Radiomics score predicts acute respiratory distress syndrome based on the initial CT scan after trauma

Objectives Acute respiratory distress syndrome (ARDS) constitutes a major factor determining the clinical outcome in polytraumatized patients. Early prediction of ARDS is crucial for timely supportive therapy to reduce morbidity and mortality. The objective of this study was to develop and test a machine learning–based method for the early prediction of ARDS derived from the first computed tomography scan of polytraumatized patients after admission to the hospital. Materials and methods One hundred twenty-three patients (86 male and 37 female, age 41.2 ± 16.4) with an injury severity score (ISS) of 16 or higher (31.9 ± 10.9) were prospectively included and received a CT scan within 1 h after the accident. The lungs, including air pockets and pleural effusions, were automatically segmented using a deep learning–based algorithm. Subsequently, we extracted radiomics features from within the lung and trained an ensemble of gradient boosted trees (GBT) to predict future ARDS. Results Cross-validated ARDS prediction resulted in an area under the curve (AUC) of 0.79 for the radiomics score compared to 0.66 for ISS, and 0.68 for the abbreviated injury score of the thorax (AIS-thorax). Prediction using the radiomics score yielded an f1-score of 0.70 compared to 0.53 for ISS and 0.57 for AIS-thorax. The radiomics score achieved a sensitivity and specificity of 0.80 and 0.76. Conclusions This study proposes a radiomics-based algorithm for the prediction of ARDS in polytraumatized patients at the time of admission to hospital with an accuracy that competes and surpasses conventional scores despite the heterogeneous, and therefore more realistic, scanning protocols. Key Points • Early prediction of acute respiratory distress syndrome in polytraumatized patients is possible, even when using heterogenous data. • Radiomics-based prediction resulted in an area under the curve of 0.79 compared to 0.66 for the injury severity score, and 0.68 for the abbreviated injury score of the thorax. • Highlighting the most relevant lung regions for prediction facilitates the understanding of machine learning–based prediction.

Prehospital Extrication Techniques: Neurological Outcomes Associated with the Rapid Extrication Method and the Kendrick Extrication Device

Most emergency medical service personnel rely on one of two techniques to extricate motor vehicle crash victims; the Rapid Extrication Maneuver (REM) or the Kendrick Extrication Device (KED). The purpose of this study was to compare pre- and postextrication neurological outcomes between these two techniques. A retrospective review was conducted of all adult patients with a vertebral column injury resulting from motor vehicle collision and admitted to a Level I trauma center between January 1, 2003 and December 31, 2010. Standardized pre- and postextrication neurological examinations were reviewed for all patients. More than half of patients (N = 81) were extricated using the KED (53.1%, n = 43) and 46.9 per cent (n = 38) were extricated with the REM. Except for the thoracic Abbreviated Injury Score, no differences between groups emerged related to the Glasgow Coma Scale score, Injury Severity Score or Abbreviated Injury Score. There were no pre- and postextrication changes for motor to all extremities and sensation to all extremities using either method. The results of this study suggest that the REM and the KED are equivalent in protecting the patient from neurologic injury after motor vehicle collision.

Is Routine Continuous EEG for Traumatic Brain Injury Beneficial?

Severe traumatic brain injury (TBI) is associated with increased risk for early clinical and sub-clinical seizures. The use of continuous electroencephalography (cEEG) monitoring after TBI allows for identification and treatment of seizures that may otherwise occur undetected. Benefits of “routine” cEEG after TBI remain controversial. We examined the rate of subclinical seizures identified by cEEG in TBI patients admitted to a Level I trauma center. We analyzed a cohort of trauma patients with moderate to severe TBI (head Abbreviated Injury Score ≥3) who received cEEG within seven days of admission between October 2011 and May 2015. Demographics, clinical data, injury severity, and costs were recorded. Clinical characteristics were compared between those with and without seizures as identified by cEEG. A total of 106 TBI patients with moderate to severe TBI received a cEEG during the study period. Most were male (74%) with a mean age of 55 years. Subclinical seizures were identified by cEEG in only 3.8 per cent of patients. Ninety-three per cent were on antiseizure prophylaxis at the time of cEEG. Patients who had subclinical seizures were significantly older than their counterparts (80 vs 54 years, P = 0.03) with a higher mean head Abbreviated Injury Score (5.0 vs 4.0, P = 0.01). Mortality and intensive care unit stay were similar in both groups. Of all TBI patients who were monitored with cEEG, seizures were identified in only 3.8 per cent. Seizures were more likely to occur in older patients with severe head injury. Given the high cost of routine cEEG and the low incidence of subclinical seizures, we recommend cEEG monitoring only when clinically indicated.

