Determinants of tracheostomy decannulation: an international survey

Study questions

We asked clinical experts around the world for their opinions regarding decannulation of tracheostomized patients. We asked four specific questions: (a) Which patient factors do clinicians rate as being important in the decision to decannulate a tracheostomized patient? (b) Which clinician and patient factors are associated with clinicians' recommendations to decannulate a tracheostomized patient in a clinical scenario? (c) To establish a definition of decannulation failure, a time frame for reinsertion of an artificial airway needs to be established. What time frame do clinicians consider for tracheostomy decannulation failure? (d) What do clinicians consider to be an acceptable rate of tracheostomy decannulation failure?

Instrument development and testing

A survey instrument based on questionnaires by Cook and colleagues [19] and Hebert and colleagues [20] was developed to examine clinical experts' opinions regarding decannulation of tracheostomized patients. We performed a computerized search of the MEDLINE databases using the Medical Subject Headings 'tracheostomy', 'tracheotomy', and 'ventilator weaning' and the text words 'tracheostomy tube', 'artificial airway', 'extubation', and 'decannulation'. We identified all relevant articles published in the English language. We also conducted semi-structured interviews with 18 attending physicians (5 ICU physicians, 5 pulmonary medicine physicians, 5 physicians who work in a mechanical ventilation weaning unit, and 3 surgeons who perform tracheostomies), 10 respiratory therapists, 2 nurse practitioners, and 2 speech therapists to generate a list of factors contributing to the decision to decannulate a tracheostomized patient. In the initial questionnaire, we included 10 determinants of tracheostomy decannulation. The determinants initially were identified from the literature review and were selected using content experts and a Delphi method. We limited our sampling to physicians and respiratory therapists.

Based on our interviews and clinical experience, we constructed two medical and two surgical clinical case scenarios considered to be representative of the types of patients commonly treated in ICUs. Each scenario included the determinants of tracheostomy decannulation and were randomly varied as follows: age (45 versus 75 years), etiology of respiratory failure (pneumonia versus chronic obstructive lung disease), difficulty of intubation (easy versus difficult), level of consciousness (alert and interactive versus drowsy but arousable), ability to tolerate capping (tracheostomy tube capped for 24 versus 72 hours), cough effectiveness (strong versus weak cough), secretions (scant thin versus moderate thick secretions), swallowing function (enteral nutrition via a gastric tube and nothing per mouth versus enteral nutrition via gastric tube and is eating Jell-O [Kraft Foods Inc., Northfield, IL, USA] and pudding), respiratory rate (18 versus 28 breaths per minute), and oxygenation (oxygenation is 95% with an FiO₂ [fraction of inspired oxygen] of 0.3 versus 0.5). The randomization procedure was designed to respect the logical constraints among the variables to ensure clinical consistency within the scenarios. Three scenarios were randomly selected to be included with each survey.

We asked respondents to provide basic demographic and professional data, including their experience in managing patients with tracheostomies and decannulations. For each of the potential determinants of decannulation, respondents were asked to rate the importance in their decision-making process by means of a 7-point Likert scale ranging from 1 (irrelevant) to 7 (very important). Some patients who are decannulated need to have an artificial airway reinserted and consequently respondents were asked what they considered as a time frame for decannulation failure as well as an acceptable rate of decannulation failure. Finally, a decannulation recommendation ('yes' or 'no') was requested for each of the three patient scenarios.

An assessment of the questionnaire's clarity and the completeness and realism of the scenarios was performed through piloting and semi-structured interviews with 10 ICU attending physicians and 10 respiratory therapists. Following these interviews, a few modifications were made and patient comorbidities were added as an additional determinant of tracheostomy decannulation to both the survey questions and scenarios (no significant prior comorbidities versus end-stage renal disease). The questionnaire and case scenarios are presented in Appendix A.

The instrument was formally tested in 11 clinicians (6 attending physicians and 5 respiratory therapists). Test-retest reliability was performed twice (2 weeks apart) and demonstrated an overall kappa score of 0.79 for the case scenarios. A clinical sensibility assessment using the methodology of Cook and colleagues [19] demonstrated the instrument to have good discriminability (8/11), clarity (11/11), utility (9/11), face validity (10/11), content validity (9/11), and minimal redundancy (1/11) [21].

