I have to be honest, I do not know how to implement a scalable robust HIE or RHIO without an efficient centralized MPI. The single most important function of an HIE is the ability to uniquely identify the persons in their population, and provide the medical information that is being requested, in a timely manner. Since, in the United States, there is no national identifier, and the Social Security Number cannot be used for identification purposes, one has to often depend on probabilistic matching criteria to uniquely identify persons. Using all identifiers available (name, dob, address, gender, etc) provides the most reliable method of unique person identification ? even if a national identifier is available, all identifiers should be evaluated, as data entry is prone to error. Take for instance the following two records:

Danny Coffey 1212 Bainbridge Ave Bronx NY 10467

Daniel Coffee 1212 Bainbridge Ave Bronx NY 10467

Are they the same person? Possibly, the names are different, but are close enough to resolve to the same person ? but, one key piece of information that is missing is the date or birth which could signify that these two individuals are father and son. Your MPI needs to be able to understand and handle discrepancies such as these. If the HIE does not have a central MPI, then each query for a patients medical information requires a query (and probabilistic match) to each participating HIO (Health Information Organization), and that could lead to excessive delays in the acquiring and sharing of information. This process would be repeated every time information was requested for each and every patient ? pretty soon, you will be saturating your systems with PDQ (Patient Demographic Query) requests. As the NHIN infrastructure gets into more use and acceptance, an MPI becomes more of a necessity, as the current NHIN Specifications calls for probabilistic matching criteria be utilized. You not only have to keep track of the patients in your own organization, but you have to keep track of the identifiers that that patient is known as in other HIO's. A tricky issue for sure.

Read More about [How to implement an HIE (Health Information Exchange) without an MPI (Master Patient Index)....