As was mentioned in the comment section of the article, it would be interesting to see if this effect really exists (by trying to shift the numbers backwards).

Furthermore, I think it is a striking reminder that even (seemingly) simple data like mortality can be subject to strong bias.