Average or Poor Hospitals Linked to Excess Medicare Mortality

GOLDEN, Colo., Oct. 16 -- More than 300,000 Medicare patients died from 2003 to 2005 because they were hospitalized in institutions that were average or poor on report cards issued today by HealthGrades, a for-profit health care quality rating service here.

HealthGrades, which evaluated Medicare records from 40.6 million hospitalizations during those years, estimated that last year a patient treated at a hospital that received five stars, the stellar rating, had a 69% lower chance of dying than a similar patient treated at a hospital rated one-star, the poorest, and a 49% lower chance of death than if treated at an average hospital (three stars).

The rating service evaluated 5,000 hospitals. A HealthGrades spokesperson said that "roughly 15%" of the hospitals received five-stars in the 28 diagnoses and procedure categories included in the report.

According to the report, if all hospitals that treat Medicare patients performed as well as HealthGrades' five-star rated best hospitals, "302,403 Medicare lives could have potentially been saved from 2003 to 2005."

Half of the excess Medicare deaths occurred in patients diagnosed with heart failure, community acquired pneumonia, sepsis, and respiratory failure treated in hospitals that received a poor rating for treating those conditions, according to HealthGrades.

Additional findings:

• Fewer Medicare patients died in hospitals in 2005 than in 2003, but the degree of improvement in mortality rate varied greatly by diagnosis.
• Compared with hospitals rated average (three-star), the risk-adjusted mortality rates at five-star hospitals improved 19% across all three years (2003 to 2005).
• A typical Medicare patient had about a 49% lower risk of dying in a top rate hospital last year than in hospital that was rated average by HealthGrades.

HealthGrades awards five stars for hospitals that performed significantly better than predicted, three for those that performed as predicted (average), and one star for those that performed significantly worse than predicted.

The ratings are issued on the basis of risk-adjusted mortality rates and complication rates. There are 18 mortality-rating categories: were abdominal aortic aneurysm repair, acute myocardial infarction, atrial fibrillation, bowel obstruction, chronic obstructive pulmonary disease, community acquired pneumonia, coronary bypass surgery, coronary interventional procedures, diabetic acidosis and coma, GI bleed, GI surgery and procedures, heart failure, pancreatitis, pulmonary embolism, respiratory failure, sepsis, stroke and valve replacement surgery.

There are categories for complication rates: back and neck surgery (except spinal fusion), carotid endarterectomy, cholecystectomy, gastrointestinal procedures and surgeries, hip fracture repair, partial hip replacement, peripheral vascular bypass, prostatectomy, total hip replacement, and total knee replacement.

For 26 of the 26 categories, HealthGrades used a multivariate logistic regression-based rating model that included predicted mortality, compared the predicted mortality to actual mortality, and tested to determine whether the difference between predicted and actual mortality was significant. Hospitals with fewer than 30 cases in a specific diagnostic category over the three-year period were removed from the analysis.

Two categories -- respiratory failure and gastrointestinal procedures and surgeries -- were rated using software developed by the 3M Corporation called APR-DRGs that were designed to factor in severity of illness and mortality risk across these two specific diagnosis groups.

HealthGrades, which rates nursing homes and physicians as well as hospitals, sells its quality rating information to employers, health plans, hospitals, and consumers.