Dr. Brook's editorial [1] and the accompanying article advocate the increased public release of patient outcome data to monitor quality of care. As noted, such a system of scorecard medicine would supposedly answer the question "Do the patients of physician A do better, after controlling for sickness, than those of physician B?"

I immediately thought of five patients in the past week who would probably score low in the "sickness" category but who epidemiologically are at high risk for early death. All are middle-aged men and women with a strong family history of coronary artery disease, hypertension, and non-insulin-dependent diabetes mellitus with evidence of microvascular or macrovascular complications. All have truncal obesity, are sedentary, and smoke. None were able to stay on a diet or exercise regularly despite encouragement by several health care providers. These patients show that overall "sickness" may largely be determined by behavior-related factors.

The measurement of behaviorally related health factors is extremely difficult. In self-reports, patients tend to underestimate their food [3] and alcohol intake [4]. Self-reported exercise activity correlates poorly with objective measures of physical fitness, particularly in sedentary persons [3, 5]. These inaccuracies may be amplified if the patient suspects that his or her responses will be used to measure health risk.

Dr. Brook correctly states that the use of physician-specific outcome data would radically change how we practice medicine. Based on his system, I would assess each patient's risk. If it differed dramatically from the "sickness" scale that he proposes, I would consider asking the patient to seek care elsewhere. This is not the patient advocacy we were trained to practice. Should a cardiologist refuse angioplasty to a patient who continues to smoke? Should a psychiatrist refuse to care for a depressed patient who is at high risk for suicide?

Drs. Topol and Califf [2] cited the case of several cardiac surgeons refusing to operate on a high-risk patient so that their scorecards remained unblemished. They also suggested that surgeons might "game" the system by entering an excess of high-risk patient data. Physicians may find themselves expending as much effort protecting their scorecards as caring for patients. This is separate from the logistic problem that the responsibility for updates would fall on the physician's office staff, already overburdened with paperwork. Brook states that the measures of quality care need not be perfect. I submit that, unless they are extremely accurate, they may affect health care in unanticipated ways. Measuring patient outcomes has great merit; improving patient outcomes is the ultimate reason for the existence of our profession. However, its current practical application seems plagued with difficulties.

I agree that to restore public confidence, we do need to work harder in policing ourselves. Most medical specialties currently require periodic retesting to maintain certification. This should assure minimum competency in terms of knowledge. Although not perfect, a process-based system evaluating conformance to established standards of care is at least manageable from a logistic standpoint and does not penalize physicians who care for difficult patients.