Author + information
- Robert H. Jones, MD, FACC1,∗∗,
- Edward L. Hannan, PhD∗,
- Karl E. Hammermeister, MD, FACC†,
- Elizabeth R. DeLong, PhD‡,
- Gerald T. O'Connor, PhD§,
- Russell V. Luepker, MD, FACC∥,
- Victor Parsonnet, MD, FACC¶,
- David B. Pryor, MD, FACC∗∗,
- Working Group Panel on the Cooperative CABG Database Project#
- ↵1Address for correspondence: Dr. Robert H. Jones, P. O. Box 2986, Duke University Medical Center, Durham, North Carolina 27710.
Objectives. The purpose of this consensus effort was of define and prioritize the importance of a set of clinical variables useful for monitoring and improving the short-term mortality of patients undergoing coronary artery bypass graft surgery (CABG).
Background. Despite widespread use of data bases to monitor the outcome of patients undergoing CABG, no consistent set of clinical variables has been defined for risk adjustment of observed outcomes for baseline differences in disease severity among patients.
Methods. Experts with a background in epidemiology, biostatics and clinical care with an interest in assessing outcomes of CABG derived from previous work with professional societies, government or academic institutions volunteered to participate in this unsponsored consensus process. Two meetings of this and hoc working group were required to define and prioritize clinical variables into core, level 1 or level 2 groupings to reflect their importance for relating to short-term mortality after CABG. Definitions of these 44 variables were simple and specific to enhance objectivity of the 7 core, 13 level 1 and 24 level 2 variables. Core and level 1 variables were evaluated using data from five existing data bases, and core variables only were examined in an additional two data bases to confirm the consensus opinion of the relative prognostic power of each variable.
Results. Multivariable logistic regression models of the seven core variables showed all to be predictive of bypass surgery mortality in some of the seven existing data sets. Variables relating to acuteness, age and previous operation proved to be the most important in all data sets tested. Variables describing coronary anatomy appeared to be least significant. Models including both the 7 core and 13 level 1 variables in five of the seven data sets showed the core variables to reflect 45% to 83% of the predictive information. However, some level 1 variables were stronger than some core variables in some data sets.
Conclusions. A relatively small number of clinical variables provide a large amount of prognostic information in patients undergoing CABG.