search

nevrome / covid19germany

R package - Load, visualise and analyse daily updated data on the COVID-19 outbreak in Germany
Other
46 stars 8 forks source link

Lack of lag of deaths #44

Closed stschiff closed 3 years ago

stschiff commented 3 years ago

I'm confused by the fact that the RKI data seems to have virtually no time lag between deaths and registered infections. As seen in the plots https://github.com/nevrome/covid19germany/README.md, the first peak of both the deaths and the infections was in calendar week 13/14, which is end of March / beginning of April. This seems to be at odds with for example the dashboard here: https://www.zeit.de/wissen/gesundheit/coronavirus-echtzeit-karte-deutschland-landkreise-infektionen-ausbreitung#diagramm-international, which suggests that deaths peaked mid-April... is it possible that the deaths in the RKI table somehow are dated to their registered infection, or something like that? They seem to be around 2 weeks too early...

stschiff commented 3 years ago

I first thought this might have to do with the difference between beginning of disease and "Meldedatum", but the shape of the death curve in https://www.zeit.de/wissen/gesundheit/coronavirus-echtzeit-karte-deutschland-landkreise-infektionen-ausbreitung#diagramm-international appears to be as periodic as expected from a "Meldedatum"... I don't think they do any kind of nowcasting.

I also tried to check at the official RKI dashboard, but they don't show a time-series of deaths...

I'm pretty sure the "NumberTestedIll" is correct, but the death-times are somehow shifted...

stschiff commented 3 years ago

I have investigated this further, and I think this phenomenon can be fully explained by the difference between the StartOfDiseaseDate and the Date variable. If I plot both the daily deaths and the daily infections summed along StartOfDiseaseDate, there is a clean lag of around 14 days between the peak of the newly fallen ill and those who died. For example, in the first wave, the new infections peak on March 16 with 5001 cases, and the deaths peak on March 30 with 258 fatalities.

The reason why the dashboard from the "Zeit", linked above, shows yet a different picture, is that they apparently have their own research going on and don't rely on the RKI numbers.

I'll close this issue.