So I did a few back-of-the-envelope calculations. What magnitude do you have in mind? Because what I got was single-digit numbers of weeks, and why bother to discuss differences as small as that?

I started by assuming that since most health care cost is concentrated in the last four years of life, that's for how long retirees can affect the statistics by moving or not moving. 4 is a small number, and the number of people who move also cannot be large compared to the general population in the area they move from, so no matter how I tweaked it, I arrived at near-zero results.