Abstract

As risk-adjusted outcome is increasingly being used to make clinical decisions and to assess and improve quality of care, it is important to develop simple, stable models for predicting outcome. Here we address the hypothesis that a risk factor for increased operative mortality at coronary artery bypass grafting may have differential effects in subgroups of patients defined by the presence or absence of other risk variables. We used a series of univariate and multivariate analyses to identify a group of ten patient-related preoperative characteristics independently predictive of operative death in the total population of 12,712 patients undergoing coronary artery bypass grafting at 43 Department of Veterans Affairs medical centers participating in the Department of Veterans Affairs Continuous Improvement in Cardiac Surgery Study. Separate logistic regression models were then developed for each of 14 clinically important subgroups defined by the presence or absence of seven risk variables. Odds ratios for operative death and estimated operative mortality were calculated from these subgroup regression models. Of 65 comparisons of pairs odds ratios of preoperative risk variables between the subgroups with and without a second risk variable, only five were found to be significant (p 0.05 without adjustment for multiple comparisons); this is only slightly more than would be expected by chance alone. Risk factors for increased operative death appear to have similar odds ratios for subgroups of patients defined by a second risk variable. This finding greatly simplifies the use of predicted operative mortality in clinical decision making and quality assessment and improvement in coronary artery bypass grafting.