Establishment and evaluation of a predictive scoring system for successful decannulation of tracheostomy in neurosurgery patients.

Objective: To establish and evaluate a predictive scoring system for successful decannulation of tracheostomy in neurosurgery patients. Methods: A prospective observational study was conducted. Patients undergoing tracheostomy in the neurosurgery departments of two tertiary hospitals in Sichuan Province were enrolled and divided into training and validation sets based on admission time. Multivariate logistic regression was used to identify predictive factors for successful decannulation, and a scoring system was developed based on the regression coefficients. Model performance was assessed using area under the receiver operating characteristic curve (AUC), calibration plots, and decision curve analysis. Temporal validation was performed using a later time cohort from the same institutions. Results: A total of 216 patients were included, with an overall decannulation success rate of 63.9%. Independent predictive factors included supratentorial lesion location (OR 8.05, 95% CI 1.87-34.72), Glasgow Coma Scale (GCS) >8 at decannulation (OR 21.21, 95% CI 4.89-92.04), < 6 airway suctioning episodes/24 h (OR 13.13, 95% CI 3.31-52.09), capping trial ≥24 h (OR 6.28, 95% CI 1.62-24.43), and serum albumin (Alb) ≥35 g/L (OR 4.71, 95% CI 1.43-15.51). The scoring system (0-15 points) achieved an AUC of 0.933 (95% CI: 0.888-0.978) in the training set and 0.930 (95% CI: 0.868-0.993) in the validation set. A cut-off of ≥10 points yielded a sensitivity of 91.9% and specificity of 84.3%. Conclusion: The predictive scoring system is scientifically valid and clinically practical, providing a simple tool to assist in decannulation decision-making and timing optimization in neurosurgical patients.