Constructing the conditional association network and identifying central and bridge nodes in aneurysmal subarachnoid hemorrhage.

Aneurysmal subarachnoid hemorrhage (aSAH) ranks among the most burdensome stroke subtypes in terms of years-of-life-lost. Existing models rely primarily on linear additive structures; explicit visualisation of factor-to-factor conditional dependencies remains limited. This study employed mixed graphical models (MGM) to restructure established clinical variables within a cross-sectional network framework, examining conditional independence structures among in-hospital clinical variables. This single-centre retrospective observational cohort included 374 aSAH patients admitted to a neurosurgical ward following definitive aneurysm treatment. The primary outcome was the discharge modified Rankin Scale (mRS), and a static network was constructed using data accrued over the entire hospitalization. Thirty-five routinely documented variables spanning demographics, severity scores (Glasgow Coma Scale, WFNS, modified Fisher), imaging characteristics, laboratory values, treatment modalities, in-hospital complications, temperature profiles and mRS underwent contemporaneous network analysis. MGM estimated pairwise conditional associations via LASSO-regularized Markov random fields (EBIC γ = 0.25, OR rule). Node strength centrality and bridge strength centrality quantified variable interconnectedness. Non-parametric bootstrap resampling (n = 1000) assessed edge weight stability. The cohort (mean age 53.9 ± 11.1 years; 60.7% female) exhibited dense interconnections among severity scores, temperature-derived indices, in-hospital complications, and therapeutic variables. Discharge modified Rankin Scale, Glasgow Coma Scale, mechanical ventilation and pneumonia occupied the highest ranks in both strength centrality and bridge strength centrality; these variables functioned as central and bridge nodes spanning the severity assessment, therapeutic intervention, and outcome evaluation communities. Conditioning on the multivariate conditional structure, the GCS-mRS relationship exhibited a pattern differing from the simple bivariate association. MGM restructured known aSAH clinical variables within a network topology, identifying Glasgow Coma Scale, mechanical ventilation and pneumonia as central and bridge nodes spanning multiple clinical domains. These central nodes represent candidates for future hypothesis-generating research, pending prospective validation.