Paroxysmal atrial fibrillation prediction using isolated premature atrial events and paroxysmal atrial tachycardia

A simplified predictor of episodes of paroxysmal atrial fibrillation (PAF) is presented. Over a thirty minute monitoring time, the number of premature atrial complexes (PAC) without normal sinus node reset is counted. When the number of such PACs changes by more than 2 over the monitoring period, a PAF prediction is made. The PAC test is supplemented with a paroxysmal atrial tachycardia (PAT) test that uses a similar approach in identifying and counting episodes of PAF. The combined PAC and PAT test yields a sensitivity of 68%, a specificity of 86% and a positive predictivity of 86% on the public test PAF database from the 2001 Computer in Cardiology.