A: The ideal payment model may depend on one’s perspective. However, the general consensus among both health policy researchers and policymakers is that we should be headed toward risk-based contacting, especially global payment models with providers at risk for the costs as well as the quality of the care they provide.

A: The strongest evidence is pretty simple: National and local studies have repeatedly found that, for aspects of care that are not targeted with some incentive, the right thing is done only just over half the time (about 55%). This is true for diagnostic tests, chronic disease management, and many other areas. In stark contrast, there is a short list of measures for which clinical organizations have incentives to perform better. For most of these, the right thing is done more than 90% of the time.

A: Pay for performance is just one approach to payment reform. The current direction of CMS and most other payers is to combine pay for performance (really, “pay for performance on a small number of quality measures”) with financial risk-based contracts.

A: There is widespread evidence of quality and safety problems in medicine. However, the studies that show this have been expensive, because legacy approaches to measuring performance are very expensive. If we could measure performance more accurately and at lower cost, we would have important tools that health care managers and policymakers could use to incentivize health care organizations to do better. Our team is making good progress with this. For instance, we have used big data and machine learning approaches to dramatically improve risk adjustment for predictive analytics like estimated the expected mortality rate and average length of stay for intensive care unit patients. These are foundational tools for those who would improve ICU performance.

A: We did a trial in New York City with physicians in small practices that had a high proportion of low income patients. These are the settings—practices with uninsured and Medicaid patients (hence lower profits) and little scale (hence little ability to spread the costs of adopting good business systems across many patients)—in which improvement should be hardest. Nonetheless, when we randomized those practices to receive pay for performance incentives, they clearly improved their care.

A: Not all pay for performance projects have shown benefit of this payment method. The success of our program probably reflects not just the impact of the general concept of pay for performance, but also three other features of the program: 1) we allowed the physician practices to help decide what parts of care were important (which ones to target with incentives), so we knew the physicians thought improving care was a good clinical goal, 2) we gave the physician practices some say in how big the incentives needed to be (matching costs to revenues), and 3) we made sure the practices could get timely information about how they were doing, so they could try different strategies and see how they improved.