Study Design
This was a single-center, retrospective cohort study conducted at a tertiary hospital in Osaka, Japan. The study followed the Declaration of Helsinki and was approved by the Institutional Review Board at Osaka General Medical Center (approval no. 30-S11-002). The board waived the need for informed consent as this was a retrospective study.
Patient Population
Records were retrospectively reviewed from patients with severe blunt trauma (Injury Severity Score [ISS] ≥ 16) treated between August 2007 and July 2015. We included patients who underwent whole-body CT scanning within 90 minutes of emergency room arrival. Exclusion criteria included patients who were transferred from other hospitals and not treated in the trauma resuscitation room, patients with traumatic cardiopulmonary arrest on arrival, pediatric patients younger than 15 years, patients who were transferred to other hospitals within 24 hours after admission, and pregnant women. Patients who underwent CT within 2 minutes after emergency room arrival and patients who received emergency surgery before CT scanning were also excluded in an effort to include patients who would receive the most benefits from CT scanning. Furthermore, in our assessment of the effects of D2BC, the dataset included subjects who received any bleeding control treatment.
Trauma Management Policy
At the halfway point of the study period (August 2011), our hospital installed an angio-CT system in a trauma resuscitation room. As this system enabled us to conduct CT examinations immediately after patient arrival without transfer to a CT room, the indication for CT scanning in patients with potentially life-threatening injury changed at this time point. However, the fundamental concepts of trauma management were based on ATLS methods throughout the entire study period.21
When patients arrived at the trauma resuscitation room, a trauma team completed the primary assessment, and an attending physician decided whether to perform CT scanning. Hemodynamically stable patients or patients who rapidly responded to the initial fluid resuscitation were assumed to be able to tolerate CT scanning in the first study period in which the CT scanner was on the same floor as, but not in, the trauma resuscitation room. In the latter period, a systolic blood pressure (BP) of 70 mmHg was regarded as the threshold for conducting CT scanning or immediate resuscitative procedures including resuscitative thoracotomy and resuscitative endovascular balloon occlusion of the aorta (REBOA). Not all blunt trauma patients underwent selective CT, but all underwent a total-body contrast-enhanced CT scan that included the head, neck, chest, abdomen, and pelvis in both study periods. If certain abnormalities were clearly identified in the primary survey or CT scanning was difficult due to hemodynamic instability, an emergency bleeding control surgery was performed in the trauma resuscitation room.
Another feature of our institution is that all trauma surgeons are trained to perform both general surgery and interventional radiology. Therefore, we can start not only emergency laparotomy and thoracotomy but also interventional radiology procedures including REBOA and angioembolization without significant time delay after patient arrival.
Data Extraction
The emergency department variables (Glasgow Coma Scale, systolic BP, heart rate [HR], shock index, body temperature [BT], respiratory rate, pH, base excess, lactate value, hemoglobin [Hb], prothrombin time-international normalized ratio [PT-INR], and activated partial thromboplastin time) were recorded as the initial set of vital signs and laboratory tests. The Abbreviated Injury Scale (AIS) of each body region was recorded, and the ISS, Revised Trauma Score (RTS), and probability of survival were calculated using the Trauma and Injury Severity Score (TRISS) method. Emergency procedures were recorded under the following groups: direct bleeding control surgery (thoracotomy, laparotomy, preperitoneal pelvic packing, and others), interventional radiology (chest, abdomen, pelvis, and others), and emergency bleeding control procedures included both direct bleeding control surgery and interventional radiology. The D2CT (time from emergency room arrival to initiation of CT) and the D2BC (time from emergency room arrival to initiation of bleeding control procedure) were recorded.
The primary outcome measure for analysis was 28-day mortality. The secondary outcome measures were 24-hour mortality and the Oxford Handicap Scale (OHS). The OHS was evaluated at 28 days or the day of hospital discharge, whichever occurred first, with handicap categories of independent (Grade 0 to 2), dependent (Grade 3 to 5), or death (Grade 6).
Statistical Analysis
Descriptive statistics were calculated as medians (interquartile range) or proportions (numbers), as appropriate. To examine the effect of D2CT on 28-day mortality, we performed a multivariable Cox proportional hazard regression model with adjustment for the following covariates described in Table S1: age, sex, mechanism of injury, ISS, RTS, presence of deadly coagulopathy, presence of acidosis, presence of hypothermia, presence of bleeding control procedure, lactate, Sequential Organ Failure Assessment (SOFA) score at day 1, HR, BT, pH, Hb, and PT-INR. We also assessed the effect of D2CT on the occurrence of 24-hour mortality using a multivariable logistic regression model with adjustment for the same covariates described above except for SOFA score at day 1, HR, BT, pH, Hb, and PT-INR to avoid overfitting. Finally, the effect on OHS was assessed by a multivariable proportional-odd logistic regression model with adjustment for the covariates considered in the Cox proportional hazard regression model. In this model, the fatal cases were categorized as Grade 6 of the OHS. For assessment of non-linear associations between D2CT and each outcome variable, we applied the restricted-cubic-spline method to D2CT in all models. All missing values were imputed by a multiple imputation method using the “aregImpute” function in the “rms” package.22 Then, we conducted similar analyses to estimate the effect of D2BC on each outcome variable in the cohort with bleeding control. In these models, the presence of a bleeding control procedure was removed from the adjustment covariates. In the logistic regression model assessing the effect on 24-hour mortality, we considered only age, sex, mechanism of injury, ISS, and RTS to avoid overfitting. Non-linearity and missing-value imputations were conducted similarly to the regression models assessing the effect of D2CT.
All statistical inference and hypothesis testing was conducted with a two-sided 5% significance level using R software version 3.6.1 (https://cran.r-project.org/).