Yale Study of Hospital Website Ratings Finds Mixed Results
There is increasing interest in using health data to help patients make choices about hospitals. A common approach is to use publicly available billing data from Medicare to compare rates of death and complications among different facilities.
The principal investigator, Harlan Krumholz, M.D., associate professor of internal medicine, cardiology and in the Department of Epidemiology and Public Health, and his colleagues, conducted a study of HealthGrades.com, Inc., one such rating system, which is promoted on a popular Internet website. The Yale group focused on the ratings of hospitals for the care of patients with heart attacks and studied 3,363 hospitals to determine whether hospital ratings from the website provided useful information to patients.
The website used the billing data to classify hospitals as having had better than expected quality (five-star) or worse than expected quality (one star), based on their mortality using an unpublished mathematical model.
The Yale investigators compared HealthGrades.com’s hospital ratings with detailed medical record information that was collected from hospitals as part of a federal project examining the quality of care provided to elderly patients hospitalized for a heart attack.
“The results were mixed,” Krumholz said. “If you aggregate the hospitals based on their ratings, there is a gradient of risk of death. Patients at five star hospitals did have a lower rate of dying from heart attacks than patients treated at one star hospitals. For the use of medications endorsed by the experts, the higher rated hospitals did not always excel in their use of treatments, and in some cases the lower rated hospitals did better.”
The website hospital ratings were not very useful when used to draw comparisons between individual hospitals. “There is tremendous heterogeneity in how hospitals did within each of the website rating groups,” Krumholz said. “There was substantial overlap in hospital performance across all the categories.”
The authors found that assuming a higher rated hospital has a lower death rate than a lower rated hospital is often incorrect. “In fact, most of the comparisons we made between higher rated hospitals and lower rated hospitals indicated they were statistically comparable,” Krumholz said. “We compared five star hospitals to one star hospitals and there was no statistically significant difference in mortality rates.”
Krumholz said the hospital ratings are only slightly better than no information at all, but convey a lot less information than a patient might expect. “In most cases,” he said, “you could not really say that the higher rated hospitals were better. The principal problem is that many hospitals are misclassified by this system, with winners and losers that may not deserve such distinctions.”
“Rather than looking at these type of ratings, it may be more helpful for consumers to know how hospitals do on accepted measures of quality, such as the timely administration of critical medications or the use of therapies endorsed by national guidelines,” Krumholz said. “Current efforts focusing on mortality data using billing data may be more misleading than illuminating for the public.”