Variability in coding of pneumonia cases skewed risk-standardized mortality rates and hospital performance rankings, investigators reported online March 17.
The bias could impede efforts to compare quality of care among hospitals, Dr. Michael Rothberg of the Cleveland Clinic in Ohio and his associates said (Ann. Int. Med. Mar. 17 [doi: 10.7326/M13-1419]).
Dr. Michael Rothberg
The Centers for Medicare & Medicaid Services partially based hospital reimbursements on 30-day risk-standardized mortality rates. To exclude nosocomial pneumonia cases, CMS included only patients with a primary diagnosis of pneumonia when estimating 30-day risk-standardized pneumonia mortality rates.
But over time, hospitals changed how they coded the sickest patients with pneumonia, Dr. Rothberg and his associates said. These patients increasingly received a principal diagnosis of sepsis or organ failure, instead of pneumonia, and thus were excluded from mortality estimates.
"These events gave the false impression that pneumonia outcomes had improved more than they had," they said, adding that "just as changes in coding over time could lead to erroneous conclusions about decreasing mortality rates, variation in coding across hospitals could lead to biased estimates of their relative mortality rates."
The investigators conducted a cross-sectional study of more than 250,000 hospitalizations of adults with a principal or secondary diagnosis of pneumonia at 329 U.S. hospitals between 2007 and 2010.
When the definition of pneumonia excluded patients with primary sepsis or respiratory failure, 4.3% of hospitals had mortality rates that were significantly better than the mean, and 6.4% had rates that were significantly worse. But when the expanded definition was used, 12% of hospitals had mortality rates that were better than the mean, while 23% had rates that were worse. Performance ranking changed for 28% of hospitals.
When the expanded case definition was used, outlier status worsened for 41% of the hospitals with the highest proportions of patients with primary sepsis or respiratory failure, but improved for 20% and worsened for none of the hospitals with the lowest proportions of these patients.
"Efforts to broaden the scope of hospital performance measures from the initial set of measures based on processes to those focused on patient outcomes are laudable, but caution is required," the investigators said. "Misclassification could harm individual hospitals and weaken confidence in public reporting."
Adding principal diagnoses of respiratory failure or sepsis with secondary pneumonia to the definition of pneumonia could help lessen the biases, they said.
The study was supported by the Agency for Healthcare Research and Quality. Dr. Rothberg and Dr. Lindenaur also received grants from AHRQ.
*Correction, 3/19/2014: An earlier version of this story misstated the name of Dr. Scott A. Flanders, one of the editorialists.