Can I assume these are independent?

ted

New Member
#1
Hello! Trying to resolve in my mind a small dilemma.

I'm assessing the risk factors for blood loss in surgery. The data come from an observational study. I ran bivariate analyses on each of the risk factors and the outcome variable. Next, using proc genmod, built a model based on AIC to test the risk factors after controlling for the other factors in the model.

Then, after questioning some suspect data I was told two different surgery records were the result of two operations on the same patient. Found out that of the 168 surgeries, there were 160 unique patients, where the majority provided just one, yet 6 provided two and one provided three. After assuming the data was independent, I'm wondering what changes I may have to make. (Note to self: Question everything before you accept someone's data!)

My question: How do I account for this dependence? If I don't, how do I justify? I know the real question is whether this association persists after controlling for the explanatory variables in the model. My investigator thinks it's OK, however, the reason I asked about the data was BECAUSE they were alike and found out it was on the same patient. I switched to glimmix and included a random effect for subject and it was estimated to be zero; I guess it is close to being linearly related to the residual variance. I don't think this effect will have an impact on my results, but I need a better answer than "I think it's fine!".
 

Miner

TS Contributor
#2
Try removing the 8 duplicate/triplicate surgeries and repeating your analysis. Then compare the results with and without this data to see whether it changes your conclusions.