Longitudinal mediation analysis on survival data with repeated events

[Guest]

I would like to include a mediation into my discrete-time hazard model. However, i also have repeated events. My data is a 3 wave panel data set. I am using Stata 14.
My moderator is a binary variable, my independent variables are binary, categorical as well as continuous.

Going into the literature, I have so far understood that I can use Stata's SEM approach to model a mediation and that I can include survival time as my dependent variable. I am not sure, however, how to include the repeated events. I was thinking of including first and second event (there are hardly more than two) as level 2 variables?

Do you think this approach is feasible in Stata?

[Guest]

I should add: Is this feasible in Stata using -gsem- as a command? I ask in particular because I have a binary mediator and I cannot quiet find out if that is problematic in SEM.

Alternatively, I was wondering whether I cannot change Baron and Kenny's Method and, given that both mediator and outcome are binary, regress a variety of logit models and from there calculate direct and indirect effects. I could use the iV at my t_1, my mediator at t_2 and my dV at t_3. *

If it gets too complicated, I might have to just skip the part of the repeated events and focus on the first event. The mediation analysis is a lot more important.

*This is an example which does just that: file:///C:/Users/Maris/Downloads/dichmed.pdf, though with cross-sectional datat.