Hospital Deaths Go Public

How well thousands of individual U.S. hospitals do on the ultimate test of hospital competence—whether very sick patients live or die—is now revealed on the Web for all to see, in hard numbers. The Hospital Compare page maintained by the Centers for Medicare & Medicaid Services (CMS) this week began displaying actual hospital death rates for heart attack, heart failure, and pneumonia patients. Until now, only relative performance had been shown—whether mortality for such patients at a particular hospital was about the same as the national average, better than average, or worse than average.

This is a big step forward. The death rates, says Harlan Krumholz, director of the Center for Outcomes Research and Evaluation at Yale-New Haven Hospital, which developed the mortality measure, "provide insights into where hospitals are compared to where they should be."

Translation: Hospitals with high death rates will be held more accountable than when they could hide behind "worse than average," which is simply too vague to carry much impact. "We hope this will elicit conversations," says Krumholz. "Board members should be interested in the performance of their hospital. Communities should be interested in the performance of their hospitals."

The rates are adjusted to take patients' risk factors into account—age; current or previous heart problems; and existing conditions like diabetes, asthma, and dementia, for example. And deaths are counted if they occur within 30 days of admission, in or out of the hospital, because poor care can have a lasting effect—about one-third of deaths occur after patients are discharged.

The new mortality rates are not a consumer tool for picking one hospital over another. For one, the numbers are still imprecise. For statistical reasons, the smaller the number of heart attack, heart failure, and pneumonia patients, the less certain a hospital's calculated death rates for the condition. So each figure is accompanied by an "interval estimate"—a horizontal bar showing the range into which the actual rate could fall.

The larger the hospital, the narrower the range. Cedars-Sinai Medical Center in Los Angeles, for instance, had 370 Medicare pneumonia patients in the year that ended in June 2007. Their death rate was 7.5 percent (better than the national average of 11.4 percent), with a range from roughly 6 to 10 percent. Ronald Reagan UCLA Medical Center only had 76 Medicare patients with pneumonia; their death rate of 9.6 percent had a much wider range of about 7 to 13 percent.

I wondered how well the death numbers track performance on the "process measures" that CMS has been putting online since 2003. Some of the measures reveal hospitals' diligence in following recognized steps when dealing with heart attack, heart failure, and pneumonia patients—unblocking the arteries of heart attack victims within 90 minutes, evaluating the heart's pumping ability in heart failure patients, and giving flu shots to pneumonia patients, for instance. When I checked six medical centers in Baltimore, the Washington, D.C., area, and Los Angeles, how often these patients died there did not seem to have much to do with performance in the related process measures.

Six isn't much of a sample, but my impression reflects reality, says Krumholz. In 2006, he was lead author of a study in the Journal of the American Medical Association that found only limited correlation between hospital heart attack deaths and the process measures. To be precise, how well a hospital did in the heart attack process measures could explain just 6 percent of the differences in death rates from one hospital to another, probably because the process measures are just a few of a multitude of factors that influence death rates. Still, says Krumholz, the process measures are all part of the complex portrait of care.

Besides the new mortality data, CMS added its first pediatric indicators—two process measures that reflect the percentage of children hospitalized with asthma who got two different types of medications. For now, the results are hardly helpful. Thousands of hospitals have at least some pediatric beds, but only 79 centers agreed to make their figures public. (It's voluntary for new measures.) "The number will increase," promises Donald McLeod, a CMS spokesman.

Should that happen, hopefully the public won't have a problem finding specific hospitals that have "children" in the name but are part of a larger facility. Looking for Inova Fairfax Hospital for Children in Virginia, for example, would be fruitless. It is listed as Inova Fairfax Hospital. "It's just the name of the hospital as reported to us," says McLeod. "There may be a kink we need to straighten out."