To TQIP or Not to TQIP? that is the Question

The Trauma Quality Improvement Program (TQIP) reports a feasible mortality prediction model. We hypothesize that our institutional characteristics differ from TQIP aggregate data, questioning its applicability. We conducted a 2-year (2008 to 2009) retrospective analysis of all trauma activations at a Level 1 trauma center. Data were analyzed using TQIP methodology (three groups: blunt single system, blunt multisystem, and penetrating) to develop a mortality prediction model using multiple logistic regression. These data were compared with TQIP data. Four hundred fifty-seven patients met TQIP inclusion criteria. Penetrating and blunt trauma differed significantly at our institution versus TQIP aggregates (61.9 vs 7.8%; 38.0 vs 92.2%, P < 0.01). There were more firearm mechanisms of injury and less falls compared with TQIP aggregates (28.9 vs 4.2%; 8.5 vs 34.8%, P < 0.01). All other mechanisms were not significantly different. Variables significant in the TQIP model but not found to be predictors of mortality included Glasgow Coma Score motor 2 to 5, systolic blood pressure greater than 90 mmHg, age, initial pulse rate in the emergency department, mechanism of injury, head Abbreviated Injury Score, and abdominal Abbreviated Injury Score. External benchmarking of trauma center performance using mortality prediction models is important in quality improvement for trauma patient care. From our results, TQIP methodology from the pilot study may not be applicable to all institutions.

Pediatric Trauma Experience in a Combat Support Hospital in Eastern Afghanistan over 10 Months, 2010 to 2011

We reviewed the pediatric trauma experience of one Combat Support Hospital (CSH) in Afghanistan to focus on injuries, surgery, and outcomes in a war zone. We conducted a review of all pediatric patients over 10 months in an eastern Afghanistan CSH. We studied 41 children (1 to 18 years; mean, 8.5 years; median, 9 years), 28 (68.2%) with penetrating injuries. Blasts (13 patients) and burns (nine) were the most common mechanisms. At arrival 19 (46.3%) underwent endotracheal intubation, four (9.8%) had no palpable blood pressure, 10.6 per cent (four of 38) a Glasgow Coma Score of 5 or less, 30.6 per cent (11 of 36) base deficits of 6 or less, and 41.7 per cent (15 of 36) hematocrit 30 or less. Red cells were given in 14 (34.1%) and plasma in 11 (26.8%). Of 32 total nonburn patients, 12 (37.5%) had multiple system injuries. Three-fourths of injuries were severe (75.8% [47 of 62] Abbreviated Injury Score 3 or greater). Thirty-two patients (78.0%) required major operations: burn and wound care, orthopedic, chest, abdominal, vascular, and neurosurgical. Second operations were performed in 16 (39.0%), most often burn and orthopedic procedures. Six died (14.6%), 13 were transferred to other hospitals (31.7%), and 20 were discharged to home (48.8%; two not noted). Broad experience in operative trauma care, pediatric resuscitation, and critical care is a priority for military surgeons.

