Results:
In the study cohort of 321 patients, the most common indications for
tracheostomy were a head and neck cancer surgical procedure (247/321,
77%) and acute airway management (58/321, 18%). The majority of the
procedures were performed by otolaryngologists (297/321, 93%) via an
open surgical technique (256/321, 80%), typically with a Bjork flap
(236/321, 74%). The cohort was predominantly male (227/321, 71%) with
the mean age being 62 years old and 42% of the cohort being older than
65 years old.
Acute postoperative tracheostomy complications were not common in this
study (29/321, 9%). The more common complications were bleeding (33%),
tracheitis (15%), and dislodgement (10%). Other, less common,
complications included dysphagia, mucus plugging, revisions, and wound
dehiscence.
Initially, bivariate analysis with Chi-Squared and Fischer Exact tests
for association were used to analyze the risk of complications during
the acute post-operative day period (days 0-14) based on sex, BMI
>30, chemotherapy, and obstructive lung disease (COPD, OSA,
or asthma). Obstructive lung disease (13.4% complications, p = 0.039)
and chemotherapy (4.3%, p = 0.04) were possibly associated with higher
risk of complications while BMI >30 and sex were not (p =
0.435 and p = 0.828, respectively). Logistic regression models to
further characterize the relationship included chemotherapy, obstructive
lung disease, age, BMI, nutrition (albumin), and neck circumference (cm)
as continuous predictors. Through backward selection, the final model
contained only obstructive lung disease (COPD, OSA, asthma) as a
predictor of complication. The associated was significant <
0.1 and insignificant <0.05 (p = 0.0502), implying this
association may be moderate.
Similarly, early post-tracheostomy complications (post-operative days
0-6) were evaluated with bivariate analysis, and radiation was found to
be a possible predictor of early complication (1.1%, p = 0.029).
Chemotherapy was associated with early complication as well (p = 0.007),
though chemotherapy and radiation were highly correlated with each other
(p = 0.000) and the decision was made to include only radiation in
multivariable analysis. Logistic regression models for early
complications included radiation, age, BMI, nutrition (albumin), neck
circumference (cm) as continuous predictors. Through backward selection,
the final model contained only radiation as a predictor of early
complication with moderated significance under 0.1 level (p = 0.060).