A new study finds high death rates, repeat hospital stays and soaring medical bills among Medicare patients who have mechanical pumps placed in their chests to help their failing hearts.

The devices have great promise but are being used too often in the wrong patients and at the wrong hospitals — in people who are too sick to benefit, and at hospitals that do not treat enough patients to gain the expertise needed for their complex care — the researchers say in an article being published on Wednesday in The Journal of the American Medical Association.

The pumps, known as ventricular assist devices, or VADs, were introduced in the 1990s and are used for several reasons: to keep patients alive until they can receive heart transplants, to prolong life and relieve symptoms in those who are not transplant candidates and to rescue people whose hearts fail after cardiac surgery. About 1,000 to 2,000 people a year receive VADs, which vary in complexity and cost $11,000 to $75,000.

The underlying illness that the devices treat is heart failure, in which the heart muscle is too damaged or diseased to pump enough blood. The condition affects nearly five million people in the United States and contributes to about 287,000 deaths a year. Its main causes include heart attacks, high blood pressure and coronary artery disease.

The researchers, from Duke University, examined the records of 2,943 Medicare patients who received the devices at 570 hospitals from February 2000 to June 2006. About half, with an average age of 63, needed the pumps to treat chronic heart failure. The other half, age 69 on average, were given VADs because their hearts failed after coronary bypass or valve surgery.

Over all, 51.6 percent of the chronic patients but only 30.8 percent of the postsurgical ones were still alive a year later. Over the course of the study, the one-year survival rate in the postsurgical group actually became worse instead of better, possibly, the authors said, because the pumps may have been implanted in sicker and sicker people. Many of the surgery patients never even made it out of the hospital.

In both groups, among people discharged with a VAD, about half wound up back in the hospital within six months. At one year, the average Medicare payment for inpatient care for the chronic patients who received pumps was $178,714, and for the postsurgery cases $111,769. (The surgical group had lower costs because on average the patients did not live as long.) Those figures do not include doctors’ fees or costs for outpatient services like physical therapy.

People with heart failure have high death rates to begin with, so VADs are hard to evaluate, the researchers said. But they noted that so far, the devices had not achieved a goal set in 2004 by the National Heart, Lung, and Blood Institute: 50 percent survival at two years and minimal time in the hospital.

“This technology was developed and released in anticipation of it getting better over time as physicians had more experience with it,” said one of the study’s authors, Dr. Kevin A. Schulman, an internist and professor of medicine and business administration at Duke. “It’s very clear that hasn’t been the case. We also see there are even worse outcomes than average at centers with very low volume.”

Another author, Dr. Adrian F. Hernandez, a Duke cardiologist, said: “Certainly there are patients who do quite well. We have other examples, people who go on to see their grandchildren getting older and have a quality of life like what they had before heart failure. Or it can become the last leg of life support, with the respirator and the dialysis machine.”

The findings mean that doctors need to pick their patients more carefully and try to use the pumps only in those with decent odds of surviving and doing well, the authors said. They also suggested that whenever possible, use of the devices should be limited to “centers of excellence”: hospitals with good track records and a steady flow of cases.

Government grants paid for this study, though the authors said they accepted other research money from various drug and device makers, including some that make VADs.

Robin Bostic, a vice president at Thoratec, a leading maker of VADs, said the study lumped together different kinds of patients and devices, lacked information about why patients were rehospitalized and did not provide enough specific information to help doctors make treatment decisions.

A heart surgeon not involved in the study, Dr. Mehmet C. Oz, vice chairman of surgery at NewYork-Presbyterian Hospital/Columbia, said an important message from the study was that chronic heart-failure patients fared much better with VADs than did those who were given the devices in desperate efforts to rescue them from failed operations.

In addition, Dr. Oz said, the research makes clear that “a small center who’s dabbling in the field should not be allowed to put VADs in.”

He also said the study was valuable because it was conducted by university researchers using Medicare data, rather than by industry, which is often the source of information about devices.

But he emphasized that VADs had great promise and said that continuing research would improve them and make them an increasingly important option for people with heart failure.