Will Pay-for-Performance Improve Health Outcomes?

Health Reformers of all political stripes seem to place a lot of weight on the ability of payment reform to improve health care quality and value. One common argument is that the current fee-for-service incentive structure for doctors and hospitals is a recipe for high cost and marginal quality because it rewards providers on a per-procedure basis. That means that doing a lot of unnecessary or only marginally beneficial things generates more revenue than doing a few very effective things.

Enter the idea of pay-for-performance, a name that's relatively self-explanatory. Simply put, providers would be rewarded with higher compensation for improving the health outcomes of their patients, or the population(s) that they serve. So in cases where patient counseling has been shown to be effective, a doctor would be incented to spend the time talking to a patient, instead of opting for some intervention that reimburses better but is not as effective, such as prescribing a medication or performing a procedure. Not only does this sound good for patient health, but the idea that you could reduce the rampant overuse of many medically unnecessary services should equal great cost savings. Better health, lower cost, those are the ingredients for better value.

The problem? It might not work, and could actually cost a lot more. This according a new study in the British Medical Journal, which analyzed the impact that a shift to pay-for-performance had on specific health outcomes after it was implemented in the UK in 2004. An international group of researchers from Harvard in the US, Nottingham in the UK, and the University of Alberta in Canada analyzed data from 358 primary care practices in the UK, plucking out 470,725  patients with high blood pressure. They then looked at patient-level outcomes (management of high blood pressure and related health outcomes) over a seven-year span of time, four years before the switch and three years after the switch to pay-for-performance.

Their findings?

Payment reform did not improve outcomes. There was no change in the incidence of strokes, heart attacks, heart or renal failure, or mortality. Compounding the conundrum, pay-for-performance requires longitudinal information systems such as electronic medical records that track patient outcomes over time, meaning potentially significant investments in technology and people-hours.

How could this be?

One dilemma is that pay-for-performance is only as good as the measures used to determine performance, and in some cases there are dramatic unintended consequences that actually lead to worse health. How? Diabetes and high blood pressure are good examples. With these conditions, the most basic performance measure would be some measurement cutoff: Hba1c-a measure of blood glucose- and blood pressure (BP), for example. Providers who bring all their diabetics below a certain Hba1c cutoff would be rewarded, the same for BP. Those precise cutoffs would generally be based on some evidence that they reduce the risks of other disease-related complications. The trouble is that what often happens is that lots of effort is spent getting folks who are close to the cutoff down to the cutoff, while those who are farthest from that performance measure, and will perhaps never get to it, receive less focus. Ironically this may mean that providers would have less incentive to work with patients at greatest danger (highest hba1c or BP) who would gain the most from even small improvements, and more incentive to work with patients who are borderline cases and have the least to gain by reaching performance cut-offs.

Could performance measures increase harms?

Making matters worse, those who are closest to the cutoff may also be the most likely to suffer side effects from treatment: Having blood glucose or BP drop TOO low as a result of medication, for example.

Naturally the study isn't the final word on how payment reform may or may not improve population health. And the above weaknesses in performance measures by no means suggest that we should do away with them. These are reminders, however, of just how complicated the interplay of payment, health, measurement and policy are.

Is there a better way to incent better performance from health care providers? What about uncoupling payment from outcomes, putting doc's on salary? What's your take?