Cheers on the study!

Let me first describe what I meant by fringe though. Take a random adult going to the hospital, not elderly, with no other major health conditions. When he walks in the door, what are the distributions of issues that he might end up having? Sepsis is going to have a probability of near 0. By contrast the typical patient that might present with sepsis - elderly, other major health conditions, well into senescence - he is generally indeed 'knocking on deaths door.' He might not answer this time (though there's a decent chance he will!), but he will imminently.

Your study compared hospitals based on a number of factors. The most significant was high volume, but in that case the difference between the highest volume hospitals and lowest was a 13.3% rate of readmission vs a 11.2% rate of readmission for hip replacement, and 12.4% vs 11% for knee replacement. Again I think this is another example of when you look at the actual data, outcomes fall quite close.

Beware their method of taking a sampling and breaking it into buckets and comparing those buckets. If even hospitals/patients were identical (which I'm certainly not claiming) and so the results were literally just random noise on a distribution, you'd see a major difference between the top and bottom buckets due to the nature of random distributions - 68-95-99.7 and all that. Their results show a signal beyond that, but it's generally a very misleading way of presenting data because of this issue.

Pre-pandemic hispanics had a life expectancy of about 82, which I described as comparable to the 85 of billionaires. I'd certainly expect billionaires to be higher for the endless reasons outlined in the already linked Hispanic paradox. The fact that it's only 3 years, less than 4% longer, is the point.