Hoping someone can help me out with a question I have on an analysis looking at prosthetic hip fracture incidence following hip replacement surgery (exposure=type of implant, with reference being the standard of care implant).
Sample: There are roughly 9,000 observations, all who have received one of a few types of prosthetic hip implants. And there are only about 80 cases of hip fracture over follow-up (~5 years).
At first it seemed like a pretty straight-forward cox regression scenario; but now I'm realizing that some people (~n=1,000) either had both hips replaced at once (bilateral replacement), or had one hip done and came back at a later date to get their other one done...
So my main question is how do I account for multiple hips? Do any of the following?
1) Do I analyze observations as 'visits to the hospital', where repeat patients are counted twice (or I only use the first hip replacement); or
2) are observations the hips themselves, in which case each person has two observations and it's a matched analysis? And if it's the latter, what do we do about hips that weren't replaced at all, i.e. have no implant (which would be most, since the vast majority only had one hip replaced).
Or 3) Do I restrict the sample to only single implants...
Would really appreciate some advice, and/or any similar studies if someone knows of any. Let me know if I can clarify anything. Thanks a ton in advance!