Professor Dr John P. A. Ioannidis of Stanford University, one of the world’s leading epidemiologists , has always viewed scientific research with a critical distance . In epidemiology in general, and the current pandemic in particular, he considers the reliability of predictions to be rather problematic . Among the reasons he cites for this are the poor quality of the available data, incorrect assumptions in the modelling and the lack of numerical stability in the calculations . An example is cited where 4 different methods already fail in the task of predicting Covid 19 deaths even one day in advance for the individual American states.
In retrospect, the predictions concerning the influence of the British mutant B.1.1.7 of the SARS-CoV-2 virus on the incidence of infection also turned out to be rather inaccurate. Whereas in Germany, for example, Professor Drosten predicted “100,000 new corona infections per day” for spring and summer , Research Luxembourg did not let itself down: due to a 50% higher infection rate of the British variant, a peak in mid-May, comparable to that of the second wave, was to be expected (left graph). However, a much more dramatic course with more than 1200 new infections per day had also been considered (right graph) .
In the weeks and months that followed, the forecasts were diligently updated at regular intervals, reminiscent of the fate of a snowman during a thaw. The forecasts on 25 February, 4 March, 8 April and 12 May 2021 are shown below.
The maximum of the “3rd wave” was finally reached at the end of March with just under 250 new infections, i.e. 6 weeks earlier and almost a factor of 3 smaller than predicted.
An article on science.lu from March  tries to explain how such model calculations are to be interpreted:
However, in these models the scenarios are calculated in such a way that the behaviour of society does not change. But when the situation comes to a head, people usually change their behaviour – or politicians decide on new measures. So that medium- or long-term model calculations are fraught with great uncertainty. They indicate what would happen if things continued as they have been trending in recent days. In other words, they are not meant to be a predictive tool. Rather, they indicate the trend for the next few days and weeks if the trend of the last few days continues. They are a useful guide. But not fortune-telling. This is where misunderstandings often arise in the public debate.
We leave it to the reader to assess these remarks. Perhaps it will be granted to him to recognise which public debate is referred to in the quotation.
Moreover, such predictions are all the more astonishing, since it is generally known that the seasonal activity of coronaviruses always subsides in April .
Comparing the second and third wave, some peculiarities stand out. We look at the positive rates (positive PCR tests per total number of tests performed) of the tests performed in the different categories:
Large Scale Test including airport (“LST & airport”)
diagnostic tests in symptomatic individuals (“diagnostic”)
Testing of contacts of those tested positive at the end of quarantine (“tracing”)
The weighted average value determined from these categories (see ) is referred to as “global”.
These sizes and the proportion of the British mutant are shown in the following graph (sources [10,11], unfortunately no separate data for “tracing” and “diagnostic” are available for week 45/2020 to 4/2021):
The third wave does not have a pronounced maximum in terms of positive rates.
For the LST, the value of 0.5% is no longer exceeded as of January; in the second wave, the maximum was still just over 2%.
The positive rate of diagnostic tests fluctuates around 2% from February onwards, compared to 10% in November.
The global positive rate is similar: fluctuating around 2.5%, whereas in the autumn it rose to 6%.
Only in contact tracing do we see a clear increase: here the values are between 10 and 20%, with the last measured value in November reaching 10%.
A correlation between the time course of the proportion of the British mutant and the positive rates is not discernible.
The positive rates of 2021 subside just as they did in spring 2020, but somewhat later. The seasonality of the SARS-CoV-2 virus therefore seems to be in effect once again. Therefore, there is no causal need to justify this development by vaccination.
When looking at the composition of the number of cases, it also becomes clear that especially the tests in the context of contact tracing contribute a significant share (in some cases up to two thirds!). The maximum in week 12 also takes place at a time when the total number of tests is at its highest. This once again confirms the dependence of the number of cases on the total number of tests .
In summary, it can be stated here: a “real” wave actually looks different. We would expect the positive rate of diagnostic tests (“diagnostic”) to change substantially and reach a pronounced maximum, as was the case in the two previous waves.
It is also not very convincing that those who are “found” positive by contact tracing make up the majority of cases. After all, these people generally have no to mild symptoms, as they would otherwise have been listed in the statistics under the category of diagnostic tests.
Let us also look at hospital occupancy and mortality:
While the occupancy of non-intensive care beds follows the number of cases, intensive care bed occupancy reaches its maximum only in week 16.
The average number of intensive care bed occupancy and the number of deaths per week initially correlate highly during 6 months (R2=0.9), but drift apart from week 12 onwards for inexplicable reasons. After all, there is no obvious reason why significantly more patients should survive the disease after a certain point in time, so there is a need for explanation here. The situation is similar for the ratios of intensive care bed occupancy or deaths per number of cases; here the divergence is even more pronounced from week 12 onwards.
With the start of the vaccination campaign at the end of December, 170,000 vaccine doses had already been administered at the time of the peak of intensive care bed occupancy in the 16th week. The claim that the promise of preventing severe courses through vaccination has been fulfilled can certainly be doubted. After all, the number of intensive care patients at this point is comparable to that of autumn.
Finally, it should be mentioned almost in passing that according to recent studies [14,15], the virulence of the British variant hardly differs from other mutants. However, political decisions were made with regard to a potential danger (for example, the compulsory wearing of masks in schools), which could certainly be called into question in retrospect on the basis of the new data.
What happens next? The next mutant is already in the starting blocks: B.1.617.2 or also called “Delta”. It is already being treated as the trigger for the “fourth wave” . The LNS has already identified a 30.9% share of this mutant (week 22) . So it will possibly be a hot autumn.
Whatever the case, we are already eagerly awaiting the coming forecasts.