Sleep underestimation and subjective sleep appraisal: Limited value of routine polysomnographic and electroencephalographic features.

OBJECTIVES

Whether routine polysomnographic (PSG) and electroencephalographic (EEG) features distinguish patients who underestimate sleep remains unclear. We examined associations of clinical, subjective, PSG, and EEG features with sleep underestimation and tested whether EEG features added discrimination beyond a parsimonious three-predictor baseline model.

METHODS

We analysed 225 adults undergoing clinically indicated in-laboratory PSG in this single-centre retrospective cross-sectional study. The misperception index ([objective - subjective total sleep time]/objective total sleep time) was dichotomised at the cohort median (0.148) and analysed continuously. Group differences across 717 variables underwent Benjamini-Hochberg false discovery rate (FDR) correction. Out-of-fold (OOF) discrimination was compared between a three-predictor baseline model (sleep quality, Epworth Sleepiness Scale score, and age) and an EEG-enhanced model; continuous-index analyses were exploratory.

RESULTS

Fifty-six participants (24.9%) overestimated sleep. Objective total sleep time did not differ between groups, whereas self-rated sleep quality did (q = 0.002). None of 600 EEG features survived FDR correction (smallest q = 0.064). At the median threshold, OOF areas under the receiver operating characteristic curve (AUCs) were 0.6786 (baseline) and 0.6564 (EEG-enhanced; ΔAUC = -0.0221; paired DeLong P = 0.4483). Across 50 repeated partitions, median ΔAUC was -0.0261 for EEG features and +0.0096 for four macrostructural PSG variables. Obstructive sleep apnoea prevalence did not differ significantly between groups.

CONCLUSIONS

Self-rated sleep quality was more consistently associated with sleep underestimation than routine PSG and EEG features. EEG features did not demonstrate incremental OOF discrimination beyond the baseline model. Subjective sleep appraisal should be assessed alongside PSG; underlying mechanisms require prospective investigation.