Good hospital care probably is around the corner, if all you need is a routine procedure—sometimes even major surgery. But a difficult condition or a complex operation demands a higher standard of care, especially if the risk is compounded by advanced age, frailty, obesity, or some other complication.

That describes the patients U.S. News set out to help, starting in 1990, with "America's Best Hospitals." The mission, unchanged over 17 years, has been to identify centers that take on and excel at tough procedures and conditions—rare cancers, worsening heart failure, seemingly untreatable leg-artery blockages. That is why most of the institutions ranked are referral centers, where the sickest patients are sent for advanced care. Such hospitals follow—and often pioneer—new treatment guidelines. They conduct bench-to-bedside research. And they exploit the latest advances in imaging, surgical devices, and other technologies.

In our online package, you'll find hospitals ranked in 16 specialties, from cancer and heart disease to pediatrics and urology. (Geriatrics was dropped this year, because it is oriented more to primary care than to specialized hospital treatment.) Out of the 5,189 hospitals evaluated, just 176 scored high enough this year in such measures of quality as mortality and patient volume to be ranked in any specialty.

In 11 of the 16 specialties, hard data largely determine a hospital's position. (In the others five, as explained below, the rankings are based only on hospitals' reputation among specialist physicians.) Initial eligibility for these data-driven specialties requires a hospitals to meet any of three standards: membership in the Council of Teaching Hospitals, affiliation with a medical school, or—to open the door to non-teaching hospitals—availability of at least nine out of 18 key technologies like shaped beam radiation, an advanced cancer therapy. This year, fewer than a third of the hospitals qualified.

In individual specialties, hospitals had to admit and treat a minimum number of patients in 2002, 2003, and 2004 with sufficiently complex conditions in 2002, 2003, and 2004—a three-year total of at least 415 in orthopedics, for example—or had to be nominated by at least one physician as among the best in that specialty in 2004, 2005, or 2006.

A hospital's overall performance is summed up by its U.S. News Score. It is made up of three equal parts: reputation, mortality, and a mix of care-related factors such as nursing and patient services. The 50 hospitals in each specialty with the highest scores are listed. The 14 hospitals on the Honor Roll demonstrate exceptional breadth of excellence: In six or more specialties, their scores were at least two standard deviations above the mean.

Here's a breakdown of the three major elements of the scores. The glossary explains all of the individual measures that appear in the rankings.

Reputation. For each specialty, a sampleing of board-certified physicians was randomly selected from the American Medical Association's Masterfile of all 860,000 U.S. doctors (200 per specialty in 2006 and 2005, 150 in 2004). The doctors were mailed a survey form and asked to list the five hospitals they feel are best in their specialty for difficult cases, without consideration of cost or location. The number shown represents the percentage of responding physicians who named the hospital. This year, about 47 percent of the 3,200 doctors contacted responded to the survey.

Mortality. The number shown is a ratio. It compares the number of Medicare patients with certain conditions who died in the hospital in 2002, 2003, and 2004 with the number of deaths expected after the severity of their condition is factored in. Below 1.00 means better than expected; above 1.00, worse than expected. Severity adjustments were derived from 3M Health Information Systems software (All Patient Refined Diagnosis Related Group).

Other factors. The final third of the score depends on such quality-of-care measures as nurse-to-patient ratio and the number of key technologies available. Most of the data came from the American Hospital Association's 2004 member survey; for the few ranked hospitals that did not respond to the survey, 2003 data were used instead.

Reputational specialties. Ranking was by reputation only in ophthalmology, pediatrics, psychiatry, rehabilitation, and rheumatology, because mortality data are unavailable for pediatric facilities and are irrelevant or unreliable in the others. Ranked hospitals were named by at least 3 percent of the responding physicians.

The 2006 rankings were produced by RTI International, a leading research organization in Research Triangle Park, N.C.

A complete explanation of the methodology (PDF)