Funny thing: in my job, I've calculated the relative survival rate of prostate cancer (the difference in mortality risk between people diagnosed and a similar population in age, sex, and race). Cause of death is not considered, only that there was a diagnosis (avoids complicate deaths and captures indirect influence). It'll often be that men diagnosed with prostate cancer have a better 5-year survival rate than men who did not receive a diagnosis. And that's after excluding non-invasive tumors.
Why? No definite idea right now. Possible correlation: a diagnosis happens when a man actively cares about his health (regularly brings up troubles with his primary care physician, asks for screening tests), which also implies he has the money to do so. This is also bolstered by the fact the survival stats don't include diagnoses at time of death. So it misses the guys who never got tested before it killed them.
By the way, if you look up the relative survival stats on the CDC's website, they won't show anything above 100% relative survival. Im short, they never let the added risk of death go below zero. Which is dumb, in my opinion. It's editing data because it violates the assumption the model is perfect.
Does this correlation still hold if you look at men below a certain age? Since prostate cancer increases with age then you could get a bias towards the more healthy as the non-prostate cancer peers die earlier of some others cause.
I can't answer that until tomorrow (vacation). We try to reduce bias from competing causes of death by using the Pohar Perme method [0].
Though I now wonder if there is a larger difference for prostate cancer than other cancer types. It's commonly believed by cancer researchers that, if they autopsied everyone, they would find a prostate tumor in most men over 65.
Why? No definite idea right now. Possible correlation: a diagnosis happens when a man actively cares about his health (regularly brings up troubles with his primary care physician, asks for screening tests), which also implies he has the money to do so. This is also bolstered by the fact the survival stats don't include diagnoses at time of death. So it misses the guys who never got tested before it killed them.
By the way, if you look up the relative survival stats on the CDC's website, they won't show anything above 100% relative survival. Im short, they never let the added risk of death go below zero. Which is dumb, in my opinion. It's editing data because it violates the assumption the model is perfect.