Letters Section Editor: Robert M. Golub, MD, Senior Editor.
In Reply: The validity of comparisons of rates of risk-adjusted clinical outcomes as measures of hospital performance can be biased by (1) the exclusion of important risk factors from predictive equations, (2) the improper use of hospital-acquired complications as risk factors,1 and (3) inconsistent reporting of valid risk factors included in predictive equations. Our study was designed to address the first of these sources of bias, took great care to avoid the second, and, as Drs Heller and Schnell correctly observe, did not address the third directly. However, potential clinical data elements were grouped from the most objective (ie, numerical laboratory data) to the least objective (ie, composite scores such as ASA Class), and more subjective variables were included in predictive equations only when they added significant predictive power to equations containing more objective clinical information. This approach ensured that more consistently documented clinical data would not be displaced from the models by more subjective findings.
Pine M, Jordan HS, Elixhauser A, Fry DE, Hoaglin DC. Hospital Mortality Risk Adjustment Using Claims Data—Reply. JAMA. 2007;297(18):1983-1984. doi:10.1001/jama.297.18.1984