This is a registry-based unmatched cohort study. We identified trauma patients requiring intubation during resuscitation in the ED, then extracted relevant trauma related parameters and outcome data from the trauma registry of the hospital and patients’ health records. The reporting of this study is in compliance of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement.  Ethical approval was sought from Research Ethics Committee, Kowloon Central Cluster, Hospital Authority of Hong Kong (KC/KE-18-0232/ER-3).
This study was conducted in a trauma centre in Hong Kong. The centre is one of the five trauma centres in the region and it is also a regional tertiary and quaternary referral centre with full complements of clinical services. Daily ED attendance of the centre is around 500 patients. As a designated trauma centre, our unit receives primary trauma diversions from ambulance service and also secondary trauma diversion from other regional hospitals. Yearly major trauma (injury severity score (ISS) >15) attendance is around 300.
In this hospital, TRT composes of specialist EP and trauma surgeons. Anaesthesiologists and other subspecialties will be consulted according to the injury pattern. The decision to solicit help from anaesthesiologists for airway management is at the discretion of the attending EP specialist.
The hospital trauma registry captures any patients who met TRT activation criteria (see appendix), who were triaged as critical or emergency in the ED(triage guideline), who died (excluding death prior to arrival to ED) and were admitted to intensive care units (ICU)/high dependency units (HDU). Injury and outcome data were prospectively collected and entered into the registry.
We screened all patients requiring endotracheal intubation in the ED from the trauma registry of the centre during the period of 1st Jan 2015 to 31 Dec 2018. Patients were included if age 18 or above. Pregnant patients were excluded, as well as those presented with cardiac arrest or transferred from other hospitals.
Subjects recruited were sorted into two cohorts according to the specialty (EP or AN) providing trauma airway management.
Our primary outcome is 30-day all-cause mortality.
Secondary outcomes include 1) proportion of successful intubations, 2) need of surgical airway or 3) rescue manoeuvres 4) respiratory or airway complications, 5) ICU and HDU mortality, 6) post-intubation cardiac arrest, 7) post-intubation hypotension and 8) post-intubation hypoxia.
Rescue manoeuvres was defined as any use of laryngeal mask airway or bronchoscopic intubation instead of direct laryngoscopy or video laryngoscopy. Airway and respiratory complications is a composite outcome defined as any ventilator-associated pneumonia (VAP)/ hospital-acquired pneumonia (HAP) or tracheostomy for prolonged intubation. HDU/ICU mortality is defined as mortality during ICU/HDU care. Post-intubation cardiac arrest was defined as any cardiac arrest during resuscitation in the emergency department after injection of induction and/or paralytic agents whichever was earlier. Post-intubation hypotension was defined as the first systolic blood pressure(SBP) after placement of endotracheal tube less than 90mmHg before leaving ED. Post-intubation hypoxia was defined as the lowest pulse oxygen saturation (SpO2) after placement of endotracheal tube less than 93% before leaving ED.
Patients’ demographic data including age, gender and comorbidities were also collected. Comorbidites included AIDS, cirrhosis, diabetic mellitus, hepatic failure, immunosuppression, leukemia/myeloma, Non-Hodgkin’s lymphoma, solid tumor with metastasis, chronic respiratory condition. The definition of these conditions follows the definition used in Apache IV score(12). Renal failure was defined as patient requiring renal replacement therapy (peritoneal dialysis or hemodialysis). Data on patients requiring long term residential care (Elderly home, etc) was also collected. Clinical variables collected included vital signs (blood pressure, oxygen saturation and Glasgow coma scale), duration of ED resuscitation, time to intubation, base excess, emergency operations (including three-in-one procedure for pelvic fracture, laparotomy, neurosurgery, closed or open reduction with fixation, or others), activation of massive transfusion protocol, presence of neck collar, trauma diversion status and TRT activation status. Duration of ED resuscitation is the duration between patient transferred in and out of resuscitation room. Time to intubation is the time from ED registration to injection of paralytic agent. If no paralytic agent was used, time of recording first vital signs documented after intubation was used.
