Thanks for posting your responses. As written, smoking doesn't work as a behavior for your outcome "treatment for bladder cancer" (it would work for "bladder cancer" itself) Though if smoking behavior affects treatment options, then it does work. The ecological study design you suggest for #2 is an interesting approach. Would you further subdivide the regions into smaller areas, such as zip codes? I don't know how common hematuria is, so I'm not sure if we have to worry about making the region too small to detect differences between regions. I'm also wondering if hematuria can come up frequently for the same person, as in have multiple episodes of it or is it a lasting condition until it is treated and goes away. This would affect how the regional diagnosis data would be handled.