Hi Joe,
I agree with Maria and Vicky that interrupted time series using the patient as the unit of analysis would be a good analytic strategy. Maria's point of trying to come up with alternate outcomes is also interesting. However, it might be difficult to figure out an adequate control for the dichotomous outcomes she mentioned. Should the patient be their own control? And is there a target for the proportion of patients who have mental health visits and primary care visits? You would also have to think very carefully about your time period. Is three months an adequate amount of time ot observe a difference?
The good thing about a study at the VA though is that your sample size always has the potential to be big!
Good work on your protocol so far, as others have said, it's been nice to see it progress over time!
Maria