A press release on the Kaiser Permanente web site showcases Doctor Robert Sallis, who conducted a study in which the data on 50,000 KP members diagnosed with COVID-19, finding that people who were less active were 20% more likely to be hospitalized and 30% more likely to die after diagnosis.

Yes, it’s just as nonsensical as you are quite likely thinking.

We know that various comorbidities, such as diabetes, asthma, old age, and kidney disease can result in worse outcomes when someone gets COVID-19. And we also know that these same conditions are associated with lower levels of physical activity.

So Dr. Sallis has found that … sicker people are less active.

But he interprets it as meaning that exercise can save you from dying if you get the disease. That might be true, but the press release comes nowhere near proving it.

Of course, it’s possible the actual paper in the British Journal of Sports Medicine has a more powerful analysis. So I found it.

Umm ….

The activity levels were self-reported. Red flag.

The authors mention confounders. That’s it. They make no effort to account for them, they just vaguely discount them: “There is also the potential issue of residual confounding due to unmeasured confounders or measurement error. However, many studies have demonstrated numerous strong benefits from PA, especially among those who suffer from a variety of chronic diseases.” Yes, but that doesn’t mean that exercise automatically has benefits in every case. It would in principle be possible to do a better analysis, in which you compared people in matched subgroups (groups matched by age and pre-existing conditions, where the only difference between members is level of physical activity). They didn’t bother to do that.

Let me be clear: if I had access to the data I could do that analysis. It isn’t mysterious. They just didn’t.

So … this paper is, let’s say, very far from convincing.