Is tracheostomy a better choice than translaryngeal intubation for critically ill patients requiring mechanical ventilation for more than 14 days? A comparison of short-term outcomes

This was a retrospective cohort study using anonymized data from an ICU clinical information system (IntelliVue Clinical Information Portfolio, Philips) derived from a single center. The database included vital signs and ventilation status recorded automatically from monitors and ventilators, laboratory data retrieved automatically from our hospital’s central laboratory report system, fluid balance, and clinical and nursing procedures. Demographic data recorded routinely at ICU admission by supervising nurses included age, sex and the Acute Physiology and Chronic Health Evaluation II (APACHE II) score. Two analytical models were constructed: a multivariable time-dependent Cox model, and propensity score matching for performing tracheostomy that adjusted for potential confounding factors to evaluate the effect of tracheostomy on the outcomes of patients with PMV.

Data were obtained from patients admitted to the 75-bed adult ICU at the tertiary referral center of southern Taiwan between January 1, 2012, and December 31, 2012. Patients who were older than 16 years and received MV ≥24 h were screened, those who underwent MV for at least 14 days were included in the study. Patients with a tracheostomy present at the time of initiation of MV were excluded. For patients who had more than one ICU admission with MV ≥14 days during the study period, only the first admission was included in the analyses to avoid the potential selection bias of including patients without tracheostomy in whom the risk of death was higher.

In our clinical practice, tracheostomy is normally undertaken after an episode of failed extubation or reintubation, in the presence of unrelieved upper airway obstruction, when airway protection or regular pulmonary toilet is indicated, when PMV is needed, or for the avoidance of the complications of prolonged translaryngeal intubation. All tracheostomies are formed surgically. The decision to perform a tracheostomy is at the discretion of the attending physician, but only after informed consent has been obtained from the patient, their next of kin or their legal representative.

Study variables included in the analyses were: age; sex; comorbidities; Charlson comorbidity index (CCI) score; APACHE II score within 24 h of ICU admission; ICU type (medical or surgical ICU); presence of a do-not-resuscitate (DNR) order; time to tracheostomy (if performed); diagnosis at ICU admission; requirement for non-invasive ventilation (NIV) after extubation; duration of MV and successful weaning from MV; ICU and hospital lengths of stay; and ICU and in-hospital mortality. Laboratory data including blood urea nitrogen (BUN), serum creatinine concentration, blood gas analysis, and white blood cell (WBC) and platelet counts were recorded. In this observational study, we could not mandate that routine laboratory tests be performed at particular times. We therefore recorded variables obtained on days 1 to 5 of ICU admission (week 1, w1) and days 8 to 12 (week 2, w2) to ensure that all data were available for the analyses. The least favorable value of each variable was chosen for analysis if more than two values existed within the same time period. ICU admission diagnosis and comorbidities were identified according to the International Classification of Diseases, ninth revision and categorized as shown in the supplementary material (Additional file 1). Non-invasive ventilation was initiated after extubation as needed, if it was judged that patients were at high risk of extubation failure [13‐16]. Successful weaning was recorded when MV was not required again after discontinuation during hospitalization, regardless of outcome.

We categorized the patients into a tracheostomy group and a translaryngeal tube group. Continuous variables are expressed as the mean ± standard deviation (SD), or the median and interquartile range (IQR). Categorical variables are expressed as frequencies and proportions (%). Continuous variables were compared with Student’s t test or the Mann–Whitney U test (for skewed data); categorical variables were compared with the chi-squared test or Fisher’s exact test, as necessary. Levels of significance are expressed as p values; p <0.05 was considered statistically significant. Demographic, clinical and laboratory values were used to determine whether tracheostomy was independently associated with both in-hospital mortality and successful weaning. As the formation of a tracheostomy was considered to be a time-dependent covariate, a multivariable time-dependent Cox model was constructed. Median values were used to categorize continuous variables for Cox models. The factors with a p value <0.1 in the univariate analysis, together with age and sex regardless of the univariate p value, were entered in the Cox models. The results are presented as adjusted hazard ratios (aHR) with 95 % confidence intervals (CIs). The impact of tracheostomy on outcomes was assessed using a propensity score for a case-matched comparison to deal with significant heterogeneity between patients undergoing tracheostomy or persistent translaryngeal intubation in a retrospective cohort study (Table 1). The case-matched comparison was performed as previously described [17]. All univariable predictors of tracheostomy with p <0.25 together with age and sex were entered into the multivariable model predicting tracheostomy. Stepwise logistic regression was performed to remove covariates that had multivariable p values of >0.25. A propensity score for undergoing tracheostomy was calculated for each patient using the coefficients of the final regression equation. Then, each tracheostomy patient was matched with a single translaryngeal intubated patient who had a similar propensity score (within 0.1 on a scale from 0 to 1). When more than one matched patient was identified for a case, the closest admission date was used to select the matched patient. We also performed a sensitivity analysis using a more stringent propensity score matching within 0.05. The matched Wilcoxon test, the McNemar test, and mixed model or general estimation equations were used for case-match study comparisons. Analyses were performed using SPSS, version 20 for Windows (SPSS Inc., Chicago, IL, USA).