Comparison of an automated thrombolytic predictive instrument to both diagnostic software and an expert cardiologist for diagnosis of an ST elevation acute myocardial infarction

Abstract

Because the electrocardiograms (ECGs) of patients with symptoms suggesting an acute thrombotic coronary occlusion are typically read by physicians relatively inexperienced in this skill, it is important to develop automated decision support. A Thrombolytic Predictive Instrument (TPI) is now available along with the standard diagnostic software in a commercially available electrocardiograph. This study evaluates the performance of the predictive software in comparison to both an expert cardiologist and standard diagnostic software. True sensitivity and specificity cannot be determined because acute coronary angiography was not performed. The specificities determined by this study were excellent (98% and 99%), and the sensitivities were very good (72% and 78%). These results that the TPI will be only rarely applied to patients who do not indeed have an acute coronary thrombosis. However, the reasons for even this small number of presumably falsely TPI positive patients should be determined and analyzed. It is unlikely that alterations of the thresholds for TPI activation will significantly improve on this very good level of sensitivity, without prohibitively decreasing specificity