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 (alpha(S)), long-term DFA exponent (alpha(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, EKG(ref)) 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<0.001). HRV metrics from EKG250 and EKG(DWH) were also significantly correlated (r=0.7, P<0.001) for all metrics except for HF (r=0.276). These data show that HF power is compromised in the EKG(DWH) signal and caution must be exercised in interpreting the HF power calculated from this EKG.