Ultra-short heart rate variability recordings: Are they reliable for use with resting sows?

The objective of this study was to determine whether ultra-short heart rate variability (US-HRV) datasets (30-, 60-, 120-, 180-, and 240-s) were comparable with short-term heart rate variability (ST-HRV) datasets (300 s) collected from individually- and group-housed inactive sows. A total of 62 ST-HRV datasets were collected from 32 inactive sows (parity: 3.2 ± 0.25 SE) while housed in individual farrowing stalls during the nursing period (i.e. baseline period; n = 26 datasets) and after reintroduction to group housing (i.e. post-mixing period; n = 36 datasets) within a commercial swine production environment. Each ST-HRV dataset was divided into ultra-short-term dataset lengths (30, 60, 120, 180, and 240 s), resulting in a total of 62 ST-HRV datasets and 310 US-HRV datasets. Average inter-beat interval length (RR), root mean square of successive differences (RMSSD), and standard deviation of RR intervals (SDNN) were calculated for each dataset. Short-term HRV datasets were then compared with their respective US-HRV datasets using Pearson correlations, Bland-Altman analyses, and Friedman tests of equality. Using the most stringent criteria, (i.e., Pearson > 0.9, Bland-Altman 5% difference threshold, Friedman test P > 0.05), RMSSD can be reliably estimated using 240 s of data. Alternatively, 180-s datasets provide reliable, but less robust, RMSSD results where the Bland-Altman 10% difference threshold is reached while maintaining a Pearson correlation coefficient above 0.9 and comparable Friedman test results (P > 0.05). No US-HRV dataset length was capable of producing comparable Friedman Test results for RR. Finally, no US-HRV dataset length was sufficient for SDNN to reach the Bland-Altman 5% difference threshold. However, the Bland-Altman 10% difference threshold, a Pearson correlation coefficient ≥ 0.9, and comparable SDNN Friedman Test results were obtained with 240 s of data.