Sleep-architecture signatures for differential diagnosis of schizophrenia, bipolar, and major depressive disorders: A...

Sleep architecture is an essential lifestyle factor influencing health and well-being, and it may offer a trans-diagnostic window to understand the shared pathophysiology across schizophrenia (SCH), bipolar disorder (BD), and major depressive disorder (MDD). However, direct comparative evidence of sleep architecture features among these three major psychiatric disorders remains scarce, especially among patients in the first-episode phase.

Whole-night polysomnography (PSG) was performed in 109 SCH, 107 BD, and 258 MDD patients. All patients were first-episode and were drug-naive or had received minimal treatment prior to enrollment. LASSO regression-selected demographic, clinical, and PSG variables entered multinomial logistic models. Receiver Operating Characteristic (ROC), calibration, and decision-curve analyses were used to assess performance.

LASSO retained BMI, course, sleep efficiency (SE), REM latency (REML), and N2%. Shorter disease course, higher BMI, lower SE and shorter REML differentiated SCH from MDD (all p < 0.05); higher BMI, longer course, shorter REML, and elevated N2% separated BD from MDD (all p < 0.05). The receiver operating characteristic curves (AUC) were 0.750 (MDD vs SCH) and 0.717 (MDD vs BD). The decision curves showed net clinical benefit across the wider threshold probabilities.

A brief, five-variable panel-integrating three key PSG parameters with BMI and disease course-demonstrated modest but clinically significant discrimination among SCH, BD, and MDD, supporting sleep-based adjunctive tools that refine differential diagnosis and advance precision psychiatry.