BACKGROUND Clinical trials and practice guidelines have identified clinical criteria for the use of coronary angiography and revascularization procedures after thrombolysis for acute myocardial infarction. The effect of these criteria on clinical practice has not been extensively evaluated. METHODS We used classification-and-regression-tree (CART) and logistic-regression models to study the patients in the first Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries trial, to identify the variables that best predicted the use of angiography and revascularization procedures after thrombolysis. RESULTS Among the 21,772 U.S. patients in the trial, 71 percent underwent coronary angiography before discharge from the hospital. Of these, 58 percent underwent revascularization (73 percent receiving angioplasty). The CART model for the use of angiography showed that age was the variable most predictive of angiography; only 53 percent of patients at least 73 years of age underwent angiography, as compared with 76 percent of those under 73. Among the older patients, age was again the most predictive factor; among the younger patients, the availability of angioplasty was a more important predictor (67 percent of patients in hospitals without angioplasty facilities underwent angiography, as compared with 83 percent in hospitals with such facilities). The next most important variable was recurrent ischemia, which was more predictive at hospitals without angioplasty facilities than at those with them. Both statistical models identified coronary anatomy as the most important predictor of the use and type of revascularization. CONCLUSIONS More patients treated with thrombolysis underwent angiography and revascularization before discharge than might be expected. Younger age and the availability of the procedures appeared to be the major determinants of the use of coronary angiography, whereas coronary anatomy largely determined the use and type of revascularization. This process appeared to select low-risk patients for intervention rather than those at higher risk, who would be the most likely to benefit.