Cystatin C for Early Mortality Prediction in First msTBI: Insights from Traditional and Machine Learning Approaches

Cystatin C, a cysteine protease inhibitor, has been implicated in various central nervous system disorders, yet its clinical significance in traumatic brain injury (TBI) remains incompletely understood. This study aimed to evaluate the association between serum cystatin C levels and early mortality in patients with moderate to severe TBI (msTBI) admitted to the ICU, and to develop a machine learning-based predictive model for 28-day all-cause mortality. A total of 369 patients were included and categorized into tertiles according to their serum cystatin C concentrations. Multivariate Cox analysis confirmed that elevated cystatin C levels were independently associated with significantly increased risks of both 28-day and 90-day mortality in msTBI patients (HR > 1, P < 0.05). Using the Boruta algorithm, cystatin C was identified as the fourth most important predictive variable. A Cox proportional hazards (coxph) model achieved the best predictive performance, with an area under the curve (AUC) of 0.8644 in the training set and 0.8422 in the test set. These findings indicate that high cystatin C levels are independently associated with early mortality in msTBI, highlighting its potential utility as a prognostic biomarker. The developed model may serve as a practical tool for clinical risk assessment and inform intervention strategies.