FierceHealth had an interesting commentary on a proposal by a non-profit group, the National Alliance for Health Information Technology, on promoting the use of a national patient identifier (see their position paper).
Finally someone has has the guts to challenge the assumption that a national patient Identifier is not possible. Currently, we are forced to do algorithmic gymnastics in order to match a patient correctly, and even then, are at best 90% effective. This has led to the dependence on high priced MPI solutions.
The problem will only get worse as more health IT infrastructure is deployed. From a safety standpoint, 90% is not acceptable. Allowing patients to "opt-in" for use of a national patient Id is a great way to address any possible privacy concerns.