I said it was a good read, but not a good argument yet, as there are possible bias in the methodology and percentages. It need more reviews from the peers and comparison with other hospitals from other countries (that's a given)
If you take a sample of 409 persons with 61,4% (251persons) over 65yo, you can't compare it that easily with another sample of 1202 with only 48,9% (587persons) over 65yo. If half of those old people die because they are more at risk, you have 31.37% mortality rate for control group and 11,93% mortality rate for the second group... but still ... half of the old people still died. So for the same mortality, we have a reaaaally different mortality rate. No drugs involved here, just casual statistics from the sample. Thus I can say the second group have a lesser mortality rate because I kept the window open....
The numbers are still in the range of the others observations, as mortality in hospital goes from 10 to 30%, but their cox model is... a little weird, based on the samples.