Development of clinical tracheostomy score to identify cervical spinal cord injury patients requiring prolonged ventilator support.

Abstract

Cervical spinal cord injuries often necessitate ventilator support (VS). Prolonged endotracheal tube use has conveyed substantial morbidity in prospective study. Tracheostomy is recommended if VS is anticipated to be 7 days or longer, which defines prolonged ventilation (PV). Identifying these patients on arrival and before tracheostomy need is readily evident could prevent morbidity while lowering hospital costs. We aimed to create a tracheostomy score (trach score) to identify patients requiring PV and who could benefit from immediate tracheostomy. A review of patients with cervical spine fractures and cervical spinal cord injuries from 2005 to 2017 from the Pennsylvania Trauma Outcome Study database was performed. Patients were excluded for missing data, no use of VS or death in less than 7 days. Patients were selected for a training set or validation set by state identification number. We used automated forward stepwise selection to select a logistic model. Significant continuous variables were dichotomized to create a simplified screening score (trach score) and this was applied to the validation set. Needing ventilation for 7 or more days was positively associated with higher Injury Severity Scores having a complete or anterior injury, and having a motor cord injury from C1 to C4. Application of the logistic model to the validation data produced a receiver operating characteristic curve with area under the curve of 0.7712, with 95% confidence limit (CL) of 0.6943 to 0.8481. The validation receiver operating characteristic curve was statistically better than chance using a contrast test with χ with p value less than 0.01. In the validation set, a trach score of 0 correlated to 33% needing PV, a score of 1 with 67% needing PV, 2 with 85%, and 3 with 98%. Use of the trach score identified the majority of patients requiring prolonged VS in our study. An early tracheostomy protocol using predictive modeling could aid in reduction of intensive care unit length of stay and improving ventilator weaning in these patients. External verification of this predictive tool and of an early tracheostomy protocol is needed. This work is a retrospective prognostic cohort study and meets evidence Level III criteria.