Use of a Bayesian statistical model for risk assessment in coronary artery surgery