G Zhang1, R Mukkamala. 1. Department of Electrical and Computer Engineering, Michigan State University, 2120 Engineering Building, East Lansing, MI 48824-1226, USA.
Abstract
BACKGROUND: A potential practical approach for continuous and minimally invasive cardiac output (CO) monitoring in intensive care unit (ICU) patients is to mathematically analyse an arterial pressure (AP) waveform using an existing radial artery line ('pulse contour analysis'). We recently proposed a technique to estimate the relative CO change by unique long time interval analysis (LTIA) of an AP waveform. We aimed to test this technique in an ICU patient population and compare its accuracy relative to other techniques. METHODS: We studied a public, electronic ICU patient database. We extracted 1482 pairs of radial AP waveforms and thermodilution CO measurements (via single bolus injections) from 169 patients. We applied the LTIA and previous pulse contour analysis techniques to the AP waveforms. We assessed the calibrated CO estimates against the thermodilution measurements. RESULTS: The overall root-mean-squared-error of the LTIA technique was 18.8%. This total level of accuracy was not better than the previous techniques. However, the average magnitude of the thermodilution changes was only 12.3% (9.9 sd). When the magnitude of the thermodilution change exceeded 30%, 50%, and 70%, the median squared-error differences between the LTIA technique and the most accurate previous technique were -45 (-322:69 quartiles) (P=0.005), -128 (-704:23) (P=0.006), and -862 (-2871:306)%(2) (P=0.055), respectively. The LTIA technique was therefore superior in detecting clinically important CO changes. CONCLUSIONS: The LTIA technique attained an overall accuracy that may be considered clinically acceptable after taking into account the known thermodilution error and became progressively more accurate than previous techniques with increasing CO changes.
BACKGROUND: A potential practical approach for continuous and minimally invasive cardiac output (CO) monitoring in intensive care unit (ICU) patients is to mathematically analyse an arterial pressure (AP) waveform using an existing radial artery line ('pulse contour analysis'). We recently proposed a technique to estimate the relative CO change by unique long time interval analysis (LTIA) of an AP waveform. We aimed to test this technique in an ICU patient population and compare its accuracy relative to other techniques. METHODS: We studied a public, electronic ICU patient database. We extracted 1482 pairs of radial AP waveforms and thermodilution CO measurements (via single bolus injections) from 169 patients. We applied the LTIA and previous pulse contour analysis techniques to the AP waveforms. We assessed the calibrated CO estimates against the thermodilution measurements. RESULTS: The overall root-mean-squared-error of the LTIA technique was 18.8%. This total level of accuracy was not better than the previous techniques. However, the average magnitude of the thermodilution changes was only 12.3% (9.9 sd). When the magnitude of the thermodilution change exceeded 30%, 50%, and 70%, the median squared-error differences between the LTIA technique and the most accurate previous technique were -45 (-322:69 quartiles) (P=0.005), -128 (-704:23) (P=0.006), and -862 (-2871:306)%(2) (P=0.055), respectively. The LTIA technique was therefore superior in detecting clinically important CO changes. CONCLUSIONS: The LTIA technique attained an overall accuracy that may be considered clinically acceptable after taking into account the known thermodilution error and became progressively more accurate than previous techniques with increasing CO changes.