Earlier quoted context omitted.
> For everybody or for some people? What is the causal chain? The study it is based on is from Al Aly [1], who is using a dataset of (mostly elderly, unhealthy) veterans who regularly seek hospital care from the US government hospital system (VA). It is an older, sicker population, and no, the results aren't generalizable. Moreover, his methods are dubious. I've dug into one of Al Aly's other paper's downthread [2],…
I only have a MSc, but with a more biomedical focus. I disagree with your overall impression and contextualization (even if you may be right on some of the finer points). I don’t agree with your assessment of Topol’s credibility, or on that of the relevance and quality of the Nature article.
For example, there's absolutely nothing stopping Al Aly from using a real, consistent control (e.g. influenza, or even just the same negative control) instead of a grab-bag of different illnesses. The reason he switches them with every paper is probably because his synthetic weighting scheme produces inconvenient results from experiment to experiment when the same control is used. But we can only speculate, because we don't have the data. Remember: he's re-weighting his covariates using these controls! This is a real, objective problem.
My opinions about Topol are my own, which is why I said they're editorial, but I'm tired of the guy's constant doom-and-gloom messaging and inability/unwillingness to critically appraise data. Even in the most charitable, rational take on Al Aly's work, you can't casually extrapolate it to the entire population the way that Topol does.