I would hesitate to interpret this so naively. For example, there are some vascular surgeons who take on patients that nobody else will touch. These are patients with many bad prognostic risk factors, but they will die without the surgery.
The surgeons who take these patients may end up having terrible stats, but be the most skilled if only because their mortality rate is not pushing 100%.
Comparing this statistic the way people compare gas mileage is a disincentive for these surgeons to operate on patients that need help, but are high risk.
Selection bias (by the physicians) should make that extremely challenging. You could adjust for underlying comorbidities, though that will often not give enough adjustment to make an impartial expert observer feel that it is sufficient.
Some states already do this. For example, for CABG and valve surgery, NY State publishes raw, as well as risk-adjusted mortality rates (RAMR) broken down by surgeon: (warning, PDF) http://www.health.ny.gov/statistics/diseases/cardiovascular/... (begins on pg 21)
Fun fact: Dr. Oz is in there and it even looks like he might be one of the better performing surgeons too. Another fun fact: it can be hard to draw inferences about individual providers this way; check out just how wide some of the the 95% CIs for the RAMRs are, and a lot hinges on how good your risk adjustment model is (details on pg 13).
CMS is actually doing that already (at least for hospitals).
Early next year, CMS will post healthcare quality metrics that it's collected from providers who bill to CMS. In 2015, hospitals will start getting reimbursed based on quality, which is a step in the right direction, however, it is chocked with problems (i.e. treating the best patients will advantage a hospital)