Objective: To test the ability of a logistic regression model (LRM) that predicts acute cardiac ischemia to make an early diagnosis of acute myocardial infarction (AMI); the ability of the LRM to predict AMI was also compared with the presenting electrocardiogram (ECG).
Setting: A small rural Irish coronary care unit.
Methods: Clinical and ECG data required by the LRM to predict acute coronary ischemia were recorded in 600 consecutive patients admitted with suspected AMI. Estimates of the LRM were ranked into equal deciles in declining probability of acute cardiac ischemia (pACI), and presenting ECGs were placed into one of seven categories.
Results: At presentation 50% of AMI patients were in the two LRM deciles with the highest pACI, and 49% of AMI patients had ECGs with greater than 2 mm ST elevation associated with reciprocal changes. ECG categories had a 76% sensitivity for the early diagnosis of AMI and the LRM had an 84% sensitivity. The specificity, accuracy and positive predictive value for the ECG categories were 92%, 84% and 85%, respectively. The specificity, accuracy and positive predictive value of the LRM were 84%, 84% and 75%, respectively. The areas under the receiver operating characteristic curve of the LRM and ECG categories were almost identical (91% and 90%, respectively).
Conclusion: AMI can be diagnosed early with comparable accuracy either by placing presenting ECGs into one of seven categories, or by the LRM. The best method and 'cut-off' point for the diagnosis of AMI varies according to clinical circumstances. Categorizing ECGs requires more skill in ECG interpretation, but takes less time. The previously reported performances of the LRM were replicated, confirming portability of its use into different clinical settings and patient populations.