When can a doctor refuse to treat a patient? Not surprisingly, there are rules. In 1987, Dr. John Bower, a kidney specialist at the University of Mississippi Medical Center, was sued after dismissing from his practice a patient who regularly missed dialysis appointments, verbally abused nurses and even threatened to kill Bower and a hospital administrator. Bower cited medical noncompliance and violent threats as grounds for terminating care. The Fifth Circuit Court of Appeals, in New Orleans, agreed with him, ruling that doctors can refuse to treat violent or intransigent patients as long as they give proper notice so that the patient can find alternative care. Forcing doctors to treat such patients, the court said, would violate the 13th Amendment, which prohibits involuntary servitude.

Sometimes doctors can refuse to treat on the basis of conscience. I once took care of a man in his 50's who had metastatic tongue cancer and respiratory failure requiring a ventilator. His family refused to turn off the machine and let him die, choosing instead to escalate treatment with even more aggressive interventions. Medicine is a stochastic science -- no doctor can predict the future -- but in this case the outcome was never in doubt. Advanced cancer patients who end up on ventilators die during their hospitalizations. Life support was futile, and continuing to insert catheters and tubes into this man seemed inhumane. After the attending physician and I consulted with the hospital's ethics committee, we told the family that we could no longer obey their wishes. We gave them the option of transferring the patient to another hospital. They didn't want to do that; treatment was scaled back, and the man died a few days later.

In rare cases, doctors have objected to treatment on moral grounds. In February 2001, it was reported that surgeons in Melbourne, Australia, were refusing to provide heart and lung transplants, coronary-bypass surgery and lung-reduction surgery to smokers. At one hospital, surgeons insisted that patients be free of nicotine, alcohol and other drugs for at least six months before going into the operating room. ''Why should taxpayers pay for it?'' one surgeon asked. ''It is consuming resources for someone who is contributing to their own demise.'' The policy predictably drew an outcry from human rights organizations and many doctors. The Australian Medical Association said it was ''unconscionable'' for a surgeon to take a moral stand on treating a patient. Ethicists said it was discriminatory because lifestyle could be blamed for many illnesses, including obesity and certain types of cancer.

In the last decade, there has emerged a new, unexpected impetus for denying treatment to patients. Consider the case of P., a jovial, heavyset man in his 50's with a weakness for chocolate and cheesecake. In his mid-30's, P. had a heart infection that required surgery to replace a damaged valve with one from a pig. About 10 years later, his kidneys failed for unknown reasons, and he had to start getting dialysis treatments. A religious man, he took it all in stride as the will of God.

Now a cardiac catheterization showed severe limitations in blood flow to several parts of his heart. P.'s coronary arteries looked like sausage links, sectioned off by numerous tight blockages. During the catheterization, fluid filled his lungs, and he had to be treated with intravenous morphine and nitroglycerin and emergency dialysis. Once his condition stabilized, he was scheduled for coronary-bypass surgery.

In the hospital, P. continued to have severe chest pains -- small heart attacks, really -- that were inadequately relieved by medications. One morning, he nearly doubled over while washing up. Because of his medical conditions, a surgeon decided that P. was a ''high risk'' surgical candidate and should be transferred to an affiliated hospital to have his operation.

The surgeon said that agencies monitoring surgical outcomes had been putting tremendous pressure on surgeons to produce good results. He was referring to ''report cards'' on cardiac surgeons. Over the past decade, while surgeons with higher-than-expected mortality statistics have lost operating privileges, others with lower-than-predicted rates have taken to advertising on the radio. Because the surgeon and his colleagues had been aggressive about treating very sick patients like P., they had incurred higher mortality rates and had been penalized by the state department of health. This was an insult he could no longer countenance, and so he had decided that all high-risk surgeries would be transferred to another hospital.

And so P. was transferred. A few days later, after his operation, he died.

When I was in medical school, internal-medicine residents used to joke about never calling a surgeon. Their message, however facetious, was that if you sent your patient to a surgeon, he would want to operate. That, after all, was what surgeons did -- and the more challenging the case, the better. Any second-guessing about surgery was supposed to be done by the referring physician.

Times are changing. In a survey a few years ago, 63 percent of cardiac surgeons in New York State said that because of report cards, they were accepting only relatively healthy patients for coronary-bypass surgery. Fifty-nine percent of cardiologists confirmed that it had become harder to find a surgeon to operate on their most severely ill patients. ''Hospitals getting cited are turning down high-risk patients,'' Dr. Eugene Grossi, a professor of surgery at N.Y.U. School of Medicine, recently told me. ''Some of the so-called best hospitals are only doing the most straightforward cases.''

