Context:

Although evidence of quality problems has been available for years, purchaser interest in quality-based purchasing (QBP) is a recent phenomenon. Furthermore, employers who support quality-based purchasing have expressed uncertainty about how to measure quality, especially outcomes, and what incentives to offer to stimulate performance improvement.

Objectives:

The objectives of this project were to develop a conceptual model of how incentives influence provider behavior, to summarize what is known from randomized controlled trials about the effectiveness of different QBP strategies, to describe ongoing QBP research, and to perform simulations to determine whether outcomes reports are too influenced by chance events to be used in QBP.

Data Sources:

We used online databases (e.g., MEDLINE^®) and bibliographies of retrieved articles for the literature search and government and foundation listings to identify ongoing research. For the simulations, we used data from public reports of myocardial infarction outcomes in California.

Study Selection:

For the literature review, we sought studies in which providers had been randomized to an incentive group or a control group. We included only projects involving interventions purchasers could plausibly adopt (payment strategies or public reporting of performance). Studies of interventions that were beyond purchaser purview (e.g., implementing clinical guidelines) were excluded.

Data Extraction:

We extracted information about the type of incentive used and the clinical and economic context in which it was applied.

Data Synthesis:

We evaluated 5,045 publications. Nine were randomized controlled trials, and many of these did not report key characteristics of the incentive or the context in which incentives were applied. Incentives used included additional fee-for-service, quality bonuses, and public release of performance data. The results were mixed: among the 11 performance indicators evaluated, 7 showed a statistically significant response to QBP strategies while 4 did not. We also found 18 ongoing research projects, none randomized. These will yield data about the approaches to QBP currently in use, provider awareness of and concerns about QBP, and some preliminary estimates of the potential impact of QBP.

Regarding assessments of outcomes reports, we found that, under reasonable assumptions and applications, outcomes reports generate meaningful information about provider performance. Providers with good (expected) performance are unlikely to be labeled as poor quality in any given period, and very unlikely to be mislabeled more than once in a 3-year period, even if one allowed approximately 10% of hospitals to be labeled poor performers annually. In addition, hospitals with superior performance were quite likely to be identified as such at least once in 3 years.

Conclusions:

Little is known about the impact of QBP on clinical performance. However, it does appear that basing incentives on measurements of outcomes is feasible without undue risk to the reputation or financial status of good hospitals. Ongoing research will only address some of the gaps in our knowledge about QBP, suggesting that much more additional research is needed. This should include comparisons of alternative QBP approaches and qualitative assessment of the barriers to and facilitators of quality improvement in response to QBP incentives.