Uses of statistical editing of real-time ambulatory electrocardiographic recordings for quantitative ventricular...

Real-time ambulatory monitoring analyzes each heart beat, counts events, and stores ECG samples for later visual verification. Typically, a physician examines these to determine whether the computer algorithm accurately identified arrhythmias. Physician editing is performed using best clinical judgement. We developed a simple statistical editing procedure for adjusting false positive and false negative computer detections. In 20 subjects having 24-hr monitoring we compared statistically edited ventricular premature beat (VPB) counts and pair/run counts with the unedited monitor counts and with physician assessment using a visual counted gold standard. The agreement of the statistically edited count with the visual standard was 65% for total VPB, 85% for VPB pair/runs, and 90% for a risk score based on ventricular ectopy. Corresponding agreements for unedited monitor count were 15, 25, and 30%, respectively. Physician assessment was not sufficiently precise to allow quantitative count estimates. This study indicates a statistical editing procedure substantially increases the level of agreement between the visual standard and the monitor count of VPB frequency and complexity. Statistically edited data are suitable for quantitative counts of VPB and other arrhythmic events in research and in medical diagnosis and treatment. This editing procedure can be a useful adjunct to any ambulatory monitoring system.