like to support this work:

World


China only


World without China

We can read from the data outside China, that they are growing on an expected exconential curve. No sign of any control, otherwise we would see a turing poing approaching and the curve to tip over to the right. The T2 appears here much bigger, as we are far way from the turing point, so Nmax has a huge error. Fitted with the logistic function, we see a doubling time of about 6 days, like in the data in China. This is extremely fast and leaves little time for reaction. If we do not act now and start intensive control on airports, this curve will continue and the prediction for at least the next 5-10 days is quite certain. In 10 days we will see around 4000 cases outside China. Don't talk, act now!


Death rate as function of average delay between diagnosis and death

The currently known death rates are calculated on the day when a death is registered and by the number of infections on that day. However, the time between being diagnosed for the corona virus Covid-19 and the death may be several days. In those days the number of infected people rise exponentially. So the death should be counted on the day of being diagnosed. Therefore, if we later know the average time delay of being diagnosed to death, we can calculate from this graphs the true death rate by using the formular in the graph! Asume the average person feels sicks, goes to hospital, gets diagnosed with covid-19, enters then rather quickly a severe condition and dies after one week. So, if the average time delay will be one week, we can read from the graph for China death rate of about 4.5%!

Method 1: Using the last value of death/recovered rate.
Method 2: Using the average death/recoverd rate.

Method 2 is less sensitive to changes in daily death rate numbers.


.ยท. Impressum