Data-driven outcome prediction and rational resource allocation in severe traumatic brain injury: a multicenter retrospective study integrating survival analysis and machine learning ensembles.

Abstract

Introduction: Severe traumatic brain injury (sTBI), is a leading cause of death and disability among young and middle-aged populations worldwide.

Objective: sTBI is associated with high mortality and disability rates, and reliable risk estimation with quantified uncertainty is crucial for optimizing intensive care unit (ICU) bed allocation and nursing resource deployment. This study aimed to establish a prognostic model for sTBI patients using multicenter clinical, imaging, and nursing data, and to explore its value in guiding clinical resource allocation.

Methods: A total of 1000 sTBI patients from multiple medical centers were enrolled in this retrospective cohort study. Clinical data including vital signs, CT/MRI imaging features, and detailed neurosurgical nursing records were collected.

Results: Multivariate Cox regression analysis identified age, GCS score on admission, pupillary light reflex, CT findings such as midline shift, and hematoma volume, mean arterial pressure, and nursing-related indicators such as incidence of pulmonary infection, and pressure ulcer grade as independent prognostic factors for sTBI patients. The resource allocation strategy guided by risk stratification significantly reduced ICU bed occupancy rate from 89.2% to 75.6% and improved the utilization efficiency of specialized nursing resources, while the overall prognosis of patients was not negatively affected.

Conclusion: The ML ensemble model constructed based on multicenter clinical, imaging, and nursing data has high accuracy in predicting short-term and long-term outcomes of sTBI patients. Risk stratification using this model can provide a scientific basis for rational allocation of ICU beds and resources, which is of great significance for improving the overall treatment level of sTBI patients.