Cross-sectional study of anxiety-cognition network patterns in rehabilitation patients using network analysis

This study aimed to explore the network relationship between anxiety and cognition and its mechanism, providing a theoretical basis for formulating anxiety-related intervention strategies. Data were collected from 146 patients in the Rehabilitation Department of Dongfang Hospital, Beijing University of Chinese Medicine. The Hamilton Anxiety Scale (HAMA) and Montreal Cognitive Assessment (MoCA) were used to measure 14 anxiety dimensions and 7 cognitive dimensions, respectively. A regularized partial correlation network (Gaussian graphical model) was estimated using the graphical LASSO with the extended Bayesian information criterion (EBICglasso, γ = 0.5) on a cor_auto correlation matrix, and centrality, edge accuracy, and network stability were assessed using nonparametric bootstrapping. Network analysis showed that the anxiety dimensions of Behavior at Interview, Depressed Mood, Anxious Mood, Fears, and Tension, and the cognitive dimensions of Attention and Abstraction had higher centrality. The anxiety dimensions of Cognitive Impairment and Behavior at Interview had the highest betweenness in the "anxiety-cognition" network, while Genitourinary Symptoms had smaller strength and closeness compared to other network dimensions. The CS-coefficient (centrality stability coefficient) values were 0.596 for expected influence, 0.363 for strength, 0.281 for closeness, and 0.048 for betweenness. The network structure results were stable. In anxiety, Behavior at Interview, Depressed Mood, Anxious Mood, Fears, and Tension were key intervention candidate nodes. Because the cross-domain edges anchored on HAMA5 (Cognitive Impairment) partly reflect the construct overlap between subjective and objective cognition, the anxiety-cognition findings most directly relevant for intervention design are those connecting affective and behavioral HAMA items to specific MoCA dimensions.