Survey administration

We searched the internet to identify the email addresses of the authors of tracheostomy articles identified in our initial search of the medical literature. These clinicians (n = 69) were sent, via email, a cover letter explaining the purpose of the study and a unique username and password that provided access to a secure web-based questionnaire (Microsoft SQL Server 2005; Microsoft Corporation, Redmond, WA, USA). Participation in the study was voluntary, and written consent was not requested but was inferred by survey completion. Reminders were sent to those clinicians who did not respond to the first mailing within 8 weeks. A second reminder was sent to those clinicians who did not respond to the first reminder within 8 weeks. A snowball sampling technique was employed and each questionnaire asked respondents to provide recommendations for additional clinical experts to survey [22]. The study was approved by the Institutional Review Board of Massachusetts General Hospital (Boston, MA, USA).

Statistical analysis

The strategy for the primary analysis was to answer each of the four specific study questions. Survey responses were summarized using nominal (proportions), ordinal (median and interquartile range), and interval (mean and standard deviation) measures. Physician and respiratory therapist responses were compared using t tests, χ² tests, and Fisher exact test for outcomes with rare events. Nonparametric comparisons were performed using the Mann-Whitney and nonparametric trend tests. Logistic regression was performed to examine associations between clinician factors, patient factors, and clinicians' decannulation recommendations. Clinician factors (profession, time since graduation, primary work facility, years of tracheostomy experience, number of tracheostomy patients treated yearly, and decannulation experience) and patient factors (scenario, age, comorbidities, etiology of respiratory failure, difficulty of intubation, level of consciousness, ability to tolerate capping, cough effectiveness, secretions, swallowing function, respiratory rate, and oxygenation) were first examined using univariate analyses. Variables that were significant at a p value of 0.1 or less were included in the multivariable analyses. Variables were selected by means of backward stepwise regression and comparison of the regression sum of squares. Two interactions were tested: etiology of respiratory failure and patient age as well as cough effectiveness and secretions. Our patient scenario data were clustered within respondents. To account for the interdependence of these observations, we used robust estimates of variance (generalized estimating equation) [23]. Statistical analyses were performed using Stata version 9.0 (StataCorp LP, College Station, TX, USA) with two-tailed significance levels of 0.05.

Response rate

In the end, the survey was sent to 309 clinicians (238 physicians and 71 respiratory therapists) at 118 medical centers in 10 countries (USA, Canada, Italy, Spain, France, Germany, UK, Greece, Australia, and Japan) between May and December 2006. Of the 309 clinicians who were sent the survey, 225 (73%) responded. The response rates for physicians (173/238 [73%]) and respiratory therapists (52/71 [73%]) were similar (p = 0.927).

Respondent characteristics

Determinants of tracheostomy decannulation

Clinicians rated level of consciousness, ability to tolerate tracheostomy tube capping, cough effectiveness, and secretions as the four most important determinants in the decision to decannulate a tracheostomized patient (Figure 1). Patient comorbities, etiology of respiratory failure, swallowing function, respiratory rate, and oxygenation were judged to be of moderate importance. Patient age was the single factor that was rated as being irrevelant. Physicians rated level of consciousness (median score: 6 versus 5; p < 0.001) as significantly more important and ability to tolerate tracheostomy tube capping (median score: 6 versus 7; p < 0.001) as significantly less important than respiratory therapists did.

Responses to scenarios

Decannulation failure

The distribution of clinicians' responses for describing a decannulation failure is summarized in Figure 2. The most frequent response to the question of what clinicians considered a time frame for decannulation failure was 48 hours. The median response was 96 hours. Both the median and mode for what clinicians considered an acceptable rate of failure for tracheostomy decannulation were 2% to 5%. Compared with physicians, respiratory therapists preferred shorter time frames for defining decannulation failure (median response: 96 hours versus 48 hours; p = 0.002 for test of proportions) but identified similar acceptable rates of decannulation failure (median response: 2% to 5% versus 2% to 5%; p = 0.066 for test of proportions).