Do pay-for-performance health programs really work?

Earlier this month, Medicare finalized the rules of a new program—mandated by the Affordable Care Act—that will pay hospitals based on the quality, not just the quantity, of care they provide. Centers for Medicare and Medicaid Services (CMS) Administrator Donald Berwick has called the initiative a “historic change” to the current system, which too often incentivizes doctors to provide more care at higher costs—even when that extra care doesn’t translate into better health. But, if the idea has generated lots of excitement in policy circles, actual implementations of pay-for-performance programs have yielded mixed results. Even the most carefully planned incentives schemes, it turns out, can lead to a lot of unwelcome surprises. The new program, the Hospital Inpatient Value-Based Purchasing Program (HIVBP), marks an important effort to address, in practical ways, many of the criticisms leveled at pay-for-performance schemes in the past. But the question still remains: Will it be enough?

Enthusiasm for the idea of pay-for-performance has exploded in recent years. In 2003, 52 such programs existed in the United States. At the time of the last major survey in 2007, that number had grown to 256. Several of these programs are actually Medicare demonstration projects and pilot studies. So what have we learned thus far from the experience of these programs? Unsurprisingly, the success of a pay-for-performance program hinges on how well it defines and measures good performance. In general, there are two ways to reward healthcare providers: 1) based on the health outcomes of patients or 2) based on how closely hospitals adhere to recommended processes of care. Take patients admitted to a hospital for heart attacks. A program focused on outcomes might measure the proportion of those patients who died within a month of their hospital visit and reward hospitals with lower mortality rates. A program focused on processes, on the other hand, would reward hospitals that followed clinical guidelines by, for example, giving higher marks to those hospitals that consistently administered standard drugs to break up blood clots to heart attack patients upon arrival. Both approaches have their advantages and disadvantages. HIVBP, for its part, uses a mixture of outcome and process measures and so has endured criticism from both sides.

The problem with rewarding outcomes, like low mortality, is that it unfairly punishes hospitals that admit sicker patients. An elderly man admitted with kidney failure just doesn’t have a very high chance of survival no matter the quality of the hospital. In 1986, Medicare publicly released mortality rates for its beneficiaries by hospital for the first time. The worst performing hospital, under this rubric, turned out to be a hospice caring for terminally ill patients. Imagine a pay-for-performance program taking money away from a hospice and giving it to a hospital in wealthy suburb treating a well-off, relatively healthy population because the hospice had worse patient outcomes. In fact, many of the early comments to drafts of the rules for HIVBP criticized Medicare for doing something similar by selecting outcome measures of performance that would unfairly punish hospitals in poorer areas that admit sicker patients. HIVBP has a simple solution to these criticisms, however. The program provides incentives for hospitals that improve upon their past performance. Hospitals serving poor, unhealthy communities, in other words, don’t have to compete against hospitals in more well-off areas. Those hospitals just have to compete against themselves.

The Medicare program also rewards hospitals based on how well they adhere to certain standard practices of care, which helps avoid some of the criticisms associated with rewarding outcomes. Unlike the health of their patients, hospitals clearly have more control over how they administer care, and thus they can’t as easily blame outside influences for a poor grade. For this reason, Medicare has designed the incentives for its program to emphasize processes of care—but process measures, unfortunately, introduce their own problems.

Critics have argued that a focus on simple, uniform standards promotes “cookbook medicine,” which limits the freedom of doctors to decide the best course of action for their unique set of patients. Indeed, if Medicare sticks with the same measures of processes for too long a period of time and doesn’t keep revising its standards, then the program risks encouraging an overly simplified model of care. Hospitals would likely get complacent and just try to hit the benchmarks on the small set of indicators that Medicare rewards, without thinking about how to put in place the infrastructure to improve quality of care across the board. To address this, the government plans on continually adding other indicators over time, and it will rotate out indicators when the great majority of hospitals have achieved a high level of performance on them. Still, Medicare must carefully monitor the effect of the program to ensure that hospitals don’t achieve a high level of performance simply by diverting resources from other unrewarded, but nonetheless important areas of care.

Another reason that policymakers need to closely watch what happens with this program is that pay-for-performance schemes often have unintended consequences. Even though Medicare has carefully designed its new incentives to reward both achievement and improvement, it could turn out that it is much easier for high performing hospitals with lots of resources to maintain a high level of quality than it is for low performing hospitals to invest in the infrastructure and training that’s necessary to improve. For instance, major improvement might require a designated staff member in the hospital devoted to quality assurance, but a small hospital with an already overworked staff may not have the resources to commit to that. In fact, several providers in a study of pay-for-performance programs in Medicaid health plans reported just that. If this turns out to be the case for Medicare as well, then the program will end up taking away money from smaller hospitals in rural areas that don’t have the resources to improve and giving it to successful, teaching hospitals in exchange for maintaining their already high quality. Another potential, negative consequence would be if providers started turning away hard-to-treat patients in order to increase their scores. Clearly, this would not be the optimal—nor the ethical—outcome Medicare desires.

Medicare has started relatively small by choosing to redistribute only 1 percent of reimbursements, so it will have the chance to evaluate the effects of a promising program without it having disastrous consequences if it fails. This conservative strategy also means, of course, that the program won’t have too large an impact, at least in the short term. Still, it’s the beginning of a sorely needed change in a healthcare system that rewards quantity over quality, with terrible consequences for the federal budget and for patients. The first generation of pay-for-performance programs has seen some modest successes and some failures. The current Medicare program admirably tries to address many of these shortcomings. In the end, we can’t know if pay-for-performance will really work for the country’s healthcare system until we try it out—and, undoubtedly, the time has come to try. “This is just the opening bid,” said Leslie Greenwald, a principal scientist at the Research Triangle Institute International and co-editor of a recent book on pay-for-performance in healthcare. “We need to change the mindset that more is better.”

Jake Marcus is a post-bachelor fellow at the Institute for Health Metrics and Evaluation.

Follow @tnr on Twitter.