Background: Medical costs vary substantially among patients. Understanding the baseline factors that predict subsequent cost may allow better selection of therapy for individual patients. Understanding the postprocedure events that increase cost should help to improve efficiency and effectiveness of coronary revascularization. Methods: Data on 4-year costs were collected from patients randomly assigned to coronary angioplasty or bypass surgery as part of the BARI (Bypass Angioplasty Revascularization Investigation) trial. Regression models first examined factors known at the time of randomization that prospectively predicted initial procedure cost and long-term cost. Subsequent models tested the value of postrandomization events as explanatory variables for cost. Results: The independent baseline predictors of higher initial percutaneous transluminal coronary angioplasty cost included 3-vessel disease (+12%) and acute presentations (+22%), whereas the independent predictors of higher initial coronary artery bypass grafting cost included the number of comorbid conditions (+5% per condition) and female sex (+7%). The independent baseline predictors of 4-year cost included heart failure (+26%), diabetes (+22%), comorbidity (+10%), and angioplasty assignment in patients with 2-vessel disease (- 15%). Postrandomization models showed higher initial and long-term costs were strongly correlated with the number of repeat revascularization procedures (+30% to +128%) and the occurrence of clinical complications (+8% to + 131%). Conclusions: Two- vessel disease identifies patients likely to have lower costs after angioplasty, whereas heart failure, comorbid conditions, and diabetes identify patients likely to accrue higher costs after either angioplasty or bypass surgery. Long-term costs can be potentially reduced by interventions that decrease procedural complications or reduce the need for repeat revascularization.