|
|
Predictive Model of Atrial fibrillation in 5600 Isolated Coronary Artery Bypass Surgeries
L. U. Nwakanma1, C. W. Hogue2, D. E. Alejo1, R. E. Thompson3, W. A. Baumgartner*1, J. V. Conte*1. 1Division of Cardiac Surgery. The Johns Hopkins Medical Institutions, Baltimore, MD, 2Department of Anesthesiology and Critical Care Medicine. The Johns Hopkins Medical Institutions, Baltimore, MD, 3Department of Biostatistics. Johns Hopkins School of Hygiene and Public Health, Baltimore, MD,
Background: New onset atrial fibrillation (AF) is a costly and common complication occurring in up to 40% of patients after coronary artery bypass grafting (CABG). Predictive models have been proposed but several of those are based on studies with smaller sample sizes. We used peri-operative risk factors to predict the relative risk of developing post-operative AF in a large cohort of patients undergoing CABG.
Methods: Pre-operative risk factors from consecutive patients ≥ 30yrs old undergoing isolated CABG surgery from 1995 to 2006 were prospectively entered into our Society of Thoracic Surgeons database. Multivariable logistic regression predictive models for development of post-operative atrial fibrillation were generated using peri-operative risk variables. The model was used to compare the predicted probability of AF with the known outcome in patients divided into deciles by probability.
Results: There were 6,348 patients, mean age 65±11 yrs old (range, 32 to 95 yrs). Of these, 421 patients had pre-operative AF and were excluded from the analyses. Of the 5671 patients with complete data, 1661 developed post-operative AF (29.3%).The variables with the most impact on the predictive model were increasing patient age and height, hypertension, the need for prolonged ventilation (24 hours or more) and Caucasian race. The model showed a good concordance of 71.3% between observed and predicted incidence of post-operative AF, a receiver operating characteristic curve area of 0.68 (c-index), and the Hosmer-Lemeshow test had a chi square value of 4.7 (p=0.79), showing a good fit for the model.
Age had the most impact on the model with a decrease of c-index to 0.587 when age is removed from the regression model.
Conclusions: Use of a validated predictive risk model can reliably stratify patients into risk groups for developing AF after CABG surgery. This may help direct aggressive prophylactic treatment to selected high risk patients. 
Back to 2008 Annual Meeting
Back to Program Outline
|
