Objective: To assess the clinical utility of an ultrasonic monitor of cardiac output (USCOM), its reliability in tracking cardiac output (CO) changes and agreement with thermodilution (TD) measurements of CO.
Design: Prospective comparison study.
Setting and participants: 55 adults undergoing thermodilution (TD) CO monitoring in a cardiothoracic or general intensive care unit between December 2006 and December 2007.
Main outcome measures: USCOM and TD measurements of CO on two occasions in each patient were compared by Bland-Altman analysis for bias and limit of agreement. A mean percentage error <30% was considered acceptable. Per cent change in cardiac index (CI) was determined by each method. Doppler profiles obtained by the USCOM were assessed against an ideal standard ("acceptable").
Results: 55 patients had measurements on 110 occasions, but Doppler waveforms were not obtained on 18 of these (16%), leaving 39 patients with paired comparisons for analysis (including 27 men; mean age, 64.7 [SD, 14.5] years). Mean TD CI was 3.4 +/-1.0L/min/m(2) (range, 2.0-6.0L/min/m(2)). The bias was 0.6L/min/m(2) (95% confidence limits [CLs], 0.4-0.8 L/min/m(2)), and the mean percentage error was 56% (95% CLs, 45%-65%). Twenty-two Doppler profiles (28%) were classed as acceptable; the mean percentage error for these was 62% (95% CLs, 38%-65%). On 15/19 occasions (74%) where TD CI changed > 15%, USCOM CI also changed >15%, but three of these changes (16%) were in the opposite direction. USCOM CI changed >15% on 9/20 occasions (45%) when TD CI did not.
Conclusions: Poor agreement with TD and a substantial rate of failure to obtain an USCOM measurement suggest that this device is unsuitable as a monitoring tool in intensive care.