Over a period of 14 months (November 2008–December 2009), we evaluated all the patients admitted to two ICUs: a 26-bed, closed clinical–surgical ICU belonging to a tertiary hospital with 700 beds, and an 8-bed general ICU without neurocritical patients belonging to a secondary hospital with 300 beds. The Ethics Committees of both centers approved the study, though informed consent was not requested, since the study was of a purely observational nature.

The study included patients tracheostomized in the context of prolonged mechanical ventilation (MV) (>21 days) or prolonged weaning19 (group 1), and patients tracheostomized due to neurological defects or the incapacity to manage respiratory secretions (group 2). Neurological defects were defined as a motor component on the Glasgow Coma Scale (GCS) of <6 points on the day on which tracheostomy was performed. The patients were considered unable to manage their respiratory secretions when two or more failed extubation procedures were registered because of the retention of secretions, or when the patient was not extubated after tolerating a weaning test because the respiratory secretions had to be aspirated ≥3 times in an hour or ≥8 times in 8h. Patients presenting criteria for inclusion in both groups were assigned to group 1 to the effects of analysis.

The exclusion criteria were a patient age of <18 years, tracheostomy performed prior to admission to the ICU, a motor component of the GCS of <6 at the time of attempted decannulation, tracheostomy indicated for long-term protection of the airway, neuromuscular disease (e.g., lateral amyotrophic sclerosis, Guillain–Barré syndrome), patients with non-resuscitation instructions, or the transfer to chronic care units of patients with a need for partial ventilation support. Patients with early tracheostomy (MV<7 days) were also excluded, because such procedures were only performed in patients with serious brain damage at the time of admission to the ICU.

Daily patient assessment was carried out, using the following criteria to time weaning19: patients in the recovery phase of the disease giving rise to the need for MV; respiratory criteria (PaO2/fraction of inspired oxygen (FiO2) >150mmHg with positive end-expiration pressure (PEEP)<8cm H2O and arterial pH>7.32); and clinical criteria (absence of ECG evidence of myocardial ischemia, no need for vasoactive drugs or dopamine (≤5μg/kg/min), heart rate<140beats/min, hemoglobin>8g/dl, temperature<38°C, no need for sedation, presence of respiratory stimulus, presence of adequate spontaneous cough reflex, and GCS>8 points–exclusively comprising the ocular and motor components).

The tracheostomized patients were progressively weaned from MV according to a clinical algorithm comprising progressive supportive pressure reduction or T-tube disconnections, according to the decision of the supervising clinician.20 When the patients tolerated at least 12 consecutive hours of disconnection during two successive days they were switched to a continuous T-tube. Weaning failure in turn was defined as the reintroduction of ventilatory support in the 72h following 24h without MV.

After tolerating 24 consecutive hours of disconnection, the patients were evaluated for possible decannulation. Firstly, an occlusion test21 was performed to discard possible airway obstruction. Briefly, the cannula was replaced by a fenestrated cannula, withdrawing the internal cannula. The tip of the cannula was then occluded for 5min. If the patient showed changes in heart rate, respiratory frequency or arterial pressure suggestive of tracheal stenosis, bronchoscopy was performed.

In the second step of the protocol, evaluation was made of the capacity of the patients to avoid aspirations (see below). Those patients with a normal swallowing test received an oral diet (involving a cannula without cuff inflation) in appropriate cases; in the rest of the cases, an enteral diet was provided through a nasogastric or jejunal tube until definitive disconnection from MV.

The third step in the decannulation protocol included a clinical evaluation by the supervising physician to assess the capacity of the patient to adequately manage the respiratory secretions. The assessment was mainly based on the frequency of the need for suctioning and the characteristics of the respiratory secretions, as explained further below.

The patients were decannulated when they met the following criteria: (1) negative occlusion test, discarding airway obstruction at tracheal level; (2) adequate capacity to manage the respiratory secretions, defined as a need for suctioning ≤2 times every 8h; and (3) low aspiration risk (normal swallowing test).

When the patients failed to meet these criteria, the tracheal cannula was switched to a fenestrated cannula with an internal diameter of ≤7mm and an internal tube. In the case of patients with a low aspiration risk (normal swallowing test), the replaced cannula cuff was, moreover, not inflated.

The tracheal cannula cuff was inflated when suctioning proved necessary with a frequency of <4 times in 8h, and MV disconnection had been maintained during >48h. If plugging was tolerated during 24 consecutive hours, and not removed to suction the respiratory secretions, decannulation was carried out.

