IOM Report: Paying for Performance in Healthcare. Casting the Die While Tinkering With a Broken System, by David C. Aron, MD, MS, Cleveland, Ohio: Louis Stokes Cleveland Veterans Affairs Medical Center, and Case Western Reserve University School of Medicine.

When Julius Caesar crossed the Rubicon with his army, a civil war was inevitable. In uttering the famous words "iacta alea est" (the die is cast), he recognized that while the outcome was not certain, there was no turning back; the step was irrevocable. The Center for Medicare and Medicaid Services has introduced pay for performance (P4P) at the physician level, and although the outcomes are not certain, there is no turning back. The Institute of Medicine (IOM) report Rewarding Provider Performance: Aligning Incentives in Medicare, a remarkably comprehensive document, provides the background and context for this effort and makes a series of recommendations to guide implementation.1 This report, the third in the IOM's Pathways to Quality Health Care series, focuses on the level of the individual provider. P4P was not viewed simply as a mechanism to reduce costs or to reward those who perform well, but rather as a means to align incentives to encourage the most rapid feasible performance improvement by all providers, especially through coordination of care across provider settings and time. The panel came to 2 conclusions: (1) the systematic and deliberate use of payment incentives that recognize and reward high levels of quality and quality improvement can serve as a powerful stimulus to drive institutional and provider behavior toward better quality and (2) the incentives introduced by P4P, by themselves, will not be sufficient to achieve the broad institutional and behavioral changes needed unless certain operating conditions are met, such as the use of electronic health records, public reporting, beneficiary incentives[horizontal ellipsis]. In reviewing these conclusions, I will address several questions: (1) is P4P necessary? (2) is P4P possible? and (3) is P4P possible when it is necessary or necessary if it is possible? Setting my cards on the table, I come to these questions from the standpoint of a physician in the Department of Veterans Affairs, a "closed" system to be sure, and one who has witnessed and participated in one of the most extraordinary healthcare system transformations, a transformation that was accomplished without, until this year, provider-level P4P.

Is P4P necessary? That is, does quality need to be improved and can P4P produce improvement?

There is no shortage of evidence documenting serious gaps in the quality of medical care. Deficiencies are present in every aspect of care-inpatient and outpatient, preventive and curative-and in the coordination of care among providers and healthcare settings. Moreover, there is marked variation in quality. The American healthcare system is also fabulously expensive, consuming about 15% of the gross domestic product. The combination of rising costs and increasing visibility of quality deficits has prompted both private and public sector initiatives, many organized around the desire to increase the value of the services paid for and the accountability of those who provide the services. Medicare's current payment system involving fee for service makes little distinction between care of high or low clinical quality or between appropriate care and inappropriate care, especially overuse of high-cost (high reimbursement) procedures. The current system also values procedures over purely cognitive activities and provides little incentive to promote coordinated care, preventive care, or even primary care in general. Since "every system is perfectly designed to get the results it gets," we should hardly be surprised. If money drives the current state, it makes intuitive sense that realigning incentives is an appropriate strategy. Medicare's mandate to introduce P4P reflects the dual pressures of quality and cost and a fundamental faith in the power of money to change behavior. Of course, whether P4P is the cure depends on the etiology of the disease of poor quality/high cost. In fact, remarkably, the evidence to support the concept of P4P is quite weak. First, although there are many P4P programs (well more than 100 in the private sector), the number of studies addressing effectiveness is relatively small, the results are mixed, and the study designs are not sufficiently rigorous to confidently attribute any observed improvement to P4P.

Is P4P possible? The hypothesis that money will change behavior depends on factors such as physician buy-in and the ability to accomplish what is being asked. Perhaps the major factor in the VA's transformation was the presence of a robust electronic medical record system that facilitated both improvement and data collection. If underdeveloped information systems are a root cause, the financial incentives of P4P will have to take into account the business case as a whole. Will physicians perceive the use of financial incentives as inconsistent with their sense of professionalism? Will physicians perceive P4P as a way to increase efficiency rather than quality and to reduce their reimbursement since long-term P4P is designed to be revenue neutral; there will be winners and losers. The Panel was quite cognizant of this and outlined a series of design principles for P4P in its implementation. Some of the principles relate to the goals of P4P, for example, reward care that is of high clinical quality, patient-centered, and efficient; reward significant provider improvement as well as achievement of excellence; foster care coordination among providers; and reward data collection and reporting functions and encourage adoption of improved information technologies. These principles can be accomplished only if another principle is met: use performance measures that reliably define good care and optimal health outcomes and that they are measured in a way that reliably and fairly identifies different levels of quality at the provider level.

