Challenges in comparing risk-adjusted bypass surgery mortality results
Results from the Cooperative Cardiovascular Project
Eric D. Peterson, MD, MPH, FACC*,
Elizabeth R. DeLong, PhD*,
Lawrence H. Muhlbaier, PhD*,
Allison B. Rosen, MD, MPH*,
Hope E. Buell, MS*,
Catarina I. Kiefe, MD, PhD and
Timothy F. Kresowik, MD, MPH
* The Duke Outcomes Research and Assessment Group, Duke University Medical Center, Durham, North Carolina, USA
The Alabama Quality Assurance Foundation, and the University of Alabama at Birmingham Center for Outcomes and Effectiveness Research and Education, Birmingham, Alabama, USA
The Iowa Foundation for Medical Care, West Des Moines, Iowa, USA
View larger version (17K):
[in a new window]
|
Figure 1 This figure the ROC curves for the four bypass surgery risk models. The C-index is equivalent to the area under each ROC curve.
|
|
View larger version (14K):
[in a new window]
|
Figure 2 A, The observed to expected mortality rates for each quintile of patient risk. Each risk quintile contains approximately 750 patients. The diagonal line represents perfect agreement between observed and expected mortality estimates. B, The same information after the models have been internally recalibrated in the CCP database.
|
|
View larger version (16K):
[in a new window]
|
Figure 3 A, Each hospital’s unadjusted mortality rates and their risk-adjusted mortality using the Parsonnet and Hannan risk models. Note: the 28 Hospitals are ordered on the x-axis by the unadjusted mortality rate. B, This same information after the Parsonnet and Hannan models have been internally recalibrated in the CCP database.
|
|
|
