[yorwba]

Life expectancy is calculated from the mortality statistics of the recent past: the share of newborns that made it to one year old, one-year olds that made it to two years, etc. etc. accumulated to get the probability of a newborn making it to age X and then dying, assuming conditions stay the same (which they never do).

"At birth" just specifies the starting age. You could also look at life expectancy at 5 to exclude infant mortality, or even life expectancy at 69, though that won't tell you about the future survival of any particular 69-year-olds any more than life expectancy at birth tells you about the future survival of babies born today.

So the drop in life expectancy in countries with many covid deaths is likely to be followed by a jump back up as the number of covid deaths decreases, while China might experience something similar depending on how the current trend of increasing case counts develops.