Detecting sleep apnea using non-linear measures of heart rate variability

Abstract Background Sleep apnea is highly prevalent yet frequently underdiagnosed due to the cost and complexity of polysomnography. Heart rate variability (HRV) offers a scalable alternative for early screening, but conventional HRV metrics often overlook the multi-scale autonomic disturbances characteristic of apnea. Methods We evaluated scale-dependent detrended fluctuation analysis (sDFA), which quantifies how heartbeat-interval correlations evolve across temporal scales, using RR-interval data from Sleep Heart Health Study ( $$n=5804$$ ). The discriminative performance of sDFA was compared with conventional HRV measures across mild, moderate, and severe apnea, and within cardiovascular disease (CVD) subgroups. Propensity score matching was applied for age and body mass index, and analyses were stratified by sex. Results Across apnea severity levels, sDFA consistently outperformed conventional HRV measures in discriminating individuals with sleep apnea. Performance gains were particularly evident in severe apnea and remained robust in participants with CVD, a subgroup in which traditional HRV metrics showed reduced discriminative ability. sDFA revealed scale-specific signatures of autonomic dysfunction that were not captured by conventional time- and frequency-domain HRV measures. Conclusion Multi-scale analysis of HRV using sDFA enhances the detection of sleep apnea across severity levels and cardiovascular risk profiles. These findings highlight the limitations of conventional HRV metrics and support sDFA as a promising tool for scalable, HRV-based sleep apnea screening, with potential for integration into wearable and ambulatory monitoring systems.