The primary purpose of using a Master Patient Index is to avoid having to consider multiple pools of patients. The very job of an MPI is to coordinate patients across systems and, in the best scenario, this would be done silently for both applications (client code calling the API) as well as users. Using an MPI shouldn’t mean that you now have two pools of users; rather, it should be that your single pool of users is coordinated with other servers/sites sharing the same MPI. The more invisible this outside of the API (to users & clients), the better.
For example, showing duplicate hits for the same patient (one from the local pool of patients and one from the attached MPI) isn’t integrating with an MPI; rather, it’s accessing an external MPI and leaving it up to the client application and the end user to perform the job that the MPI should be doing.