Latent trajectories of cerebral perfusion pressure and risk prediction models among patients with traumatic brain injury: based on an interpretable artificial neural network
Abstract Objective: This study aimed to characterize long-term cerebral perfusion pressure (CPP) trajectory in traumatic brain injury (TBI) patients and construct an interpretable prediction model to assess the risk of unfavorable CPP evolution patterns. Methods: TBI patients with intracranial pressure (ICP) monitoring and artery blood pressure records were identified from the Medical Information Mart for the Intensive Care (MIMIC)-IV 2.1 and eICU Collaborative Research Database (eICU-CRD) 2.0. The research process consisted of two stages. First, group-based trajectory modeling (GBTM) was used to identify different CPP trajectories. Second, different ANN algorithms were employed to predict the trajectories of CPP. Results: A total of 331 eligible patients were ultimately enrolled in the study. The GBTM classified patients into 5 CPP trajectory groups, including declined then rose (13.3%), low then rose (26.6%), moderate (31.7%), rapidly elevated (25.7%), persistently high (2.7%). Group 1 and Group 5 trajectories were merged into Class 1 based on in-hospital mortality, and the Boruta algorithm was used to identify features that distinguish it. The best 6 predictors were invasive systolic blood pressure coefficient of variation (ISBPCV), venous blood chloride ion concentration, PaCO2, PT (Prothrombin Time), CPP coefficient of variation (CPPCV), and mean CPP. Compared with other algorithms, Scaled Conjugate Gradient (SCG) performed relatively better in identifying the unfavorable trajectory class. Conclusion: This study identified two CPP trajectory groups associated with elevated risk and three with reduced risk. PaCO2 might be a strong predictor for the unfavorable CPP class. The ANN model achieved the primary goal of risk stratification, which is conducive to early intervention and individualized treatment. Keywords: Cerebral perfusion pressure trajectories; Traumatic brain injury; Artificial neural network; Group-based trajectory modeling