Mechanism of trauma was grouped into three categories, namely burns, low energy trauma (such as fell from height less than 2 metres, or penetrating injuries), and high energy trauma (such as fell from height more than 2 metres or motor vehicles accidents). Trauma related scores (Probability of survival , Injury Severity Score (ISS), Revised Trauma Score (RTS) were also extracted for analysis.
We were particularly interested in chest trauma, head, neck and facial trauma which may predict intubation difficulty. As such, we have extracted the maximum abbreviated injury scale (MAIS) from the registry, by analysing the body region code of 1, 2, 3, 4 and 6 (head, face, neck, chest and spine) of the Abbreviated Injury Scale (AIS) separately. Spinal injuries other than cervical spine were excluded.
Data Sources and Measurements
Clinical and laboratory data were retrieved from the trauma registry and the Clinical Management System (CMS) via the clinical data analysis and reporting system (CDARS) of the Hospital Authority. Patient’s case notes were also accessed to retrieve data and ascertain diagnosis. The trauma registry was also the source of trauma related data, such as clinical parameters during ED resuscitation, mechanism of injuries and injury severity scores, namely, AIS, ISS and RTS. The outcomes (mortality/survival) and interventions given during pre-hospital, ED resuscitation and inpatient period were also retrieved.
Diagnosis of ventilator associated pneumonia and hospital acquired pneumonia were based on the diagnosis made by the treating physicians.
Coding of the ISS and AIS were performed by an experienced trauma nurse specialist trained in the AIS coding.
In addition to the inherent bias in a retrospective study, we noted that EPs were found to intubate more patients with isolated head injury. It is well known that probability of survival of severely head injured patients is poorer when compared with that of patients with other system injuries having the same ISS. There are also many parameters showing baseline imbalance. Hence a planned logistic regression was performed to control for confounding factors resulted from differences in trauma severity and pattern.
Based on a cursory review of trauma registry data of the year 2017. There were a total of 107 intubated patients. Mortality among these patient were 53.5% and 29.7% when the airway were managed by emergency physician (EP) versus Anaesthesiologists (AN) respectively.
Based on the above observation, the required sample size were found to be around 195, according to the Fleiss equation with continuity correction, taking power (1-beta) of 90% and alpha of p<0.05.  To achieve the expected the sample size, we have screened the trauma registry data from 2015 to 2018.
In this study, the following grouping of quantitative variable were performed prior to analysis.
MAIS more than 2 in each of the body regions were considered as serious. This is based on the definition of the abbreviated injury score.
Patient with ISS > 15 is considered as suffering from major trauma, as customary in trauma researches.
IBM SPSS Statistics Version 25, Microsoft Excel version 16.23 and Apple Inc. Numbers were used to handle calculation and data analysis.
Patient demographics were reported with descriptive statistics, using mean, median and standard deviations (SD) and interquartile range (IQR). Normality of continuous variables were tested with Shapiro-Wilk Test. Difference between normal variables were analyzed with Student's t tests while non-normal variables were managed with Mann Whitney’s sign rank tests. Categorical outcome variables such as 30-day all-cause mortality were analyzed with the Chi-square tests or Fisher-Freeman-Halton-exact tests.
Missing data would be managed with multiple imputations, using missing at random assumption. The multiple imputation function of SPSS were used to impute10 sets of complete dataset by the Markov Chain Monte Carlo method. The maximum number of parameters in imputation model was set to 10. Each dataset were then inspected by two authors for consistence. Then the pooled averaged values from the 10 sets will be used for analysis.
We carried out a multivariable logistic regression to control for any potential confounding and interaction. Baseline characteristics with potential difference (P<0.25) were included in the univariate regression analysis (Table 3). Statistical outcomes are to be regarded as significant if P is <0.05, adjusted odd ratios would be reported.
Analyzing the missing data by listwise deletion were planned as a sensitivity analysis.