Johns Hopkins Hospital in Baltimore and Mt. Sinai Hospital in New York are among the nation’s elite hospitals. But when it comes to how they practice medicine, the doctors at the two institutions are very, very different.

At Johns Hopkins, nearly half of patients enroll in hospice in the last six months of life. That number stands at one-quarter at Mt. Sinai. At Hopkins, one-third of patient deaths will occur in the hospital (others are likely at home or in hospice). At Mt. Sinai, 44 percent of patient deaths happen in a hospital bed.

“If we look at patients with chronic illness, there have been any number of studies that find most people don’t want to die in a hospital, not more than 15 percent,” said David Goodman, director of the Center for Health Policy Research at the Dartmouth Institute for Health Policy and Clinical Practice.

“The preferences are pretty clear. It’s hard to find a place where the levels of people dying in the hospital are similar to what we know about general preferences,” he said.

The Dartmouth Atlas has already shown big variation in the amount of care doctors provide in different areas of the country. Their latest research, released Tuesday, suggests this variation is rampant among the most respected medical institutions.

Researchers at the Atlas looked at how the top 23 academic medical centers, as ranked by U.S. News and World Report, provide care to their patients. Their results show huge variations in how the very best hospitals care for their sickest patients.

A patient at New York Presbyterian Hospital can expect to spend 20 days in the hospital during the last six months of life. The average Mayo Clinic patient would have 10 days in the hospital over the same time span.

At the University of Utah, the average patient sees 20 doctors during the last six months of life. At the Cedars-Sinai Hospital in California, that number is 73 physicians.

A patient at UCLA’s Ronald Reagan Hospital is three times as likely to have a fall or injury while in the hospital than one at Cedars Sinai, also in Los Angeles.

“We know these differences cannot be explained by the prevalence of diseases,” Goodman says. “Yes, populations differ. Those are small differences we see compared to the dramatic differences in care.”

That matters for patients who stand to get markedly different care depending on where they seek treatment. It also matters for medical residents, who are likely to pick up different habits of care depending on where they train.

“We believe that medical students should take into account the way physicians deliver health care,” says Anita Arora, a co-author of the study and recent graduate of the Geisel School of Medicine at Dartmouth College. “The culture of practice determines a hidden training curriculum. Residency shapes the way physicians are going to practice the rest of their lives.”