How did such diffidence, antithetical to the surgical ethic, arise in the pursuit of quality improvement? When did surgeons start becoming afraid of sick people? To answer this, it is necessary to go back to the late 80's, the dawn of what Dr. Arnold Relman, former editor of The New England Journal of Medicine, called ''the era of assessment and accountability'' in medicine. It was a period when concepts like ''total-quality management'' and ''continuous-quality improvement'' (modeled after the Japanese management strategy of kaizen) were diffusing throughout the corporate world. People decided that it was time to start applying these concepts to medicine. Cardiac surgery was chosen as the prototype for this ''outcomes research'' because the data are plentiful -- more than 500,000 coronary-bypass operations are performed each year -- and patient follow-up is good.

Several programs were begun during this period. The Health Care Financing Administration, a federal agency, and several states developed systems to keep track of patient outcomes. In the early 90's, New York and Pennsylvania started publishing mortality statistics on hospitals and surgeons performing coronary-bypass surgery. (Although this data had been collected for years, it had never been released to the public.) Hospitals at first submitted pseudonyms, until the states forced them to use surgeons' real names. Soon the statistics appeared in newspapers and on the Internet.

The purpose of these report cards was to improve the quality of cardiac surgery by pointing out deficiencies in hospitals and surgeons. The idea was that surgeons who did not measure up would be forced to improve. Patients, it was argued, had the right to know which was the ''best hospital'' and who was the ''best surgeon'' before deciding where to seek medical care. As consumers, they were entitled to know what they were getting for their money. (Surgical report cards were just the tip of the iceberg of this ''consumer movement,'' as performance reports on nurses and health-care plans soon appeared, and ''best doctors'' lists started proliferating in books and magazines.)

Of course, mortality rates are affected by myriad factors, so models were created to predict surgical risks and avoid penalizing surgeons who took on the most difficult cases. For example, a 50-year-old man, otherwise healthy, who underwent coronary-bypass surgery was judged to have a mortality risk of about 1 percent. For a 70-year-old on intravenous nitroglycerin with a history of heart surgery, congestive heart failure, emphysema and other medical problems, the risk was about 20 percent or higher.

Many surgeons, however, felt that the models underestimated surgical risk, particularly for the sickest patients. They criticized the models for oversimplifying heart surgery. They argued that the models did not have adequate predictive ability. Among other things, they said, the models did not account for simple bad luck.

Surgeons began to fastidiously report -- some say overreport -- medical conditions that could affect the outcome of surgery. Surgeons started coding for everything, Grossi told me. In some New York hospitals, the prevalence in surgical patients of emphysema, a condition known to increase surgical risk, increased from a few percent to more than 50 percent after report cards came into use. Some surgeons, according to one physician, even made a habit of routinely putting patients on intravenous nitroglycerin because being on the drug conferred added risk. Others tried to ''hide'' surgical deaths by transferring patients to hospice programs right before they died.

Despite these excesses, in the beginning there were high hopes for quality-improvement programs. In the first few years, there were major gains in cardiac surgical outcomes, at least according to some studies. Veterans Affairs hospitals saw their mortality rates drop 14 percent below what their model predicted. Mortality declined by 24 percent for a consortium of New England hospitals with their own quality-improvement programs. The most striking results were in New York State, where mortality rates for coronary-bypass surgery declined a whopping 41 percent and outcomes improved for all hospitals at all levels. In a 1994 article in Annals of Internal Medicine, the cardiologists Eric Topol and Robert Califf wrote that ''appropriate implementation of score cards could ultimately lead to a substantial improvement in the quality of U.S. cardiovascular medicine.''

But not everyone believed that report cards were causing real improvements in care. Some experts said that the better outcomes simply reflected nationwide trends. In Massachusetts, for example, cardiac surgical mortality declined by about 40 percent even without a statewide program for reporting outcomes. Others entertained a more disturbing possibility. Were surgeons' numbers improving because of better performance, or was it because patients like P. were not getting the operations they needed?

Last January, researchers at Northwestern and Stanford tried to answer this question. Using Medicare data, they studied all elderly patients in the United States who had heart attacks or had coronary-bypass surgery in 1987 (before report cards were used) and 1994 (after they took effect). They compared New York and Pennsylvania, states with mandatory surgical report cards, with the rest of the country.

What they discovered was that there was a significant amount of ''cherry picking'' in the states with mandatory report cards. For example, they found that between 1990 and 1993, patient health-care expenditures over the year before coronary-bypass surgery dropped by 7 percent in New York and Pennsylvania but stayed roughly the same in the rest of the country, suggesting that coronary-bypass operations started being performed on healthier patients. This was matched by a drop in the number of operations for sicker patients, who experienced ''dramatically worsened health outcomes'' as a result, including more congestive heart failure and recurrent heart attacks.

The researchers concluded that surgical report cards in New York and Pennsylvania led to substantial selection bias by surgeons and that patients generally were worse off for it.

