乌鸦传媒

Hospitals that instituted pay-for-performance programs for certain conditions saw some short-term mortality reductions in MI, heart failure, and pneumonia, but the results were not maintained over the long term, researchers found.

In a study involving a total of 161 hospitals in England, overall 30-day in-hospital mortality was greatly reduced in the hospitals that participated in the pay-for-performance programs during the first 18 months and continued to drop for the rest of the study, according to , of the Institute of Population Health at the Manchester Centre for Health Economics in England, and colleagues.

However, by the end of a longer, 42-month follow-up period, the lower 30-day mortality rate at pay-for-performance hospitals was not significant, and nonparticipating hospitals had greater mortality reductions overall for the three conditions studied (by 0.7 percentage points, 95% CI 0.3-1.2), they wrote in the Aug. 7 issue of the .

On the other hand, when mortality rates for patients with conditions not covered by incentive programs were analyzed, the pay-for-performance hospitals had consistently better outcomes than the controls -- on average 1.2% lower (95% CI 0.4-2.0) -- suggesting the programs might have a "spillover" effect.

"Although short-term improvements in the quality measures for conditions related to incentives were sustained in the long term, our analyses provide no evidence that the incentives have a long-term effect on 30-day mortality," the authors wrote.

And in the U.S.

, senior associate director for policy at the American Hospital Association (AHA) in Washington, told 乌鸦传媒 that the study was an interesting one, but "in terms of implications for pay-for-performance here in the U.S., it's a little tough to draw direct conclusions because the structure of the programs is different and the financing structure in the U.K. is different."

In addition, "when you're looking at outcome measures, they are affected by care provided by hospitals and other providers, but outcomes can also be influenced by a patient's clinical factors and the kind of communities patients live in, so that makes one-to-one mapping a little more difficult," he said.

, the AHA's vice president for quality and patient safety policy, said that 30-day in-hospital mortality is not necessarily the most sensitive way to measure the success of pay-for-performance programs and the quality measures that come with them.

For example, she noted that one of the quality measures the researchers looked at was smoking cessation counseling. "It would be odd to me, knowing some of the clinical science behind these measures, if smoking cessation counseling had an impact on 30-day mortality; you just won't see it that quickly."

"These measures are not perfectly aligned with the outcomes they're using, and therefore you see somewhat of a muffled impact of the pay-for-performance program."

Foster, who, like Demehin, spoke while a public relations person sat in on the call, said she has noticed spillover effects -- such as the one suggested in the study -- at U.S. hospitals. "Most notably for us, this is around the area of readmissions," she explained.

"In order to reduce heart attack, heart failure, and pneumonia readmissions, hospitals have focused on a number of strategies, and hospitals have then readily employed them to help other complex patients, once they are discharged from the hospital, to stay out of the hospital."

Modest Results

This latest finding mirrors earlier studies in which pay-for-performance programs have also had modest benefits.

found that hospitals that combined public reporting with pay-for-performance incentives had more quality improvement -- ranging from 2.6% to 4.1% depending upon the measure analyzed -- than those relying on public reporting alone.

To explore the issue further, the researchers compared 24 hospitals in northwest England that participated in a pay-for-performance program involving three conditions -- acute myocardial infarction, heart failure, and pneumonia -- with 137 hospitals in the rest of the country that did not participate in the program.

They also collected data on outcomes from all hospitals for five other conditions: acute renal failure, alcoholic liver disease, intracranial injury, paralytic ileus and intestinal obstruction without hernia, and duodenal ulcer. In addition, data on the hospitals' performance on particular quality measures were collected. The researchers looked at data on more than 1.8 million hospital admissions over a 5-year period from 2007 to 2012.

The financial incentives changed several times over the course of the program. For the first year, hospitals that scored on the top quartile on quality measures received a 4% bonus payment, while those in the second quartile received a 2% bonus.

For the next 6 months, hospitals scoring above the first year's median score for performance received an "attainment bonus" and became eligible for two more bonuses based on their improved and absolute performance scores. For the final months of the study, a fixed amount of the hospital's expected income was withheld and was paid out only if certain performance levels were reached.

Two Different Analyses

The researchers performed two kinds of analysis on the data: a "difference-in-differences" analysis that compared changes in mortality over time for all eight conditions between hospitals participating in pay-for-performance and nonparticipating hospitals; and a "triple-difference" analysis that compared changes in mortality over time in participating and nonparticipating hospitals for conditions linked to the financial incentives and then subtracted changes in mortality over time for conditions not linked to the incentives.

They found that hospitals participating in pay-for-performance programs showed improvement on all the measures of quality over the first 18 months of the study, and further improvement during the rest of the study, although the "rates of improvement slowed over time and, for some measures, especially acute myocardial infarction, plateaued at high levels toward the end of the period."

The authors also found:

• Risk-adjusted mortality decreased over time in both the participating hospitals and the rest of the hospitals for all eight conditions in the study.
• The "difference-in-differences analysis" confirmed that the pay-for-performance program had a significant overall effect on mortality in the short term (-0.9 percentage points, 95% CI minus 1.3-minus 0.4), including a significant decrease in mortality among patients with pneumonia (-1.5 percentage points; 95% CI minus 2.3-minus 0.7) and nonsignificant reductions among patients with acute myocardial infarction and those with heart failure.
• The "triple-difference" analysis demonstrated an overall short-term mortality reduction of -1.5 percentage points (95% CI minus 2.6-minus 0.5), with a significant reduction in mortality for pneumonia (-2.2 percentage points, 95% CI minus 3.4-minus 1.0) and nonsignificant decreases for acute myocardial infarction and heart failure.
• Between the 18-month and 42-month periods, the risk-adjusted mortality for conditions associated with pay-for-performance incentives fell by 1.6 percentage points in northwest England and by 2.3 percentage points in the rest of England. However, the reductions in mortality for conditions not linked to incentives were larger in the northwest region than in the rest of England.

"Thus, the short-term improvements in mortality in the northwest region as compared with the rest of England were not maintained," they concluded. "The change in mortality from before the start of the initiative to the end of the long-term period was not significant in either the between-region difference- in-difference analysis (-0.1 percentage points, 95% CI minus 0.6 to 0.3) or the triple-difference analysis (0.4 percentage points, 95% CI minus 0.6 to 1.3)."

"Spillover Effect" Possible

The researchers did say that the loss of effect from pay-for-performance programs might be due to a "spillover effect" -- that is, improvements in care in hospitals not in the incentive program, or improvements in care for hospitals in the program that were in clinical areas not targeted by the study. They found "limited evidence" for such a spillover effect.

Comment