Development and evaluation of models to predict death and myocardial infarction following coronary angioplasty and stenting.

Prior estimates of the risk death and myocardial infarction following percutaneous coronary intervention (PCI) may not be uniformly applicable due to recent significant changes in clinical practice. Accordingly, we studied 2,804 cases from January 1997 through February 1999, in order to develop risk models to predict death, and post-procedural myocardial infarction following PCI. Risk models were constructed using multivariate logistic regression, artificial neural networks and prognostic risk scoring systems. Composite logistic regression models and artificial neural networks performed similarly in predicting the risk of major acute complications (c-index for predicting death of 0.812 and 0.807, respectively). Risk scoring models, based on the composite logistic regression beta coefficients, performed only slightly worse (c-index death = 0.794). Risk score models appear to provide reasonable discrimination while offering the potential for simple clinical implementation in the estimation of the risk of death and myocardial infarction in interventional cardiology.