Are Hospital Pay-for-Performance Programs Failing?
The logic of pay-for-performance systems is simple enough: pay doctors and hospitals based on how well their patients are doing, rather than on the number of medical services they provide. The payment structure was designed to fix a central problem in American healthcare. The United States spends far more per person on healthcare than any other country, yet has the poorest health outcomes in the advanced world. Pay-for-performance, also known as value-based purchasing, was meant to encourage doctors to optimize the welfare of patients while discouraging spending on unnecessary care.
There was never any evidence that pay-for-performance works, said Li professor of international health Ashish Jha, who studies the effectiveness of economic incentives in the healthcare market. A recent paper published this April in the BMJ by Jha finds that the federal pay-for-performance program under Medicare, the Hospital Value-Based Purchasing (HVBP) program, hasn’t had any impact on mortality rates. Nor do HVBP-participating programs show any statistically significant advances over the small number of hospitals that don’t participate in the program. No hospitals, not even those with the worst mortality rates before the program was implemented, showed an improvement that could be attributed to pay-for-performance. “We looked across all different conditions and couldn’t find a single one where it seemed to have a meaningful effect,” Jha explained.
In the early 2000s, the Bush administration ran a pilot version of pay-for-performance at about 200 hospitals that agreed to tie their payments to certain quality measures. Even then, Jha said, “the broad consensus in the community was that it didn’t really work.” Despite this, the system was implemented on a national scale by the Affordable Care Act (ACA), which created the HVBP program at more than 90 percent of U.S. hospitals. “There was a sense that we needed to begin somewhere,” he explained, “and that this program would be good to test out nationally.”
Now, a few years into the ACA, Jha’s research measures the disappointing results of pay-for-performance. He blames the program’s failure on its poor design: HVBP affects only up to 2 percent of Medicare payments to hospitals, and the size of the financial incentive is determined through a complex formula with many variables, some of which doctors might not be able to control. As Jha put it: “Imagine if your boss came to you and said, ‘I’ll put just 2 percent of your salary at risk, and you have to do these 38 things.’ So you just ignore it.” Clearer and larger incentives, he argued, might begin to produce results. “Instead of incentivizing 35 things, incentivize three things.”
A recent review of literature on pay-for-performance by professor of population medicine Stephen Soumerai argues that many studies showing promising results for pay-for-performance in the United Kingdom have methodologies so flawed as to be useless. Soumerai focused on history bias, a design flaw that fails to control for medical trends that were already occurring, independent of pay-for-performance. The U.K. hospitals had already been making gains in mortality rates in the early 2000s, and some studies wrongly attributed that to the introduction of pay-for-performance. Findings like these, Soumerai warned, can wrongly encourage other nations to commit billions of dollars to pay-for-performance programs. He suggests that expensive policies should always be supported by evidence from controlled, experimental or quasi-experimental studies, like randomized control trials, where patients are randomly assigned to different intervention groups (in this case, to either a pay-for-performance or non-pay-for-performance hospital), and their results then compared.
Systematic reviews consistently show no gains from pay-for-performance. Why then does it continue to make its way into policy, in the United States and elsewhere? Soumerai, who has been publicly critical of financial incentives in healthcare, challenges the basic assumptions that go into economic models. “Health economists invent these things in their offices and advise presidents and Congresses, but they don’t talk to doctors in the field about what’s important to them, or how you might improve practice,” he said. “It’s much easier to say, ‘We’ll pay for an outcome or penalize against it.’” Policymakers need to become much better at understanding the conditions in which doctors work that aren’t captured by theory, he argued. Many doctors have an intrinsic motivation to care for patients, for example, but this might be distorted by extrinsic, financial motivation—a known cognitive bias. In some cases, Soumerai added, pay-for-performance might even be harmful—it could discourage doctors from seeing the sickest patients.
In the last few years, health economists have also begun to question the value of market incentives in influencing the demand side of healthcare. Increased cost-sharing, a scheme that passes more of the costs of medical care to patients through higher deductibles and co-pays, shares with pay-for-performance the goals of reducing unnecessary care and rewarding high-quality care. Cost-sharing was once embraced as an answer to runaway healthcare inflation—the idea was that it would force patients to shop for the best prices and reduce their spending on unnecessary care—but now its efficacy is being widely challenged. The evidence on cost-sharing suggests it causes consumers to use less healthcare, even when they need it, because patients aren’t good at understanding which medical services are necessary and which aren’t. They also lack the ability to comparison-shop—trying to get information about prices from insurance companies is frequently a traumatic ordeal—and even patients who are given shopping tools by their insurers tend not to use them.
One might find interesting analogs to pay-for-performance incentives in other areas of policy, including the use of school funding as a blunt instrument to punish poorly performing schools under the No Child Left Behind (NCLB) Act. Like pay-for-performance in the healthcare space, these policies, now widely repudiated, seem to have had the unintended effect of harming the most at-risk populations that they were intended to help (in the case of NCLB, by cutting funding from low-income schools). Programs like these, Soumerai argued, haven’t worked for the same reason they haven’t worked in healthcare: predicting human behavior is a messy and often impossible project. Policies that make public funding contingent on such predictions are inherently risky, and should be pursued only when based on a great deal of strong evidence. There’s a broadening consensus in the policy world that economic incentives need to be designed much more carefully—accounting for culture, values, and psychology—to produce the outcomes they’re intended to.
Scholars like Soumerai doubt the ability of market incentives to improve healthcare at all. Jha and others closer to the mainstream of policy debates suggest the question isn’t whether incentives inherently work or not, but whether they’ve been properly designed to accomplish the things we want them to. “There’s actually a very robust debate happening in healthcare about whether incentives can work at all. It’s a funny debate because the answer is: ‘Of course they can.’ Make the incentives big enough, and you’ll see change. Doctors have intrinsic motivation, but if the incentives clash against that, you’ll end up doing more harm than good,” Jha said. “We’re still in the early stages of understanding how humans make decisions.”