Day-to-day variability in Doppler derived cardiac output measurements was assessed in 10 infants (0-6 months of age) and in 12 older children (1-13 years, mean: 7.4 years) with congenital heart disease which had been treated surgically. There was no valvular stenosis or ventricular septal defect, all patients had technically excellent imaging quality. Standard commercial equipment (Mark 8, ATL; 3, 5 or 7.5 MHz scanheads) was used for echo recordings, which were stored on video tapes and blindly evaluated by an independent investigator. Cardiac outout (CO) was measured over the aortic and pulmonary valves on two consecutive days using standard projections. Mean CO variability of the infants was 32% and 41% for the aortic and pulmonary valves, respectively. The corresponding variability in CO in the older children was 41% and 32% for the aortic and pulmonary valves respectively. The greatest difference existed in measuring cross-sectional areas of aorta and pulmonary artery (29-37%). The Doppler parameters showed the lowest variability (13-26%), except for the pulmonary artery in the infants (46%). On heart rate, there was a "physiological" day-to-day variability of 10-20%. CONCLUSIONS There is a considerable variability in Doppler derived CO measurements. The detection of CO changes in critically ill children might be missed by this error. CO measurements over the aortic valve seem to be more reliable than over the pulmonary valve.