Importance of re-calibration time on pulse contour analysis agreement with thermodilution measurements of cardiac output: a retrospective analysis of intensive care unit patients

We assessed the effect of re-calibration time on cardiac output estimation and trending performance in a retrospective analysis of an intensive care unit patient population using error grid analyses. Paired thermodilution and arterial blood pressure waveform measurements (N = 2141) from 222 patient records were extracted from the Multiparameter Intelligent Monitoring in Intensive Care II database. Pulse contour analysis was performed by implementing a previously reported algorithm at calibration times of 1, 2, 8 and 24 h. Cardiac output estimation agreement was assessed using Bland–Altman and error grid analyses. Trending was assessed by concordance and a 4-Quadrant error grid analysis. Error between pulse contour and thermodilution increased with longer calibration times. Limits of agreement were −1.85 to 1.66 L/min for 1 h maximum calibration time compared to −2.70 to 2.41 L/min for 24 h. Error grid analysis resulted in 74.2 % of points bounded by 20 % error limits of thermodilution measurements for 1 h calibration time compared to 65 % for 24 h. 4-Quadrant error grid analysis showed <75 % of changes in pulse contour estimates to be within ±80 % of the change in the thermodilution measurement at any calibration time. Shorter calibration times improved the agreement of cardiac output pulse contour estimates with thermodilution. Use of minimally invasive pulse contour methods in intensive care monitoring could benefit from prospective studies evaluating calibration protocols. The applied pulse contour analysis method and thermodilution showed poor agreement to monitor changes in cardiac output.