Improving the Statistical Approach to Health Care Provider Profiling

15 October 1997 | Volume 127 Issue 5 Part 2 | Pages 764-768

This paper reviews and compares existing statistical methods for profiling health care providers. It recommends improvements that are based on the use of better statistical models and the adoption of more realistic, medically based criteria for judging the performance of health care providers. Unlike most profiling methods, the proposed hierarchical models allow the probability of acceptable provider performance to be calculated; thus, they can answer such questions as, "What is the probability that a given hospital's true mortality rate for cardiac surgery patients exceeded 3.33% last year?" The commonly encountered problems of regression-to-the-mean bias and small caseloads can be handled by using hierarchical models to extract more information from profiling data.

Author and Article Information

From Harvard Medical School and Harvard Pilgrim Health Care, Boston, Massachusetts; and Harvard University, Cambridge, Massachusetts.
Note: This article is one of a series of articles comprising an Annals of Internal Medicine supplement entitled "Measuring Quality, Outcomes, and Cost of Care Using Large Databases: The Sixth Regenstrief Conference." To see a complete list of the articles included in this supplement, please view its Table of Contents.
Grant Support: In part by Agency for Health Care Policy and Research grant RO1 HS 07118-01 and the Department of Veterans Affairs, Management Science Group.
Requests for Reprints: Cindy L. Christiansen, PhD, Harvard Medical School and Harvard Pilgrim Health Care, 126 Brookline Avenue, Suite 200, Boston, MA 02215.
Current Author Addresses: Dr. Christiansen: Harvard Medical School and Harvard Pilgrim Health Care, 126 Brookline Avenue, Suite 200, Boston, MA 02215.