Abstract
BACKGROUND: Ultra-short-term heart rate variability (usHRV) has been found to be associated with atrial fibrillation (AF); however, research in this area is currently limited. OBJECTIVE: This study aimed to investigate the association between usHRV metrics and AF. METHODS: This retrospective cohort study included 48,416 participants from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. UsHRV time domain and frequency domain metrics were also collected. We examined the connection between usHRV and AF in electrocardiogram samples collected from 08:00 to 18:00, 18:00 to 08:00, and all day to understand the impact of time on the findings. To address the research objectives, we used Cox regression analysis, stratified curve fitting, threshold effect analysis, subgroup analysis, and the assessment of interaction effects. RESULTS: During an average follow-up of 1.88 years, 3611 (7.5%) participants developed AF. UsHRV time domain metrics were unstable in a day, and the HR was more significant in those ≥55 years of age, with a statistically significant interaction. In contrast, the usHRV frequency metrics are more clinically significant and stable. The hazard ratios for the 08:00 to 18:00 samples were 0.79 (95% confidence interval [CI] 0.74-0.84) for log(ratio of low frequency and high frequency), 0.74 (95% CI 0.67-0.80) for log(low-frequency normalized units), and 2.26 (95% CI 1.80-2.84) for log(high-frequency normalized units), respectively. CONCLUSION: The frequency domain metrics of usHRV exhibit strong stability, surpassing those derived from time domain metrics, and offer improved convenience compared with HRV. This makes them particularly notable for their clinical significance.