Reliability Loss with Sampling Rate Reduction

Abstract

High sampling frequency is not usually available in hospital monitoring systems, what can limit the usefulness of the data, namely for repolarization measures. In this work the reliability of beat-to-beat measures using low sampling frequency is quantified with respect to the original high sampling rate. ECG recordings originally at 1000 Hz, including intensive care patients data, were downsampled to 500 Hz and to 250 Hz. Automatic delineation was applied to extract beat-to-beat intervals. The reliability was measured considering concordance correlation coefficient (CCC) and intraclass correlation coefficient (ICC); the Information-based measure of disagreement (IBMD) was used for agreement quantification. High reliability and low disagreement were generally found. Using the conventional interpretation excellent consistency was found at 250 HZ for more than 68%, 50%, 50% and 58% of the cases for RR, RT QT and QT(c), respectively. Results indicate that at 250 Hz RR, RT QT and QT(c) measures are still reliable compared with the values at 1000 Hz.