I think it's important to keep in mind as with all things, correlation doesn't automatically mean causation - less healthy people are probably more likely to be drawn to less active jobs, for instance. Maybe there is a randomised intervention study that's been done?