Bibhav,
I agree that a clustered randomized trial is feasible and suitable for your intervention and the variation of a stepped-wedge would work, as well. The issue with stepped-wedge we discussed in class was the actual implementation of it that may be an issue. The clustered randomization will help avoid contamination of the control subjects. I can understand the decision to do a stepped wedge since it would alleviate any issues with confounding of other organizational changes happening near the time of the intervention. I agree with your unit of analysis being the PCP since we discussed the benefits of having many clusters and minimal number of individuals per cluster. Essentially what you have is a cluster at the individual level... One question I have is whether all the PCP's at the 11 hospitals will eventually get this training. Will you implement at the hospital level or have one PCP at each hospital site trained and see how they go? I was a bit confused on that.