My study focuses on comparing hospitalization rates pre- and post-enrollment in a medication adherence packaging program.
I am measuring hospitalizations for all patients for one year prior to their enrollment, and for one year following their enrollment.
I believe I am making this more complicated in my head. What statistical method would be most appropriate to use to measure my results?
You end up with a contingency table -> chi²-test.
If you have covariables: logistic (binomial) model.
Dear professor Jochen Wilhelm I think it is a complicated question. It depends on the type of the disease. For example, for the diseases of type SIR, if we have 50% rate in the first year, these people can not be hospitalized in the second year (and for the second year, 50% raw rate is in fact 100%). On the other hand, your suggestion entails losing a large proportion of data. What is your opinion?
Thank you both so much for your input. My initial concern with using chi2 test was that it would be useful if I was looking at the proportion of patients who were hospitalized in the pre- vs. post-intervention periods. But instead, I wish to compare the total quantity of hospitalizations (i.e. if a patient was hospitalized 5 separate times in that year, I want my statistics to reflect this). Please correct me if I misunderstand. Thank you!
Ah, ok... so you have a number of hospitalizations (within a year pre and within a year post intervention). This can be modelled using a negative binomial model. The correlation between pre- and post intervention data from the sampe patients can be accounted for with a random intercept term (so the solution is a mixed negative binomial model).
- Badges
- Science topic
- Similar topics
- Correction