Automatic real time arrhythmia monitoring in the intensive coronary care unit.

A prospective study was carried out to evaluate accuracy of a real time computer system for arrhythmia monitoring in a coronary care unit. QRS complexes in 57 continuously recorded 30 minute electrocardiographic rhythm strips from 30 patients were classified by the computer. Blinded interpretation of each beat was made by a cardiologist. A total of 146,638 QRS complexes were classified during the monitoring period of 1,710 minutes. The accuracy rate of computer detection was 99.94%. During this period 67 artifacts were identified as beats. The computer correctly identified 5,467 of 5,729 ventricular premature beats (accuracy of detection rate 95.46%). The incidence rate of false positive diagnosis of beats other than ventricular ectopic beats was 0.1%. The rate of computer accuracy in detecting ventricular premature beats was 93.33% in two patients with atrial fibrillation. In individual patients, computer interpretation did not deviate from the cardiologist's interpretation by more than 1 beat/min. The rate of computer accuracy in detecting supraventricular premature beats was lower (82.45%). The distinctly improved accuracy of computer monitoring of ventricular arrhythmia in a coronary care unit reported is the result of a new algorithm. This approach combines major elements of cross correlation of QRS signals as well as grouping of complexes into families. Normal and abnormal templates are stored and newly detected QRS complexes are matched. The program appears to be sufficiently accurate to warrant further applications.