GEE, COX for a quasi experimental design?

Hi guys
I would like your expert advice on how to go about some analyses on a trial. Basically the study enrolled 70 patients, who were observed for 8 months, The first 4 months (1-4) they were provided with standard care at home. The next 4 months (5-8) they were administered education sessions. We collected baseline info and at the end of month 4 and 8 regarding the use of health-care services. Thus we have two binary variables: 1 at time 4 and 1 at time 8 (yes, or no). We run a McNemar test to understand if there was a difference in proportions of events before and afterand the results were significant. Now we would like to model baseline predictors for these events (time, age, gender and living alone or not). The questions are:
Can we run two separate models to understand the difference in predictability at T0 and T1? Basically two binary logistic models.
Is it possible to run a generalized estimating equation model on these data using baseline predictors, to take into account the correlations between the two follow-ups and using the same predictors as above?
Can we run two separate Cox regression models at T0 and T1 to model time to event in the sample? I have some doubts because the two models would be dependent.

Thanks for everyone who can help me!