Network Structure and Model Projection Analysis of Depressive and Anxiety Symptoms, Sleep Disturbances, and Medical Coping Styles in Hospitalized Patients With Coexisting Hypertension and Diabetes: A Multicenter Cross-Sectional Study

Background Patients with coexisting hypertension and diabetes mellitus commonly experience depressive and anxiety symptoms, sleep disturbances, and maladaptive coping. However, the conditional associations among psychological, sleep, and coping factors, as well as their projected effects under prespecified model perturbations, remain unclear. Guided by the Theory of Unpleasant Symptoms, this study examined the network structure of these factors and their model-implied responses to simulated perturbations. Methods This multicenter cross-sectional study included 529 hospitalized patients with coexisting hypertension and diabetes mellitus. A 26-node ordinal Gaussian graphical model (GGM) was estimated using the Patient Health Questionnaire-9 (PHQ-9), Generalized Anxiety Disorder-7 (GAD-7), Pittsburgh Sleep Quality Index (PSQI), and Medical Coping Modes Questionnaire (MCMQ). Expected influence, bridge expected influence, predictability, and network stability were assessed. Model-based projections across multiple perturbation levels were subsequently evaluated using the Gaussian Graphical Model NodeIdentifyR Algorithm (GGMNIRA). Projected effects under alleviation- and aggravation-oriented perturbations were further compared in a calibrated 23-node Ising network using the NodeIdentifyR algorithm (NIRA). Results The 26-node network contained 152 nonzero edges. Trouble relaxing, restlessness, and negative self-evaluation exhibited relatively high expected influence, whereas thoughts of self-harm or death, fear that something awful might happen, and trouble relaxing were among the leading nodes in bridge expected influence. In GGMNIRA, the five highest-priority nodes all belonged to the anxiety community, with trouble relaxing ranking first; only the anxiety-to-depression pathway met the model-defined decision threshold. In Ising-NIRA, fatigue and sleep problems ranked highly under alleviation-oriented perturbations, whereas fear that something awful might happen and restlessness ranked highly under aggravation-oriented perturbations. Indirect changes in other nodes accounted for approximately 87%–88% of the total projected change associated with high-priority nodes. No robust evidence of network heterogeneity was identified. Conclusions An anxiety symptom cluster characterized by worry, trouble relaxing, and restlessness was prominent in both the network structure and model-based projections. Fatigue and sleep problems showed relatively high projected priority under alleviation-oriented perturbations, whereas fear that something awful might happen and restlessness ranked highly under aggravation-oriented perturbations. These findings suggest that improvement and worsening of psychological and sleep symptoms may be associated with distinct network patterns.