Does someone have experience on how to include doses of different drugs in a regression model? Is it possible? I am working in a retrospective cohort comparing the effects of two drugs. The Outcome is cure and we are using a Logistic regression model. One problem we face is that since there is no standard dose for one of the medications there is significant variance in the doses of this drug. Stratifying and doing subgroup analysis is not a good option since this would lead to power reduction and we already have a small number of patients. Thus, we wanted to include the dose of each drug in the regression model. Using the total dose of each drug we experienced some problems(Complete separation). I read an interesting paper by Gelman et al talking about standardization of the input variables in regression models (Gelman 2008). I kept thinking if standardizing the doses of each drug would be feasible. I could for example divide the dose used by each patient by the mean dose used for all patients that used that drug and include the result in the regression model. Would this be reasonable?

More Daniel Barroso's questions See All
Similar questions and discussions