[Skip to Content]
[Skip to Content Landing]
Article
December 23, 1988

Interpreting Hospital Mortality DataThe Role of Clinical Risk Adjustment

Author Affiliations

From the Office of Research, Health Care Financing Administration, Baltimore (Dr Jencks); Baxter Healthcare Corporation, The Health Data Institute, Lexington, Mass (Dr Daley, Mr Lenhart, and Ms Walker); the Division of General Medicine and Primary Care, Department of Medicine, Beth Israel Hospital, Harvard Medical School, the Charles A. Dana Research Institute, and the Harvard-Thorndike Laboratory, Boston (Dr Daley); and the Department of Economics and Statistics, The RAND Corporation, Santa Monica, Calif (Drs Draper and Thomas).

From the Office of Research, Health Care Financing Administration, Baltimore (Dr Jencks); Baxter Healthcare Corporation, The Health Data Institute, Lexington, Mass (Dr Daley, Mr Lenhart, and Ms Walker); the Division of General Medicine and Primary Care, Department of Medicine, Beth Israel Hospital, Harvard Medical School, the Charles A. Dana Research Institute, and the Harvard-Thorndike Laboratory, Boston (Dr Daley); and the Department of Economics and Statistics, The RAND Corporation, Santa Monica, Calif (Drs Draper and Thomas).

JAMA. 1988;260(24):3611-3616. doi:10.1001/jama.1988.03410240081036
Abstract

This study uses national Medicare data as well as data that were abstracted to calibrate the Medicare Mortality Predictor System to assess the usefulness of a risk adjustment system in interpreting hospital mortality rates. The majority of variation in annual hospital death rates for the four conditions studied (stroke, pneumonia, myocardial infarction, and congestive heart failure) is chance variability that results from the relatively small numbers of patients treated in most hospitals in a year. For hospitals in the highest and lowest quartiles of observed death rates, the difference between observed rates and those predicted by the Medicare Mortality Predictor System is not quite one third smaller than the difference between observed rates and unadjusted national rates. Risk adjustment methods do not show whether the unexplained difference in mortality rates results from differences in effectiveness of care or unmeasured differences in patient risk at the time of admission. Risk-adjusted mortality rates, therefore, should be supplemented by review of the actual care rendered before conclusions are drawn regarding effectiveness of care.

(JAMA 1988;260:3611-3616)

×