IIRC, it's based on the distance from the blast radius. The confounding thing about it is the fact that there is apparently a range of distance from the explosion where people have lower incidence of cancer than people who were even further way.

That seems hard to explain.