Development and validation of a nomogram for decannulation in patients with neurological injury: A prognostic accuracy study

BACKGROUND: Tracheostomy is a lifesaving procedure provided for patients with severe neurological injury. However, there is a lack of clarity about whether patients can be decannulated within 6 months in those receiving tracheostomy and what factors can be detected as a predictor for decannulation. OBJECTIVE: The objective of this study was to explore predictive factors of decannulation in patients with neurological injury receiving tracheostomy within 6 months and construct a novel nomogram model for clinical diagnosis and treatment. METHODS: This retrospective observational study enrolled patients with neurological injury who were admitted to the ICU of neurosurgical department in the First Affiliated Hospital of Nanjing Medical University between January 2016 and March 2021. Patients were divided into decannulation group and cannulation group according to whether tracheostomy tube removal was performed within 6 months after tracheostomy. Multivariable logistic regression analysis was performed to determine associated risk factors with a bootstrap backward selection process. The nomogram to assess the probability of decannulation at 6 months was constructed based on the regression coefficients of the associated factors and validated by bootstrap resampling. Model performance was measured by examining discrimination (Harrell's C-index), calibration (calibration plots), and utility (Kaplan–Meier curves stratified by the tertile of the predicted probability calculated and subgroup analysis stratified by age and intervention). RESULTS: A total of 40.1% (147/367) of patients decannulated within 6 months. Significant variables in multivariable logistic regression analysis were age (odds ratio [OR], 0.972; 95% confidence interval [CI], 0.954–0.990), National Institutes of Health Stroke Scale (NIHSS) score (OR, 0.936; 95% CI, 0.911–0.963), early rehabilitation (OR, 5.062; 95% CI, 2.889–8.868), shock (OR, 0.175; 95% CI, 0.058–0.533), and secondary surgery (OR, 0.210; 95% CI, 0.078–0.566). The area under receiver operating characteristic curve estimated with these variables was of 0.793 (95% CI, 0.747–0.838; P < 0.001). A nomogram prediction model was constructed to predict the probability of decannulation in tracheostomized patients with a concordance index of 0.788 after internal validation. CONCLUSION: We developed a nomogram that can predict the probability of decannulation within 6 months in tracheostomized neurological injury patients. The nomogram, including age, NIHSS scores, early rehabilitation, shock, and secondary surgery, may assist clinicians in estimating patients' prognosis.