To the Editor.
—In the November 1986 issue of the Archives, Detsky et al1 presented a modified version of the index of Goldman et al2 of cardiac risk in noncardiac surgical procedures. In both these studies, a multifactorial risk index is derived using stepwise linear discriminant analysis. Whereas Goldman et al include the type of surgery among the factors used to calculate the risk index, Detsky et al use a Bayesian approach, modeled on that of "Clinical Epidemiology Rounds" (CER),3 in which the calculation of a patient's overall probability of cardiac complications involves the use of a "pretest probability" based on the given hospital's experience with the particular surgical procedure planned for the patient, and a "likelihood ratio" (LR), which is calculated from the patient's risk score that in his approach is assumed to be independent of the surgical procedure to be done. A nomogram is used