constitute a critical safety net. The same market forces that will operate to improve or eliminate the cohort of providers who perform poorly may leave populations subject to disparities in care with fewer provider options than they had before. Pay-for-performance programs must therefore be carefully designed to identify relationships that exist between populations subject to disparate care and poorly performing providers. Objective assessment will help limit cultural bias in performance measurement.

The application of performance measures in the evaluation of health care for a particular condition (e.g., diabetes mellitus) or preventive service (e.g., breast cancer screening) poses the risk of decreasing performance and thereby compromising the quality of care being provided in areas that are not the focus of pay for performance. Pay for performance could encourage this tendency to manage to the measures, focusing efforts excessively on those measures that yield the greatest financial return. At the same time, however, this could be beneficial by focusing efforts on areas with the greatest need for improvement, such as the treatment of chronic diseases.

Additionally, measures may conflict with one another, ultimately causing harm to patients. This concern reinforces the need, articulated in the Performance Measurement report (IOM, 2006b), to develop a comprehensive set of performance measures as rapidly as possible. The present report articulates the need for measures that reward three key domains of care: clinical quality, patient-centeredness, and efficiency. As noted earlier, a single-minded focus on clinical quality can lead to increased health care costs through overuse of services. A similar narrow focus on efficiency could compromise clinical quality and raise at least the appearance of a fundamental conflict of interest. And performance measures that place undue emphasis on clinical quality or efficiency are unlikely to be patient-centered.

A comprehensive portfolio of performance measures must reflect consensus around the vision of a reformed and integrated health care system designed to achieve the goals articulated in the Quality Chasm report (IOM, 2001). For example, prompt, understandable, and empathetic communication to the patient of the results of a magnetic resonance scan is as important as the technical quality and value of the imaging study itself; ideally, financial incentives should be restructured, based on valid and robust measures of performance, to encourage both.

In the health care sector, best practices are adopted at a surprisingly and disconcertingly slow rate (Lomas et al., 1993; Bates et al., 2003). While