Background <p>Distinguishing bipolar depression (BD) from unipolar depression (UD) in adolescents is clinically challenging because depressive symptoms overlap and reliable objective markers are lacking. Sleep and heart rate variability (HRV) have been implicated in mood disorders, but their value for differentiating adolescent BD from UD remains insufficiently studied. This study compared clinical characteristics, device-derived sleep metrics, and nocturnal HRV indices between adolescents with BD and UD, and explored whether physiological measures provided additional information associated with BD classification.</p> Methods <p>In this single-center retrospective cross-sectional study, 110 inpatients aged 11–17 years with a current depressive episode were included, comprising 46 patients with BD and 64 with UD. Overnight portable sleep monitoring was performed during early inpatient assessment. Clinical characteristics, device-derived sleep parameters, and nocturnal HRV indices were extracted from medical records and device-generated reports. Between-group comparisons were corrected for multiple testing using the Benjamini-Hochberg false discovery rate. Hierarchical logistic regression models were used to examine the relative contribution of clinical variables, sleep metrics, and HRV indices. An exploratory parsimonious model was internally evaluated using repeated cross-validation and bootstrap resampling.</p> Results <p>Compared with the UD group, the BD group had a longer illness duration (24.0 [12.0, 48.0] vs. 12.0 [6.0, 26.0] months, q = 0.042) and a higher prevalence of self-harm history (93.5% vs. 68.8%, q = 0.031). No selected single-night device-derived sleep metric differed significantly between groups after false discovery rate correction. In contrast, several nocturnal HRV indices, including time-domain, frequency-domain, and nonlinear measures, were significantly lower in the BD group. In the full hierarchical model incorporating clinical, sleep, and HRV variables, lower log-transformed total HRV power was the only physiological measure independently associated with BD classification (odds ratio per 1-standard-deviation increase = 0.34, 95% confidence interval: 0.16–0.69, <i>p</i> = 0.003). An exploratory parsimonious model including illness duration, self-harm history, and log-transformed total HRV power yielded an apparent area under the receiver operating characteristic curve of 0.820 and an optimism-corrected value of 0.807.</p> Conclusions <p>Adolescents with BD showed reduced device-derived nocturnal HRV compared with those with UD, particularly lower total HRV power, whereas the selected single-night portable sleep metrics showed limited discriminative value in this dataset. Reduced nocturnal HRV may represent an auxiliary objectively measurable physiological correlate for future research on the differential assessment of adolescent depressive episodes. Prospective studies with standardized diagnostic assessment, repeated or longitudinal physiological monitoring, and external validation are needed before clinical application.</p>