Abstract
BACKGROUND: Biochemical monitoring of cerebral oxygen metabolism is essential in the management of cranial brain injury (CBI). Jugular venous blood gas (JVBG) analysis provides real-time data reflecting cerebral biochemical states, offering a valuable window into brain oxygenation and metabolic demands. This study aimed to evaluate JVBG biochemical markers as indicators of CBI severity and as predictors of prolonged intensive care unit (ICU) stay, emphasizing their integration into machine learning models for clinical utility. METHODS: A retrospective analysis was conducted on 100 ICU-admitted CBI patients. Serial measurements of JVBG parameters-jugular venous oxygen saturation (SjvO2), partial pressure of oxygen (PjO2), arteriovenous oxygen difference (AVDO2), and oxygen content difference (AVDL)-were performed over five days. Spearman correlation and random forest algorithms were employed to assess relationships between JVBG biomarkers, clinical severity (via Glasgow Coma Scale), and ICU duration. RESULTS: Patients with more severe CBI exhibited significantly reduced SjvO2 and PjO2, and elevated AVDO and AVDL (p < 0.001). Strong correlations were found between JVBG markers and clinical severity scores. A multivariable prediction model incorporating age, SaO2, and PaCO2 at one day post-admission yielded excellent predictive performance for prolonged ICU stay (AUC = 0.974; sensitivity = 100%; specificity = 94.8%). CONCLUSIONS: JVBG biomarkers serve as clinically informative biochemical indicators for assessing CBI severity and forecasting ICU duration. Their integration into predictive algorithms may enhance precision in neurocritical care and support biochemical decision-making in intensive care settings.