BACKGROUND: Knowledge of the risk of atrial fibrillation (AF) for patients undergoing coronary artery bypass graft surgery (CABG) can guide decisions about prophylactic therapy. Accordingly, we sought to use tree-based methods to stratify patients into groups that will have similar risk of AF after CABG and informed decision making regarding aggressive prophylaxis of AF. METHODS: We studied 1209 consecutive patients with isolated CABG performed in 1998-1999 at Yale-New Haven Hospital. Patients with preoperative AF were excluded. Tree-based analysis was carried out to stratify patients into similar groups regarding the risk of AF. Relative risks (RRs) and 95% CIs were calculated at each level of stratification. RESULTS: Age was the most important variable. The importance of other risk factors seemed to be different for younger and older patients. Although in the younger age group (< or =60 years) severity of coronary artery disease (RR 2.19, 95% CI 1.12-3.34) followed by hypertension (RR 1.82, 95% CI 1.23-2.68) were important predictors, in the older age subgroups (61-69 and > or =70 years), nothing or only ejection fraction <40% (RR 1.31, 95% CI 1.08-1.59) was important. In the highest-risk group, AF occurrence was 55% and, in the lowest-risk group, it was 10%. In the low-risk groups, aggressive prophylaxis may not be justified in light of the smaller number of events that would be prevented, possible adverse events, and costs. CONCLUSION: Age and variables related to heart disease severity are predictors of AF. The tree-based method may be a useful tool for clinicians who seek to determine who is more or less likely to benefit from aggressive arrhythmia prophylaxis.