Determinants of Successful Decannulation in Difficult-to-wean Patients

The factors of success of decannulation are not well known in the literature and the decision to decannulate is mainly based on clinical judgment. The failure rate of decannulation is between 2 and 25% with a poor prognosis in case of failure. The objective of this study is to determine the factors associated with successful tracheostomy removal in patients hospitalized in a respiratory weaning unit. The secondary objectives are to evaluate in tracheostomized and hospitalized patients in weaning unit: * The prevalence of successful tracheostomy removal; * The prevalence of successful weaning from mechanical ventilation; * Factors associated with successful weaning from mechanical ventilation; * Demographic characteristics of these patients at admission; * Ventilatory characteristics of these patients at admission; * Biological characteristics of these patients at admission; This is a prospective, single-centre, interventional cohort study with an expected duration of 2 years. All patients admitted to the respiratory weaning unit in the Forcilles' hospital, with a tracheostomy and an expected duration of mechanical ventilation \> 48 hours will be consecutively included. All factors potentially associated with successful tracheostomy removal will be prospectively collected: severity factors related to the ICU stay, ventilatory factors, respiratory and extra-respiratory factors.

1. Rational Mechanical ventilation following tracheal intubation is a common life-support procedure to treat patients with acute respiratory failure. However, even if the acute event is resolved, the ICU length of stay may be prolonged by prolonged mechanical ventilation. Many factors contribute to prolonging mechanical ventilation and prevent progression in weaning from the ventilator: neuromuscular blocking agents, pulmonary infections, pre-existing respiratory disease, etc. In case of prolonged invasive ventilation, following several failures of weaning and re-intubation, these patients are therefore most often tracheostomy by percutaneous or surgical way in order to optimise the mechanical ventilation weaning strategy. The weaning units, initially created in the United States and then in France in the early 2000s, have improved the success rate of mechanical ventilation weaning in these difficult-to-wean patients. The factors associated with the success of decannulation are not well known in the literature and the decision to remove the tracheostomy is primarily based on the clinical impression. The rate of tracheostomy removal failure ranges from 2 to 25%, with a poor prognosis in case of failure. 2. Hypothesis The severity of the critical care patient, the factors related to mechanical ventilation, the respiratory and extra-respiratory status are independent factors for successful decannulation. 3. Main objective Determine the factors associated with successful tracheostomy removal in difficult-to-wean patients hospitalised in respiratory weaning unit. 4. Secondary objectives The secondary objectives are to evaluate in tracheostomy and hospitalised patients in weaning unit: * The prevalence of successful tracheostomy removal; * The prevalence of successful weaning from mechanical ventilation; * The factors associated with successful weaning from mechanical ventilation; * Demographic characteristics of these patients at admission; * Respiratory characteristics of these patients at admission; * Biological characteristics of these patients at admission; 5. Method This is a prospective, single-centre, interventional cohort study with an expected duration of 2 years. This study will be performed per the ethical standards of the Declaration of Helsinki and will be reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. All patients admitted to the respiratory weaning unit in the Forcilles' hospital, with a tracheostomy and an expected duration of mechanical ventilation \> 48 hours will be consecutively included. Data will be recorded using a local case report form by the investigator. Demographic data, comorbidities and medical and surgical history, biological findings, the ventilatory parameters and the nutritional state will be recorded at inclusion. All factors potentially associated with successful tracheostomy removal will be prospectively collected: severity factors related to the ICU stay, ventilatory factors, respiratory and extra-respiratory factors. On-site monitoring is planned every month. 6. Statistical plan: Continuous variables will be expressed as the mean (± standard deviation) and compared using the Student's t test if the null hypothesis is not rejected by the Shapiro-Wilk test. Continuous variables will be expressed as the median (interquartile range) and compared with the Mann-Whitney U or Kruskal-Wallis test if the null hypothesis is rejected by the Shapiro-Wilk test. For categorical variables, the proportions of patients in each category will be calculated. Then the groups will be compared using Chi squared test. A multivariate logistic regression model will be carried out to assess the relation between decannulation success and variables associated in univariate analyses. All statistically significant variables will be included in the model. Variable selection will be stepwise, based on Akaike Information Criterion. To check multicollinearity between independent variables, the variance inflation factor will be calculated before performing multivariate logistic regression. Multicollinearity will be regarded as present when the variance inflation factor is \> 5. Goodness of fit will be assessed by Hosmer-Lemeshow method. For all tests, a p-value ≤ 0.05 will be considered statistically significant. All statistical analysis will be made using R software (version 3.6.1, www.R-project.org). Sample size calculation: The investigators based the sample size calculation on the number of independent variables which will be included in the logistic multivariable regression model. The investigators use the criterion of one variable per 10 events for binary logistic regression analysis. The investigators plan to include the 5 most relevant ultrasound measures in the model. As prevalence of tracheostomy success in our unit is estimated at 0.5, the investigators plan a sample size of 100 patients.