We have established the Merged Cardiac Registry (MCR) containing over 100,000 cardiovascular surgery cases from 47 sites in the U.S. and Europe. MCR outcomes data are used by the contributors for clinical quality improvement. A tool for prospective prediction of mortality and stroke for coronary artery bypass graft (CABG) surgery (83% of the cases), known as RiskMaster, has been developed using a Bayesian model based on 40,819 patients who had their surgery from 1988-92, and tested on 4,244 patients from 1993. In patients with mortality risks of 10% or less (92% of cases), the average risk prediction is identical to the actual 30-day mortality (p > 0.37), while risk is overestimated in higher risk patients. The receiver operating characteristic (ROC) curve area for mortality prediction is 0.76 ± 0.02. The RiskMaster prediction tool is now available online or as a standalone software package. MCR data also shows that average mortality risk is identical for a given body surface area regardless of gender. Outcomes data measure the benefits of health care, and are therefore an essential element in cost/benefit analysis. We believe their cost is justified by their use for the rational assessment of treatment alternatives.