The comment of Albert Einstein that not everything that counts can be counted is honored in the breech rather than in the observance. Ease of data collection is a major criterion for choosing performance measures. These technical measures of quality, sometimes measures of process and sometimes measures of outcomes, are but a small part of clinical practice. Moreover, their discrete and typically cross-sectional nature assumes a strong link between process measures and outcomes and can also readily contribute to the fragmentation of care. A recent study conducted by Bradley and colleagues found that compliance with P4P-styled treatment strategies for acute myocardial infarction (ie, aspirin, angiotensin-converting enzyme inhibitors, [beta]-blockers) accounted only for a 6% variation in 30-day mortality.2 Six percent variation is not zero, nor does it necessary mean that there will not be a larger variation at longer periods of follow-up. However, it does raise concerns. Threshold measures bring their own set of limitations. For example, there is currently controversy about the use of a glycosylated hemoglobin value (A1c) less than 7% to identify good control in patients with diabetes. Such a level might be inappropriate in patients with multiple comorbidities, risk of hypoglycemia, or short life expectancy.3

How the measures are taken is also not trivial and is even less so when focusing on individual physicians. The technical, administrative, and financial challenges and burdens, which even in a relatively closed system like VA were substantial, would be magnified in smaller practices that lack the resources of large systems. Even rewards large enough to meet a business case could divert time and resources away from patient care. Putting these challenges aside, there remain such issues as sampling. The reliability of sampling techniques, when there are small numbers of patients with particular diseases receiving care from individual physicians, has been well demonstrated.4,5 Performance measures must be able to take into account different patient characteristics, that is, risk adjustment, or gaming the system by targeting patients in whom meeting the performance measures would be easy could reduce access for those deemed more difficult. In fact, the Panel should be complimented on its attention to potential unintended consequences and all parties should pay attention to Table 2-3 on page 38, which outlines them in some detail. However, while a set of P4P measures has been proposed, a comprehensive evaluation framework to detect these and other potential unintended consequences has not. In fairness, such a system was recommended, but it has not been developed. Since P4P also involves the use of potential public humiliation, it will be critical as pointed out by the Panel that provider achievement must be reported in ways that are both meaningful and understandable to consumers. Given all the uncertainties, the Panel wisely made its first recommendation that the secretary of the Department of Health and Human Services should implement P4P in Medicare, using a phased approach as a stimulus to foster comprehensive and systemwide improvements in the quality of health care. This brings us to the question about whether P4P is necessary when it is possible and a closer look at the experience of the VA is in order.

The achievements of the Veterans Health Administration in the Department of Veterans Affairs, an agency once solely the object of criticism, have become an example of systemwide improvement, cited both here and abroad. Comparisons of similar indicators of quality in the VA and Medicare fee-for-service systems and managed care organizations have, in general, shown the VA to be superior for these measures.6-8 The transformation began in 1995 and the efforts involved better use of information technology, measurement and reporting of performance, and integration of services and realigned payment policies. These efforts included determined leadership, focused interventions to improve care and programs to advance implementation research, and incentives to improve care, but interestingly, P4P at the provider level was not implemented until recently. The key concept in the recommendation is the term "systemwide." A major issue in quality improvement systemwide in settings other than VA is the lack of a coherent system that links health care organizations in the United States. Given the wide range of interests of the involved parties (eg, private practitioners, large physician groups, academic institutions, health insurers, and chief executive officers of those insurers, among others), it is difficult to envision how the proposed plan for P4P will work and how it will avoid the myriad of unintended consequences. Whether the lessons of the VA can be applied constructively and effectively to a "nonsystem" remains to be seen. Until the broader issues of health care delivery are addressed, we are tinkering with a broken system (or nonsystem). The Panel does include appropriate cautionary notes, but one wonders whether this is another example of the trenchant observation of Winston Churchill: "You can always count on Americans to do the right thing-after they've tried everything else." After reading this IOM report, one must conclude that we are still in the mode of trying everything else.

David Aron, MD, MS

Education Office (14W), Louis Stokes Cleveland VA, Medical Center, 10701 East Blvd, Cleveland, OH 44106. E-mail: [email protected]

REFERENCES