I wish this was available to the general public, and about 6 months to a year from now it should be, but this commentary in JAMA from August 18 does an excellent job of succinctly laying out a number of the inherent problems in current, proposed, and theorized models for compensating physicians. Dr. Brook's focus is to lay out the benefits and drawbacks of three basic physician compensation models: salary (fixed compensation), capitation (compensation based on numbers of patients cared for), and fee-for-service (compensation for doing stuff).
With salaries, doctors may feel empowered to provide a great amount of detail in their care, but there is no incentive to see the volume of patients necessary to keep a practice viable.
With capitation, doctors may keep costs down by ordering fewer tests, but the motivation will be to have as many patients as possible in their practice in order to make a living.
With fee-for-service, doctors are encouraged to order and "do" as many things as possible, regardless of the benefit to the patient. Furthermore, this model provides patients with a false sense of benefit because more testing is incorrectly assumed to be associated with better care and better outcomes (ie: annual stress tests).
Dr. Brook then goes on to discuss the next generation of payment models based on pay-for-performance. In theory, this model is an excellent approach. It would be great to design a physician payment model that rewards those that provide the "best" patient care and the highest quality. But, as they say, the devil is in the details. While it is relatively easy to identify outcomes that are good goals to strive for (discharge of heart attack patients on beta-blockers), documenting, reporting, and designing payment based on these is incredibly complex and difficult to implement.
Pay-for-performance seems to provide a modest way of improving the quality of care delivered, but is no panacea for compensation of physicians.