Nursing home quality of care is a chronic problem the country over. Nearly 95 percent of nursing homes flunk one or more quality measures on regulatory surveys. A quarter of homes have problems that threaten patients’ lives. Regulators strive to pull laggards up to the minimal standards required to keep licensure and certifications, but they can go only so far. A small group of facilities is always at the bottom. Nothing seems to move them up.

How about paying for performance? If they improve quality of care, pay them a bonus. Obamacare (the ACA) now does that for hospitals. They get paid a bonus of 1 or 2 percent of their Medicare bill if they meet hard-to-achieve quality markers for certain conditions. They lose money if they don’t show improvement, readmit too many patients because they failed to fix their problems the first time, or because they fail to control infections in their hospital. Together these bonuses and fines can add up to substantial new revenues, or substantial losses.

But the ACA does not apply to nursing homes. We wanted to know whether or not it should. Could Pay for Performance – P4P – help improve nursing home care. Not replace current regulations. Just complement them with bonuses for improving care.

We reviewed the P4P studies done over a decade in hospitals and nursing homes to see what has been learned about P4P in both nursing homes and hospitals. Over the years 2002 to 2012, 22 studies were found suitable for inclusion in our analysis. Six were randomized controlled trials, 10 were not randomized but had a control group, and six were before and after studies with no comparison group. We calculated “effect sizes” for each study: That is, how big and in what direction was the effect of receiving P4P on the 150 outcomes studied across all the 22 studies?

The range of effects was large. Most positive was a 45 percent increase in foot exams when a facility was paid extra to do them. Tests for kidney disease were positively affected (37 percent increase); 28 percent more patients received perfect asthma care when providers were paid for perfect care; eye exams in diabetic patients went up by more than 25 percent. A dozen and a half other outcomes went up by double digits. Several dozen more went up by single digits and more by less than 1 percent increase.

Unfortunately, nearly another two-dozen outcomes got worse when paid to get better. Most of the negative effects were less than a 1 percent decline. But two were in single digits, and one was a 21 percent decline: the amount of blood being pumped from the heart’s left ventricle. But it was in a rural hospital, and probably reflects an increase in a small number of patients.

Of more concern is the median effect size. That was less than 3 percent. Even if we assume that over time P4P would get providers’ attention and cause them to work harder at improvement, to say the 75th percentile (a very generous assumption), P4P promises to improve care by barely over 4.5 percent.

If P4P were cheap, easy and pretty certain to deliver improvement, even a 3 or 4.5 percent increase might be worth trying for. But alas, P4P is anything but cheap, easy or certain. In point of fact it is likely to be rather expensive for at least two reasons: The cost of bonuses; and The cost of monitoring to make sure that patients have the condition being paid for at the start of a payment period, and that it’s been fixed when payment is due. With bonuses and potential lawsuits at stake, careful, ideally repeated and validated measurement is necessary. And bonuses have to be large enough to attract providers’ attention.

But these are not the only problems with P4P. Indeed, the list of potential pitfalls with P4P is quite distressing. It typically starts with poor design. For example, what do you pay for? If you pay for things that are easy, you waste money and produce little improvement. If you pay for difficult, complex outcomes, you risk demanding that providers prevent or fix problems that may be the result of nature or other causes beyond the providers’ control. If you pay for only a few things, quality still lags in most areas. If you pay for many, providers may lose focus or get discouraged. And you can only pay for things that can be measured clearly, and for which there are enough cases to do analysis, and there is a reasonable chance of success.

And whom do you pay? If you pay top management, the money may never trickle down to those whose behavior at the bedside has to change. But if you pay aides and nurses, management may not cooperate in supporting extra care needed. And when do you pay? Monthly to keep providers aware of their potential to benefit, or infrequently so that successes and failures even out so you don’t have to take money back that you gave last month when things went well now that we’re in a month when declines are outnumbering improvements.

But these are just design problems. There is still implementation. Here is where things can really go wrong, plagued by such problems as:

“Slippery slope:” Caving in when providers demand that you pay for almost meeting success criteria; paying for processes of care that don’t guarantee better outcomes; paying for good outcomes that don’t really change patient well-being; paying for outcomes that are similar but not as difficult as the ones you wanted to pay for; paying for improved record keeping rather than real change in care practice; and paying windfalls to facilities that were already going to do it, while the ones you really want to improve are still not meeting outcome goals, not getting bonuses but not getting any better.

“Distortion effect:” What you pay for is what you get, but when you don’t pay for what you used to get, you may not get it anymore until you add it to the pay list. Care and attention get drawn away to things that earn bonuses.

“Cream skimming:” Facilities quickly figure out that they should admit only patients likely to improve, shun those not likely to improve, and load up on patients with conditions that are lucrative under your P4P system.

Staff may also not know how to fix the problems you are eager to pay for. Training may be inadequate even if knowledge is available. And no one knows whether paying for certain outcomes will cease to be effective over time. Providers may become inured. Finally, will filthy lucre drive out good will? Will those who used to work harder out of a sense of mission and duty now cease to volunteer unless they are paid?

These problems are not made up. They come from experiences of those who’ve done the studies and those who’ve studied them. But there are some positive lessons to take home. Among them are: transparency is important – providers must know what’s being measured and where the baseline from which they’re expected to improve started; providers have to be involved in choosing outcome performance targets and agree that they are achievable; and P4P is more likely to succeed if it’s coupled with other reforms, like web-based quality ratings.

In the paper we’re summarizing here[1] we did report one large and very successful randomized controlled trial of nursing home P4P which was done decades before any of the other less successful studies. Alas it was conducted by the present author and colleagues at what is now called the US Agency for Healthcare Research and Quality. It resulted in: (1) admission of more complex cases to the treatment group nursing homes (cases they normally shun); (2) longer life expectancy; (3) increased discharges; (4) reduced hospitalizations; and (4) and higher patient satisfaction. So it can be done. The problem is that it’s easy to go wrong, hard to get P4P right, it’s likely to have only small effect sizes, it’s expensive, and no one knows whether short term P4P improvements will last over the decades that we want nursing homes to deliver good quality.

On the other hand, it’s an approach that can be broadly applied to many nursing homes, which most reforms can’t. Those few celebrated nursing home success stories tend to focus on star performers, which is not helping the average Medicaid patient. And Congress really likes P4P and is likely to demand broader application of it. We’d be best served by trying to improve our design and implementation efforts rather than giving up on it even though, frankly, it’s not clear that the juice is worth the squeeze.

Dr. Weissert is director of the FSU Public Health degree program, professor of political science at FSU, and professor emeritus of public health at the University of Michigan.

[1] Weissert, William G. and Lucy Faye Frederick. 2013. Pay for Performance: Can It Help Improve Nursing Home Quality? Public Administration Review 73:S1, (S140–S151) October.