What the Covid-19 mortality rate tells us
Data is at the core of corona virus outbreak with every country publishing numbers of new cases, deaths and recoveries almost on a daily basis. WHO is coming up with daily reports of the pandemic covering these numbers across the globe. It presents an opportunity for data scientists to analyse and make sense out of it, maybe apply some models and forecast the upcoming days to then try and prevent the worse from happening. However, a very simple analysis of the data present tells us something that we all should be aware of before applying any kind of models. The simplest of analysis: calculating the mortality rates of different countries by dividing the deaths by cases, ironically is also acting as a warning to attempt further analysis. The reason is simple, the major insight that comes up after the seeing the number: the data is unreliable.
Let’s compare two countries: Germany and Netherlands. Germany has reported almost 6x times the cases reported by Netherlands but still lags behind in the number of deaths. There seems to be no major difference in medical infrastructure and resources available and the reason for this difference could be anything from selection bias in data to just a different system for reporting deaths. To give an example, UK is reporting the deaths of all those patients who had coronavirus even when the reason of death is not identified properly and their infection was mild, whereas some countries are strictly reporting deaths caused due to the virus infection. One more thing we should note here is that countries like India and Pakistan are just not doing enough tests due to which there are still a lot of un-identified cases out there which, if factored in, would increase the denominator significantly; a thing we observed when US ramped up its testing.
The conclusion: the data has a lot of statistical landmines as another article notes and we should be careful while attempting to work with it.
