Study design and data source
In this study, we investigated the use of the two devices for one-lung ventilation in clinical anesthesia using data from a population-based nationwide insurance dataset in Taiwan (Fig. 1) and used the dataset to extract records that satisfied a matched-pairs study design controlling for patient age, sex, and year of surgery. We used this dataset, with fewer confounding factors, to estimate the relative risk of adverse postoperative outcomes including pulmonary infection, respiratory failure, and mortality associated with use of double-lumen tubes and bronchial blockers for one-lung ventilation in multivariate analysis.
The data used in this study were collected from the inpatient claims database from 2000–2012 that was released by the National Health Insurance Research Dataset (NHIRD) in Taiwan. The Taiwan National Health Insurance Program was launched in 1995, and by 2011 it covered more than 99% of the population [23, 24]. The high coverage rate makes the NHIRD one of the largest population-based databases in the world. The accuracy of the adverse event records held in this database, including those for ischemic stroke, acute myocardial infarction, and in-hospital mortality, has been validated elsewhere [25-28]. The database provides the age, sex, admission details, and encrypted identification number for each patient and the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes for diagnoses and procedures. All secondary data were de-identified. The protocol for this study was granted exemption from full review by the Institutional Review Board (IRB) of Chi Mei Medical Center, Tainan, Taiwan (Chairperson Prof Chung-Hsi Hsing) on May 25, 2016 (IRB-10505-E05) in view of the anonymity of the data. The need for informed consent was waived for the same reason.
Patients
This retrospective, nationwide, population-based study included patients who underwent a surgical procedure that included lung separation by placement of a bronchial blocker or double-lumen tube between 2000 and 2012 (Fig. 1). The codes for the lung isolation devices used were identified (Sup. 1). To eliminate any influence of previous surgical events, only first inpatient surgical procedures that included lung isolation were collected. Cases with missing data or records that indicated multiple isolation devices that were used for the same surgical procedure (i.e., charges for both a double-lumen tube and a bronchial blocker) were excluded. Each patient who underwent one-lung ventilation surgery using bronchial blockers was matched to three randomly selected patients who underwent surgery using double-lumen tubes for age, sex, and the year that the surgery was performed. Patient age was classified as younger than 50 years, 51–64 years, or older than 65 years. The lung isolation technique was classified as being performed during pneumonectomy, lobectomy, segmentectomy, local excision, other pulmonary surgery, oesophageal surgery, cardiac surgery, vascular surgery, or other type of surgery. Pre-existing comorbidities that were determined from medical claims within the 3 years before surgery included heart disease, hypertension, diabetes mellitus, renal disease, obstructive pulmonary disease (chronic obstructive pulmonary disease or asthma), and dementia (Sup. 2). Patients encountered problems in airway management with associated upper airway injury recorded after index surgery were considered difficult airway (Sup. 3). The modified Charlson Comorbidity Index (CCI) [29] was used as a surrogate for the American Society of Anesthesiologists physical status classification to indicate individual health status [30, 31] and is presented as the mean and standard deviation. The hospital accreditation level was also considered. According to the Taiwan Joint Commission on Hospital Accreditation, local Taiwanese hospitals with fewer than 300 beds provide primary medical care, regional hospitals with 301–999 beds provide secondary care, and medical centres with 1000–2500 beds provide research facilities and are responsible for training the staff and tertiary medical care.
Outcomes
The primary outcomes were readmission within the first postoperative year with a diagnosis of pulmonary infection or respiratory failure (Sup. 4). The secondary outcomes were in-hospital mortality and one-year mortality after the index surgery. In-hospital mortality was identified using the “in-hospital death” code at discharge or withdrawal from the National Health Insurance Program within 30 days after discharge because of the custom of choosing the “die at home” option in Taiwan. Enrolment in the National Health Insurance Program is mandatory and is withdrawn within 30 days of death. After index surgery, both primary and secondary outcomes were traced monthly for six months. The duration of ventilator support, length of stay in hospital, and duration of stay in the intensive care unit were also compared between the two study groups.
Statistical analysis
Descriptive statistics and standardized differences were used to compare the relative balance of patient information in the double-lumen tube group and the bronchial blocker group. Patient characteristics that were considered potential confounders associated with postoperative outcomes were included as covariates in the predicted model in order to evaluate the relative risk associated with use of double-lumen tubes or bronchial blockers. The use of standardized differences [32, 33] indicated that the balance between the groups with regard to the type of surgery performed, hospital accreditation level, and underlying health status indicated by the CCI score and other comorbidities was acceptable. Conditional logistic regression analyses adjusted for the surgical procedure performed, hospital accreditation level, CCI score, underlying comorbidities, and difficult airway were performed to compare the risk of adverse postoperative outcomes between the two groups using odds ratios (ORs) and 95% confidence intervals (CIs). Correlations were considered statistically significant at P < 0.05. The detectable OR of 1.3 between cases and controls was estimated at 90% statistical power and the probability of a type I error at 0.05. Statistical analyses were performed using SAS version 9.4 software (SAS Institute Inc., Cary, NC, US.