Back to Annual Meeting

Back to Annual Meeting

APHA Scientific Session and Event Listing

Recorded presentation

George J. Stukenborg, PhD and Douglas P. Wagner, PhD. Public Health Sciences, University of Virginia School of Medicine, Blake Center Room 400B, 1224 West Main Street, Charlottesville, VA 22908-0821, 434-924-8649, gstukenborg@virginia.edu

Many states publish lists identifying hospitals with higher or lower than expected numbers of deaths for acute myocardial infarction (AMI) patients. The California Hospital Outcomes Project study of AMI is one of the most comprehensive examples. We reproduced the original California mortality risk adjustment model and developed a new mortality risk adjustment model that adjusts for comorbid disease measured by diagnoses reported as present at admission. Multivariable logistic regression was used to re-estimate the original California model and to develop the new model for 120,706 California patients hospitalized with AMI from January 1996 through November 1998. The discrimination achieved by both models was measured in a validation population of 42,678 California AMI patients hospitalized from December 1998 through November 1999. The new model obtained a C statistic of 0.86, which substantially exceeded the C statistic of 0.76 achieved by the original California study model. Indirectly standardized mortality ratios were calculated for each of the 416 hospitals in the study. Hospitals with significantly more deaths than expected (or fewer deaths than expected) were identified in the same manner as the original study, using both models. The new model identified 12 hospitals with lower then expected mortality, and 23 with higher than expected mortality, compared to the 19 hospitals with lower than expected mortality and 36 with higher than expected mortality in the original study. The large improvement in discrimination obtained using present at admission diagnoses resulted in substantially fewer hospitals identified with higher or lower than expected mortality.

Learning Objectives:

• develop understanding of methods for measuring comorbid disease and adjusting for differences in mortality risk in outcomes research using large administrative data sets

Presenting author's disclosure statement:

Any relevant financial relationships? No

Recorded presentation