Abstract Introduction Insomnia is among the most prevalent sleep disorders, yet current clinical frameworks struggle to explain why many patients report poor sleep despite apparently normal polysomnography (PSG). This mismatch, known as subjective–objective sleep discrepancy (SOSD), complicates diagnosis and treatment. Conventional staging collapses sleep into categorical states and may overlook continuous dynamics. This study introduced a probabilistic, information-theoretic approach to characterise sleep as a continuum, aiming to explain insomnia subtypes and the mechanisms underlying SOSD. Methods PSG recordings from 904 individuals with chronic insomnia and 104 good sleepers were analysed. Instead of assigning each epoch to a single stage, a hypnodensity approach generated probability distributions across Wake, N1, N2, N3 and REM. From these time series, two metrics were derived. The first, intrusion, captured the extent of mixing across sleep–wake probabilities during a given epoch. The second, instability, quantified abrupt changes from one epoch to the next. Participants were grouped as good sleepers, insomnia without subjective–objective discrepancy, and insomnia with discrepancy. Group differences were tested using regression models, while machine-learning classifiers evaluated diagnostic performance. Associations with spectral EEG power and conventional PSG outcomes were examined. Results Both insomnia groups showed higher levels of intrusion and instability than healthy sleepers, indicating more mixed and unstable sleep–wake dynamics. However, the patterns underlying these abnormalities differed. Individuals with insomnia and SOSD exhibited prominent wake intrusions during sleep, accompanied by increased beta and gamma EEG activity, whereas those with insomnia but no SOSD displayed greater sleep intrusions during wakefulness, associated with elevated delta and theta activity. A model based solely on intrusion and instability measures reliably distinguished the three groups, achieving strong classification accuracy, and these markers predicted traditional PSG outcomes including sleep efficiency, WASO and total sleep time. Conclusion A continuous, probabilistic representation of sleep reveals that insomnia and subjective–objective discrepancy exist along a spectrum rather than as discrete categories. Intrusion and instability capture meaningful neurophysiological differences between insomnia phenotypes and provide an interpretable bridge between subjective complaints and objective PSG features. This framework offers a scalable approach for improved phenotyping, clinical stratification, and potentially treatment selection in insomnia. Support (if any)
Herzog et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: