Single-beat determination of left ventricular systolic function at a beat with equal subsequent cardiac cycles has been proposed as an accurate method in atrial fibrillation. However, there has still been substantial variability between the values calculated from beats with equal subsequent cycles. Therefore, some refinement on the single-beat method is needed. In 100 patients with atrial fibrillation, Doppler aortic flow time-velocity integral was determined for at least 20 consecutive cardiac cycles. The values at beats with equal subsequent cardiac cycles were chosen and compared with the average values over all cardiac cycles. The values at beats with cycle lengths shorter than 500 milliseconds were usually far below the average values over all cardiac cycles. Bland-Altman agreement analysis revealed improved accuracy by gradually narrowing the range of the limits of agreement when 2 or 3 beats with equal subsequent cycles and cycle lengths longer than 500 milliseconds were used for evaluation.