This cohort study was conducted in the setting of the implementation of a perioperative quality improvement initiative within the Department of Anesthesiology and Perioperative Medicine at Oregon Health & Science University (OHSU) Hospital. On December 1, 2012, OHSU instituted a practice change to transition from ETTs with a barrel-shaped cuff design to the TaperGuard™ ETT for all surgical patients. We used an interrupted time-series to compare two cohorts of patients undergoing inpatient surgery with general anesthesia and the placement of an ETT, during a baseline period with the use of standard ETT and an intervention period with the use of the TaperGuard™ ETT. The baseline cohort included patients who had surgery between April 1, 2011 and November 30, 2012; the intervention cohort included patients who had surgery between December 1, 2012 and February 15, 2014. The collection and review of clinical information for this study was approved by the OHSU institutional review board, which waived the need for informed consent. The study was registered on clinicaltrials.gov as NCT02450929. This manuscript adheres to the applicable SQUIRE 2.0 guidelines.
During the intervention period, there were no active institutional changes to address postoperative pneumonia. Patients admitted to the ICU received a uniform pneumonia prevention bundle including oral care, head of bed elevation, daily sedation interruptions with spontaneous breathing trials, and appropriate stress-ulcer prophylaxis. There were no other institutional changes to operating room management during the two study periods, including default ventilator settings, aspiration prevention techniques, and oral care.
All elective and emergency surgical patients undergoing procedures in the operating room that required endotracheal intubation followed by postoperative hospitalization were included in the study. We excluded patients younger than 18 years of age. For patients undergoing multiple surgeries during a single hospitalization, only the first surgical event of the hospitalization was described.
Outcomes and Data Collection
The primary outcome was postoperative pneumonia during the hospitalization, identified based on hospital discharge ICD-9 codes for bacterial and fungal pneumonia and included the following specific codes: 481.00-486.99 (pneumonia) and 997.31 (VAP). Centricity (General Electric, Fairfield, CT) and EPIC (Verona, WI) anesthesia information management systems were queried for clinical data. An internal OHSU perioperative patient database that included a cohort of these patients was also used to collect baseline demographics, characteristics of anesthetic and surgical perioperative care, and postoperative variables.
Summary statistics (means and standard deviations for quantitative characteristics and frequencies and percentages for categorical factors) were estimated for patients’ demographic characteristics (i.e., age, race, gender, body mass index), perioperative factors (i.e., American Society of Anesthesiologists [ASA] physical status classification, procedural classifications) and potential confounding factors (i.e., tidal volume, rapid sequence intubation, use of non-depolarizing neuromuscular blockade, positive end-expiratory pressure [PEEP]). We first compared the characteristics between the baseline and intervention cohorts using two-sample unequal variance t-test for the quantitative characteristics or chi-squared test statistics for the categorical ones. We also presented ETT cohort (unadjusted) summary statistics for the primary outcomes: (a) VAP, using a chi-square test, (b) duration of mechanical ventilation (min), and (c) hospital length of stay (days). The latter two comparisons were made using an unequal variance t-test. We then formally evaluated whether there were differences in these outcomes by adjusting (controlling) for potential confounding factors. For the postoperative pneumonia outcome and hospital mortality outcome, we performed multivariable logistic regression to test whether the odds of postoperative pneumonia or mortality in the intervention cohort were different than the odds of postoperative pneumonia or mortality in the baseline cohort. For the two quantitative secondary outcomes, hospital length of stay (LOS) and duration of mechanical ventilation (min), we performed multivariable linear regression using robust (sandwich) estimated standard errors to remedy possible violation of the model variance assumption. We tested for differences in the two cohorts using standard (adjusted) pairwise comparison tests.
Finally, we explored whether possible effects between postoperative pneumonia and the two cohorts were modified by the following potential effect modifiers: diabetes, hypertension, COPD, tobacco abuse, ischemic heart disease, GERD, heart failure, obesity, and intraoperative use of nondepolarizing neuromuscular blockade. These analyses were performed using multivariable logistic regression models. Separate models were fitted for each of the potential effect modifiers, similar to the primary adjusted analysis for postoperative pneumonia, but with the inclusion of one additional term for the effect modifier and its interaction term with the cohort predictor. The a priori selected confounding factors included in all adjusted analyses were ASA status (five categories), tidal Volume (ml/kg), Caucasian race (yes/no), age (in years), male gender (yes/no), and PEEP (=0 or >0). All hypothesis tests, associated p-values and confidence intervals were two-sided. The statistical analyses were performed using Stata (ver. 15.1) and R (ver. 3.3.3) statistical packages.