Pay for performance in the intensive care unit—Opportunity or threat?*

Abstract

Ongoing evidence of poor-quality healthcare has stimulated the development of provider reimbursement schemes linked to the delivery of high-quality care. Our objective was to describe these programs and their potential implementation in intensive care units (ICUs). MEDLINE (2000-May, 2008) and personal files. We selected empirical studies, narrative and systematic reviews, and commentaries addressing pay-for-performance programs. Using a narrative review format, we discuss the definition of pay-for-performance, describe current implementations, suggest challenges of applying these programs to the ICU setting, and discuss alternative quality improvement programs. The ICU will likely become a target for pay-for-performance plans, considering the high cost of care, development of ICU quality-of-care measures, and interest from healthcare regulators and funders. Existing plans applied outside the ICU have varied in the amount of financial incentive and targeted provider and quality measures. Evaluations are sparse. Implementation challenges specific to the ICU include selecting evidence-based and feasible quality of care measures, motivating the entire interdisciplinary team, integrating multifaceted behavior change strategies, and developing informatics infrastructure for timely audit and feedback. Other incentive-based alternatives to improve ICU quality of care include a "centers of excellence" approach (referral of patients to centers with excellent outcomes), public reporting of ICU outcomes, and payments to hospitals for participating in quality improvement programs. Participation in pay-for-performance programs is a potential opportunity for intensivists and ICU teams to improve outcomes for their patients in partnership with regulatory agencies and healthcare funders. Because many aspects of optimal design of these programs in ICUs are unknown, robust evaluations of their effect on healthcare quality should be integrated into any implementations.