There is no perfect method to measure the exact fatality rate of a novel virus when the epidemic is ongoing. CFR is probably the least accurate number.
If anyone thinks that a CFR that varies between 0.2% and 28.8% (!) is a good way to measure the risk this virus poses to a population I need to bring out my facepalm emoji —> 🤦♀️
The death rate per 100,000 has been increasing quite rapidly in the last couple of weeks in the U.S., one assumes due to the surge in infections that started over a month ago. That current number of deaths/ 100,000 probably looks less ”palatable” than 3.3% (which is the current case fatality rate in the U.S. according to the linked data). I would assume that’s the reason he prefers to look at those numbers, as opposed to the preference being based on a legitimate epidemiological rationale.
If testing doesn’t ”capture” the true number of infected individuals, CFR will be overestimated. So a country’s testing strategy can have a significant effect on CFR.
Cases that are detected and active, but not resolved, may die in the future and this can lead to CFR being underestimated. This effect can be significant during the exponential growth phase of a pandemic.
Deaths per 100,000 is probably a more reliable way to measure the impact the disease has on a given population. However, since the accuracy/degree of correct attribution of Covid-19 related deaths is also a variable, this needs to be assessed together with excess mortality for the time period in question. Many countries show a significant discrepancy between the number of Covid-19 related deaths reported and the total number of deaths that have occurred.