Background. New-onset postoperative atrial fibrillation (POAF) is a frequent rhythm disturbance after coronary artery bypass grafting (CABG). This study investigated the independent effect of POAF on early and late mortality after isolated CABG. Methods. Data of patients who consecutively underwent isolated CABG between January 2003 and December 2007 were prospectively collected. The analysis included 5098 patients with preoperative sinus rhythm and no history of atrial fibrillation. Logistic regression analysis for early mortality and Cox regression analysis for late mortality were performed. Propensity score matching was performed to eliminate the effect of confounders. Results. Median follow-up was 2.5 years. POAF was documented in 1122 patients (22.0%). Early mortality was more frequent in POAF patients (3.1%) vs non-POAF patients (1.6%, p = 0.002), but multivariate logistic regression analysis could not identify POAF as an independent predictor of early mortality (p = 0.169). This outcome did not change after adjusting for quintiles of the propensity score of POAF (p = 0.100). Multivariate Cox proportional hazard analyses demonstrated POAF was an independent predictor of overall and late mortality with hazard ratios of 1.35 (p = 0.012 and p = 0.039, respectively). Analyses after propensity score matching showed that patients with POAF had similar hazard ratios of 1.36 for overall mortality and 1.34 for late mortality (p = 0.009 and p = 0.042, respectively). Conclusions. POAF is an independent predictor of overall and late mortality after isolated CABG but not of early mortality.