It always interesting when the domain of analytics enters the broad market. The ongoing outbreak of COVID-19 (“coronavirus”) has naturally led to much public speculation around infection and mortality rates.
Of course, the press has always struggled with interpretation of statistics (and I can only assume that highly analytical people are not drawn to become journalists). A classic case is confusing correlated and causal events. For example, it is true that the majority of tall people where big shoes, but it is unlikely that wearing big shoes makes you tall.
In viewing various stories related to the pandemic (and press conferences held by the president, the VP and various CDC executives) it is obvious that there is no common understanding of what the infection and mortality statistics actually mean. For example, President Trump and his CDC appointee threw out wildly different mortality rates in the same press conference. Not comforting.
At the time I write this, the latest statistics are as follows:
· 762 reported cases of infection
· 28 reported deaths
Given that 28/762 equals 3.7%. End of story? No – here is the first problem with using 3.7% as the mortality rate:
· It would seem likely that infection cases are underreported. There are reports that extremely healthy people heal from infection without hospitalization
· On the other hand, a death is very unlikely to be under (or mis-) reported.
This implies that the mortality rate is something less than 3.7%, right? No. The infection cases are at all stages of illness, with a weighting towards recent infection (due to the very rapid spread of the virus). Without analyzing the time-based cohorts of the infected, it is impossible to estimate the actual mortality rate – but it is very likely to be > 5% of those who are reported to be infected.
No comments:
Post a Comment