Development and validation of a risk prediction model for tracheostomy in acute traumatic cervical spinal cord injury patients.

PURPOSE: Tracheostomy may become indispensable for patients with acute traumatic cervical spinal cord injuries. However, the early prediction of a tracheostomy is often difficult. Previous prediction models using the pulmonary function test (PFT) have limitations because some severely injured patients could not provide acceptable PFT results. We aim to develop an alternative model for predicting tracheostomy using accessible data obtained from the bedside.

METHOD: Clinical, neurological and radiological data from 345 consecutive patients with acute tetraplegia were retrospectively reviewed. We applied multiple logistic regression analysis (MLRA) and classification and regression tree (CART) analysis to develop the prediction model for tracheostomy. By train-test cross-validation, we used the sensitivity, specificity, area under the receiver operating characteristics curve (AUC) and correction rate to evaluate the performance of these models.

RESULTS: According to the American Spinal Injury Association (ASIA) standards, an admission ASIA motor score (AAMS) ≤ 22, ASIA grade A and presence of respiratory complications were identified as independent predictors of tracheostomy by both models. The model derived by CART suggested that the highest signal change (HSC) in the spinal cord on magnetic resonance imaging (MRI) also affected a patient's requirement for a tracheostomy, while MLRA demonstrated that tracheostomy was also influenced by the presence of an ASIA grade B injury. The CART model had a sensitivity of 73.7%, specificity of 89.7%, AUC of 0.909 and overall correction rate of 87.3%. The sensitivity, specificity, AUC and correction rate of the MLRA model were 81.8, 86.4, 0.889 and 85.7%, respectively.

CONCLUSIONS: We suggest using the CART model in clinical applications. Patients with AAMS ≤ 1 exhibit an increased likelihood of requiring a tracheostomy. For patients with an AAMS in the range of 2-22, surgeons should consider giving these patients a tracheostomy once respiratory complications occur. Surgeons should be cautious to give a tracheostomy to patients with an AAMS ≥ 23, if the patient experiences an incomplete spinal cord injury and the HSC in the spinal cord is at C3 level or lower based on MRI. For other patients, close observation is necessary; generally, patients with complete SCI might require a tracheostomy more frequently.