Dimension-level network structure linking depression, anxiety, stress, sleep problems, and problematic smartphone use among chinese medical students.

Background: Medical students experience converging risks of emotional distress, sleep disturbance, and problematic smartphone use, but the dimension-level conditional association patterns linking these domains remain insufficiently specified. Methods: This cross-sectional study surveyed 2,587 Chinese medical students (mean age = 18.88 ± 1.01 years; 55.51% female) using the 21-item DASS-21, the PSQI, and the MPAI. Regularized Gaussian graphical models were estimated with EBICglasso for the DASS-MPAI, PSQI-MPAI, and integrated DASS-PSQI-MPAI networks. Strength, bridge strength, node predictability, bootstrapped stability, and gender-based network differences were examined. Nodes represented DASS-21 dimensions, PSQI components, and MPAI dimensions rather than individual questionnaire items. Results: Anxiety was the most prevalent emotional distress dimension (39.89%), followed by depression (34.60%) and stress (15.58%). Sleep problems were detected in 23.42% of participants, whereas problematic smartphone use was detected in 64.71%. Across networks, nodes clustered into clearly differentiated emotional distress, sleep, and problematic smartphone use modules, with stronger within-domain than cross-domain edges. In the DASS-MPAI network, stress and withdrawal showed the highest strength, whereas depression and stress showed the highest bridge strength. In the PSQI-MPAI network, withdrawal and inefficiency were the strongest central nodes, and sleep disturbance and loss of control showed the highest bridge strength. In the integrated network, anxiety and stress showed the highest strength, followed by inefficiency and withdrawal. Bridge strength identified sleep disturbance (0.264), daytime dysfunction (0.239), and anxiety (0.235) as the most prominent cross-domain bridge nodes. Bootstrap analyses supported network stability; the integrated network centrality indices showed acceptable-to-good stability. Gender comparisons revealed no significant difference in global strength (P = 0.728), but the omnibus network structure test was significant (P = 0.010). Conclusions: This study provides a dimension-level map of conditional associations among emotional distress, sleep problems, and problematic smartphone use in a single-institution convenience sample of Chinese medical students. Anxiety, stress, sleep disturbance, daytime dysfunction, inefficiency, and withdrawal emerged as central or bridge nodes in the observed networks. These findings should be interpreted as exploratory cross-sectional associations rather than causal relationships or confirmed intervention targets, but they may inform hypotheses for future longitudinal and intervention studies.