Hi Sara,
It has been great to see your protocol progress. While the interrupted time series may be more feasible than the clustered randomization given your implementation at a single site, you may want to consider expanding this to a clustered randomization (I know you had mentioned Kaiser and other Pedi clinics) if you could. The disadvantages of an interrupted time series, as discussed in class, is the inference of causation is weaker and the analysis scheme may be more complicated requiring the same size to be larger, perhaps. Rachel makes some good points to think about.
The unit of analysis will be the patients, as you stated. I think it would be interesting to look at the rates or types or something regarding complications patients see before the intervention and compare it to after.