> If the data shows members with annual checkups have better health outcomes on average than members without annual checkups, that is something that's relatively cheap and easy to do with potentially significant impact.

That or it's yet another example of selection bias. There have been so so many things like this where the epidemiological data shows a correlation with health, but there isn't actually a causal link. For example, annual checkups might correlate with better health because it's a more common behavior among people who can afford to do it, and wealthier people tend to be healthier.