The dilemma of sensitivity versus specificity in computer-interpreted acute myocardial infarction

The use of thrombolytic therapy and out-of-hospital electrocardiogram (ECG) acquisition capability has put even greater importance on the diagnostic accuracy of computerized ECG interpretation programs. Such programs must have extremely high specificity to minimize the possibility of clinicians treating inappropriate patients; thereby needlessly subjecting the patients to the risk of potentially life-threatening complications of the medication. At the same time, studies have shown that both prehospital personnel and emergency department (ED) physicians are aided by automated ECG interpretation programs with high sensitivity. These programs assist the attending personnel in rapidly identifying patients with suspected acute coronary thrombosis, which might otherwise have been undetected or not diagnosed until more obvious ECG abnormalities were present. In previous studies, clinically correlated databases have been used to develop and test sensitivity and specificity of the acute infarction detection algorithm in the Marquette 12SL ECG interpretation program. One program revision resulted in a marked increase in sensitivity (21-53%) without loss of specificity (99.5% to 99%). More recent studies have shown the sensitivity of the interpretation program to be influenced greatly by infarct location with sensitivity lower in anterior than inferior injury. Further refinement of the acute infarction interpretation criteria along with the methodology and data used are presented. Increased sensitivity without appreciable loss of specificity has been possible for detection of both acute inferior and anterior infarction; however, different methods were used for each location. Consideration of reciprocal or concomitant repolarization changes are found to be more useful for inferior than anterior injury. Methodological approaches are presented as they relate to the compromise between sensitivity and specificity.