Effect of EKG Sampling Rate on Heart Rate Variability Analysis

We studied the effect of EKG sampling rate on heart rate variability (HRV) analysis. We acquired EKG from four term hypoxic-ischemic encephalopathic infants undergoing therapeutic hypothermia. The EKG signal was acquired continuously for 4 days from the cardiorespiratory monitor through the analog port. The following HRV metrics were calculated: normalized low-frequency (nLF), normalized high-frequency (nHF), low-frequency (LF), high-frequency (HF), short-term detrended fluctuation analysis (DFA) exponent (α s ), long-term DFA exponent (α L ), root mean square (RMS) short (RMSS), and RMS long (RMSL). In addition, heart rate was used. These metrics were calculated for EKG acquired at 1 KHz (served as reference, EKGref) as well as from EKGs downsampled at 500 Hz (EKG 500 ), 250 Hz (EKG 250 ), and 125 Hz (EKG 125 ). The bedside monitors were simultaneously sending the EKG to a data warehouse, storing the EKG (EKG DWH ) at 250 Hz. All HRV metrics were also calculated for the EKG DWH . The comparison between HRV metrics calculated from EKG ref and downsampled EKG (EKG 500 , EKG 250 , EKG 125 ) was made with intraclass correlation coefficient (r). The comparisons of HRV metrics between EKG 250 and EKG DWH were also made with ICC. Our results show that HRV calculated with EKG ref and from downsampled EKG were highly correlated (r>0.8 for all comparisons, P 250 and EKG DWH were also significantly correlated (r=0.7, P DWH signal and caution must be exercised in interpreting the HF power calculated from this EKG.