In this issue of Annals [1], Topol and Califf endorse a fundamental clinical change involving the release of physician-specific information about performance. In particular, do the patients of physician A do better, after controlling for sickness, than those of physician B? Although proposals for public release of information at a physician-specific level are not new [2-6], this article is important because it comes from physicians in cardiology departments of two prestigious institutions and because it is about releasing information on the outcome of cardiovascular procedures. Until now, calls for public release of information about physician performance have usually emanated from business, government, or physicians who would like to see physicians in other specialties release their data. Does this article represent a paradigm shift? Does it represent the realization that the future U.S. health care system will involve competition about quality at a physician-specific level? If so, what are the likely consequences?

Let us consider two different worlds. World 1 represents the present system in which physician-specific performance data are not released, and world 2 represents the release of such data. In world 1, as cost containment occurs, the growth in the use of cardiovascular procedures will slow down. Because only information about price is available, physicians who charge less receive more business. Physicians who do procedures with higher quality of care or who achieve better outcomes will not be better off than those who do not. By better off, I mean having a greater likelihood of staying in business, increasing their market share, or providing more cardiovascular care than their less competent colleagues. Quality may go down as all physicians cut corners to increase their market share. Maintenance of our present system is likely to have two negative results and one positive result: First, variations in outcome by physicians are likely to remain large. For coronary artery bypass surgery, this might mean 5 additional deaths per 100 patients who received surgery [7]. Second, the professional survival of an individual physician is unlikely to depend on achieving higher quality. Third, because quality will not be an area in which competition occurs, cooperation among physicians to advance clinical knowledge will probably not decrease.

In world 2, physician-specific performance data would be used by the public, hospital administrators, directors of health plans, insurance company executives, and the government. Patients would probably select those physicians who obtain better than expected outcomes for their patients. Patients may not pay much attention to whether differences among physicians are statistically significant at the conventional level of P < 0.05. In fact, a P value of 0.1, 0.2, 0.3, or even higher, still implies that, on average, the patient would be better off going to the higher doctor on the quality list than the next lower one on the list. What is important to the patient is whether these differences are clinically significant. In this regard, we need to develop understandable and scientifically valid ways to make sure that the data we present to patients are not misleading. For example, case-mix adjusted analyses that show that last year Dr. A had 5 fewer deaths per 100 patients than did Dr. B do not mean that the difference will be the same in the future. On average, it most likely would be smaller [8].

Release of data on quality to the public would put enormous pressure on physicians. Physicians who can do a procedure competently might lose their ability to practice in a region because they are average. For instance, if a town happened to have five cardiovascular surgeons, all of whom were above average, then the last physician on the quality list might be out of work. Such a physician might have to move.

If the measures of quality used in world 2 had some validity, then their release would improve and reduce variation in outcomes by physicians. For these events to occur, the quality measures need not be perfect. We do not need to know, for example, how to measure severity of illness perfectly. However, from a physician's viewpoint, the likelihood that he or she will be unjustifiably discriminated against is related to the validity of the quality measurement system. Physicians whose patients were sicker on average but whose levels of sickness were not incorporated into the severity measure might be unjustifiably forced out of business. In addition, poor outcomes are not the sole responsibility of the physician. Physicians who are unlucky because they selected an organization that provides little support might be similarly tarnished and find it difficult to practice.

Thus, although the public is likely to benefit from world 2, some physicians might undergo considerable (and sometimes unjust) hardship. Clinical medicine is facing a major choice about how medicine will be practiced in the next century. Public release of information about quality at the physician level will dramatically change the way we practice. It may have a greater effect on the profession than economic reforms from Washington. If we are going to use this method to decrease variations in quality and to improve mean levels of quality of care, then we must examine carefully each step we make.

In addition, virtually all outcome comparisons of physicians use mortality as their outcome measure. Almost all valid clinical models that analyze what physicians do to patients (process) have used mortality as the outcome variable. Even though death is an important outcome, so is quality of life; outcome comparisons among physicians must eventually include comparisons on quality-of-life effects as well. This is especially important in the ambulatory area where death is an infrequent outcome. Finally, in this cost-conscious world, competing on broad measures of quality of life may result in the need to expend more rather than less on medical care. Physicians are being asked to spend less time, not more, with patients; however, improving a patient's sexual, social, physical, and mental well-being is certainly not consistent with such a policy.

The major technical question about the release of outcome data is: How much clinical data needs to be collected to increase the validity of the outcome comparisons to a level that the amount of misinformation is acceptable to physicians? A related question is: How much money should be invested to improve the validity of the publicly released information in order to make sure that mistakes in ranking physicians are decreased to an acceptable level? Hannan and colleagues [9, 10] in New York State have shown that adding clinical data changes the mortality rankings of some hospitals for coronary artery bypass surgery. However, the increase in validity may not be sufficient to warrant the expense of collecting the clinical data. We need an analysis describing the cost of producing data for the public about physicians and hospitals at an acceptable level of validity and describing how the cost varies by different levels of validity. This information might lead to a compromise between the public and physicians about the public's willingness to spend money on a valid data system and the physician's acceptance of the data as accurate enough to identify which physicians were producing higher quality of care. Can we have the best of both worlds?