Based on previous studies,1,5 the tracheostomized patients remained in the ICU until decannulation or until decannulation was not considered possible (>120 days after definitive disconnection from MV). The patients classified as presenting “expectable in-hospital death” at discharge from the ICU (Sabadell score 3) were excluded from the study and discharged to selected hospital wards after only 10 days of decannulation attempts.

Decannulation failure was defined as the need to intubate or re-cannulate the patients within the 96h following decannulation.

The following information was collected on a prospective basis: patient age, gender, APACHE II score upon admission to the ICU, diagnosis leading to admission, comorbidities, duration of MV, and duration of stay in the ICU and in hospital. In addition, variables related to tracheostomy were recorded: indication, performance time, technique, complications of the procedure and related adverse events (mainly malpositioning and occlusion episodes) during stay both in the ICU and in hospital. In turn, the following variables were documented upon weaning from MV: GCS, forced vital capacity (FVC) and spontaneous peak expiratory flow (PEF), the frequency of aspirations, and volume of the secretions (suctioned and total, obtained by summing spontaneous expectoration) during the previous 8h.

Vital capacity (VC) was measured with a Wright/Haloscale Respirometer® (Ferraris Respiratory Europe, Hertford SG13 7NW, England), and PEF was determined with a Mini-Wright® Peak Flow Meter (Clement Clarke International Ltd., Essex CM20 2TT, England). Both variables were measured with the tracheostomy cuff both inflated and deflated, in order to determine whether plugging affected the capacity of the model used. The clinical variables were recorded in the 8h after complying with the weaning criteria.

Aspiration risk was assessed by means of the swallowing test. Basically, the latter involves the swallowing of 50ml of water with the tracheostomy cuff deflated.22,23 The results of the swallowing test were classified as follows: (1) normal (≤5 swallows in <10s); (2) abnormal (>5 swallows in ≥10s, or clinical evidence of aspiration during the test); or (3) severe dysfunction (spontaneous aspiration of saliva or pharyngeal secretions, contraindicating performance of the test). The measurements were made on a daily basis, until the test results proved normal.

The respiratory secretions were clinically classified as watery, foamy, thick or almost solid.24 Needs for airway care were identified, determining the volume and viscosity of both the expectorated and the suctioned secretions. Viscosity was assessed by measuring the time required to travel the surface of the Muco-Safe® suctioning container (Unomedical A/S, Birkeroed, Denmark). In order to optimize simulation of the clinical scenario, the supervising physician was blinded to the results of these quantitative variables.

Continuous variables were reported as the mean and standard deviation (SD) in the case of variables with a normal distribution, and as the median and interquartile range (IQR) in the case of a non-normal distribution.

The Student-test was used to compare continuous variables, while the Kruskal–Wallis or Mann–Whitney U-test was applied in the case of samples with <30 individuals. The chi-squared test with Yates correction or the two-tailed Fisher exact test was used to contrast categorical variables.

Kaplan–Meier survival analysis was used to compare time to decannulation in the two groups.

Our working hypothesis was based on previous clinical observations of differences between the two groups in the time from tracheostomy to weaning, and in the time from weaning to decannulation. Accordingly, we constructed a multivariate model for each group, using Cox regression analysis to identify the variables associated to the time interval from weaning to decannulation. The results are expressed as hazard ratios (HR, with the corresponding 95% confidence interval (95%CI)), where HR>1 corresponds to a factor that reduces the time to decannulation.

Our sample size (fewer than 100 patients in each group) could have caused over-adjustments in the statistical model. This phenomenon is even more likely on considering that we recorded no differences before selecting the subgroups. On the other hand, our decannulation protocol is not based on scientific evidence, allowing variations in the time to decannulation, on establishing comparisons with other centers. In a recent epidemiological study, the optimum decannulation failure rate was estimated to be between 2% and 5%.9 Although this percentage is not based on evidence, our own percentage of 0% suggests that the protocol we use is very conservative, and could effectively delay the decannulation of patients who are ready for decannulation. Furthermore, the low weaning failure rate suggests that our weaning protocol is also conservative. As a result, the time from tracheostomy to weaning could be longer in our study than in other publications.18 Lastly, the definition of patient incapacity to manage the respiratory secretions always contains a subjective element, and we chose a very restrictive criterion in order to clearly separate the two patient groups.

Only the patients with a low level of consciousness at the time of tracheostomy and who posteriorly regained consciousness were included in the study, since cooperation was necessary for measuring most of the variables.

In conclusion, the classification of tracheostomized patients according to the indication of tracheostomy is a fundamental step in developing decannulation predictive models.