[ad_1]
On December 12, the Robert Koch Institute reported 241 new infections per 100,000 inhabitants in seven days for the Elbe-Elster district in Brandenburg. For the Sonneberg district in Thuringia, the RKI gave an incidence of 182, for the Friedrichshain-Kreuzberg district of Berlin as 162. With a slight margin, all the figures turned out to be clearly too low: in Elbe-Elster the incidence was in reality 473 at Sonneberg at 404, at Friedrichshain-Kreuzberg at 281. The RKI numbers were far off.
The examples mentioned may be extreme, but they pose a bigger problem: the Robert Koch Institute only includes some of the infection cases that are reported to it from the districts within a week in calculating the 7-day incidence. Therefore, the incidence values given by the RKI are consistently too low. They almost always have to be corrected upward. Many federal states also have the flawed RKI method.
If the incidence is underestimated, it is not a theoretical problem, but it has real consequences: the values serve as the basis for decision-making on measures to combat the pandemic. If the official incidence value is too low, the rules may not apply, even though the infection process actually requires it. Meanwhile, the Bundestag has even codified this incidence control function in the Infection Protection Act. Paragraph 28a defines several thresholds. The values published by the RKI are decisive.
If the official figures are not up to date, the rules may not even go into effect
the Süddeutsche Zeitung documented and evaluated RKI incidence data for all rural and urban districts over a two-week period. More than 5700 data points are included in the analysis. On average, the incidence of RKI is 17 points too low. In less than a fifth of cases, the official incidence turns out to be too high. Statistician Helmut Küchenhoff from the University of Munich and mathematician Moritz Kaßmann from Bielefeld University advised on the evaluation.
“With the RKI method, some of the cases are neglected. That is problematic,” says Küchenhoff. The RKI figures are the result of national data collection. The health authorities of the 401 counties and individual urban districts record the report of a positive corona test. From there, the data is sent to the state health authorities. They summarize the reports and report their values to the RKI. However, not all of these cases are included in the incidence calculation.
Because RKI assigns cases to the date they were reported to local district authorities. In some of the cases, this date is so old that the case falls outside the seven-day window. Then, that infection is assigned to the previous week, so its incidence increases in retrospect. But no one takes note of this corrected value. No school can be closed retrospectively. There is no retroactive curfew. “The RKI method is suitable for mapping the course of the pandemic as accurately as possible,” says mathematician Kaßmann, “but it is not suitable as a daily basis for making decisions.” But it is precisely this use that the law provides.
Nobody takes note of the subsequently corrected values
The Robert Koch Institute is aware of the problem. He too Spiegel has already reported on it. “The delay in transmission can lead to an underestimation of the incidence of 7 days, especially in the case of dynamic developments, since not all data is available yet,” said the authority when asked by SZ. “Determinants for the evaluation of the situation in situ” are the data of the local authorities, since they are updated. A contradiction with the Infection Protection Act, which even specifies the Internet address of the RKI dashboard.
All reports on the current situation in Germany and around the world, as well as the most important news of the day, twice a day with SZ Espresso. Our Newsletter Updates you morning and afternoon Free registration: sz.de/espresso. In our News app (download here) you can also subscribe to espresso or breaking news as a push message.
The SZ discussed various methods with Küchenhoff and Kaßmann to calculate more valid incidence values. Neither can accurately represent actual value. But for all of them, the deviation is significantly less than that of the RKI method. One possibility is to give more time to the offices. For example, without taking into account all the reports until the day before, but only until the day before yesterday or the day before yesterday. The distance reduces the influence of late pending registrations. Küchenhoff’s team at the University of Munich uses this method.
A very simple solution has proven to be incredibly reliable: cases are not assigned to the date of notification in the local authority, but to the date of publication in the RKI. All infections that are reported to the RKI within one week are included in the incidence. Some of these cases actually belong to the previous week, but there are also no cases from the current week that have not yet been reported. Both of these effects largely cancel each other out. Therefore, the incidence values shown in the SZ maps and diagrams are calculated using this method.