''Mandatory reporting mechanisms,'' they wrote, ''inevitably give providers the incentive to decline to treat more difficult and complicated patients.'' They questioned whether report cards had improved surgical performance. ''We offer a different explanation,'' they wrote. ''Observed mortality declined as a result of a shift in incidence of surgeries toward healthier patients, not because report cards improved the outcomes of care for individuals with heart disease.''

he incentive to refuse treatment for high-risk patients has created a kind of spiritual crisis in the field of cardiac surgery. Heart surgeons, among the most highly trained and fearless of specialists, are shrinking from taking on the toughest cases because of statistics.

Not long ago, M. was hospitalized with fever and difficulty walking. He had been living with relatives in Staten Island before his intravenous-drug habit landed him in jail. He was a frail man in his 50's with white hair and piercing narrow-set eyes, one of which had stopped moving normally. Doctors thought he had had a stroke, eventually confirmed by various brain scans, but they didn't know why. An echocardiogram, an ultrasound of the heart, provided the answer. Sitting on top of one of his heart valves was an infected mass of tissue -- a vegetation -- that was flapping around wildly, like a flag in the breeze. A portion of it probably had broken off and gone to his brain.

The echocardiogram also revealed an abscess, a collection of pus, around the valve. Few conditions in medicine absolutely require surgery, but this is one of them. Antibiotics cannot penetrate into an abscess because there are no blood vessels to deliver the drugs. Left untreated, it is bound to grow, eating away at surrounding tissue. In fact, the abscess had so weakened a portion of M.'s heart that that part of his heart had turned into a thin-walled aneurysm, bulging out with every heartbeat. If the aneurysm ruptured, M. would almost certainly go into congestive heart failure and die.

The surgeon thought that M. was probably too sick for surgery. It was going to be a complex operation, involving excision of the infected valve and a portion of the aorta, the major blood vessel emanating from the heart, and replacement with tissue grafts. Besides being a diabetic and a ''re-op,'' M. had had a stroke and a recent bout of congestive heart failure. All that added up to a surgical mortality risk of about 50 percent. The surgeon talked about report cards and admitted that there were pressures to avoid the case. ''I don't know whether it's ethical or unethical,'' he said, ''but in this day and age we are not rewarded for taking care of sick patients.''

It isn't just heart surgeons who feel this pressure. Similar pressures are being brought to bear in other areas of medicine. For example, there is evidence that report cards on interventional cardiologists have resulted in a drop in the number of angioplasty procedures performed on very sick patients. This emphasis on statistics, not patients, has chagrined many doctors. One surgeon put it to me this way: ''Anything that makes you turn down a patient because surgery is too risky, that's a good thing. But turning someone down because they make your numbers look bad -- and it's the patient's last shot -- well, that's a bad thing, and that kind of thing is happening more and more.''

Surgical report cards were supposed to embolden surgeons and protect patients by improving the quality of cardiac surgery. They appear to have failed on both counts. In fact, there is little evidence that the public pays much attention to outcomes data. ''The U.S. experience with public disclosure is in many ways paradoxical,'' researchers wrote in the journal Quality in Health Care in 2001. ''Consumers express a desire for more information about quality while, at the same time, few are able to understand or use it.'' Surgeons, it would appear, are the only ones paying attention to their report cards.

What then should be done to improve health care? A 1999 Institute of Medicine report estimated that up to 98,000 deaths occur in the United States every year because of medical errors. Most of these deaths, the report said, do not occur because of individual mistakes but because of flaws in the way hospitals and clinics operate. In the recent case of the 17-year-old Mexican girl who died at Duke University Hospital after being transplanted with a heart and lungs from a donor with an incompatible blood type, the fault was not in any one individual but rather in the structure of care: not enough people checking to make sure the blood types matched.

Perhaps the most far-reaching lesson of surgical report cards is that scrutinizing and punishing individual doctors is not the right way to improve health care. Health care is too complex; outcomes depend on many variables. Everyone, of course, wants a system with accountability -- people want to be able to point a finger at someone -- but such a system often ignores the anatomy of most medical mistakes. Surgery, for example, is a team sport, involving referring physicians, technicians, nurses, anesthesiologists and surgeons. Focusing on the performance of one piece of this structure ignores the larger process at play.

In the case of M., however, one doctor did make a difference. A few days later, a cardiologist made a personal appeal on M.'s behalf to a surgeon. Unexpectedly, he decided that M. would get his operation after all, once appropriate tests were completed to evaluate his surgical risk. M., for his part, said he would stop using drugs. Observation can change behavior; sometimes it just depends on who is looking.

Sandeep Jauhar, M.D., a cardiology fellow in New York City, writes frequently about medicine for The Times.