Intracranial Pressure Monitoring in Severe Isolated Pediatric Blunt Head Trauma

Very little research regarding standard treatments for pediatric traumatic brain injury (PTBI) exists. The objective of this study was to examine the use of intracranial pressure (ICP) monitoring devices in PTBI and to determine if its use was associated with any outcome benefit. Data were collected from the Trauma Registry over an 11-year period (1996–2006) on all blunt trauma pediatric patients (age < 14 years) with an initial Glasgow Coma Scale score ≤ 8. Data collected included: demographics, admission Glasgow Coma Scale score, mechanism of injury, Injury Severity Score, Abbreviated Injury Score, and use of an ICP monitor. Outcome measures included: mortality, complications, discharge location, and capacity. Thirty-three (25%) of 129 blunt PTBI patients had ICP monitors placed. Patients with monitors were more severely injured overall (Injury Severity Score: 25 vs 18, P = 0.001) and had more severe head injury (81% head Abbreviated Injury Score > 3 vs 55%, P = 0.01) than patients without monitors. However, there was no difference in mortality (28% vs 35%, P = 0.52), discharge location (P = 0.10), and discharge capacity (P = 0.84). After multivariable analysis to adjust for the differences between the two study groups, the use of ICP monitor provided no survival benefit (adjusted odds ratio: 1.1; 95% confidence interval [CI]: 0.3–4.1; adjusted P value = 0.85). The use of ICP monitor was, however, independently associated with a higher risk of developing extracranial complications (adjusted odds ratio: 4.3; 95% CI: 1.2–16.4; adjusted P value = 0.025). In conclusion, the use of ICP monitors in pediatric patients with severe isolated head injury provided no survival benefit and was associated with an increased risk of complications.

How We Die: The Impact of Nonneurologic Organ Dysfunction after Severe Traumatic Brain Injury

Although nonneurologic organ dysfunction (NNOD) has been shown to significantly affect mortality in subarachnoid hemorrhage, the contribution of NNOD to mortality after severe traumatic brain injury (TBI) has yet to be defined. We hypothesized that NNOD has a significant impact on mortality after severe TBI. The trauma registry was queried for all patients admitted between January 2004 and December 2004 who died during their initial hospitalization after severe TBI (head Abbreviated Injury Score 3 or greater). Cause of death and contributing factors to mortality were determined by an attending trauma surgeon from the medical record. The data were analyzed using both Fisher's exact and Wilcoxon rank sum. One hundred thirty-five patients met inclusion criteria. Sixty-seven per cent were males, 83 per cent were white, and the mean age was 38.5 years. Mean length of stay was 2.9 days. Fifty-four patients (40%) had isolated TBI (chest Abbreviated Injury Score = 0, abdominal Abbreviated Injury Score = 0). Of the 81 deaths attributed to a single cause, 48 (60%) patients died from nonsurvivable TBI or brain death, whereas 33 (40%) died of a nonneurologic cause. Cardiovascular and respiratory dysfunction (excluding pneumonia) contributed to mortality in 51.1 per cent and 34.1 per cent of patients, respectively. NNOD contributes to approximately two-thirds of all deaths after severe TBI. These complications occur early and are seen even among those with isolated head injuries. These findings demonstrate the impact of the extracranial manifestations of severe TBI on overall mortality and highlight potential areas for future intervention and research.

Pelvic Fracture Risk Assessment on Admission

Patients with pelvic fractures (PF) have high incidences of associated injuries and mortality. To identify patients with PF at the highest risk for mortality on admission to a Level I trauma center, investigators analyzed 566 PF in 12,128 patients (1996 to 2005). PF were categorized on arrival as high risk (HR) or low risk (LR) by initial blood pressure, examination, radiographs, and CT. HR PF included open fractures, open-book injuries, PF 4 or greater abbreviated injury score, and hypotension (89 mmHg or less systolic blood pressure); all other cases were categorized as LR PF. Patients with PF had 6 per cent (35 of 566) mortality compared with 3 per cent (300 of 11,529) without PF. Mortality was 24 per cent (25 of 103) in HR PF, including 11 per cent (one of nine) of open fractures, 25 per cent (12 of 49) of open-book injuries, 23 per cent (14 of 62) of 4 or greater abbreviated injury score, and 33 per cent (20 of 60) of hypotensive patients compared with 3 per cent (13 of 454) of LR PF. Compared with LR PF, patients with HR PF were younger (43.5 vs 53.8 years) with higher injury severity scores (28.7 vs 11.9) and longer hospital stays (10.6 vs 7.4 days). PF mortality remains high despite treatment in a Level I trauma center. Trauma mechanism, initial blood pressure, pelvic radiography, and CT can be used to predict a patient's mortality risk.

Blunt Gastric Injuries

Gastric rupture after blunt abdominal trauma is a rare injury with few reports in the literature. The purpose of this study was to review our experience with blunt gastric injuries and compare outcomes with small bowel or colon injuries. All patients with hollow viscus perforations after blunt abdominal trauma from 1992 to 2005 at our level I trauma center were reviewed. Of 35,033 blunt trauma admissions, there were 268 (0.7%) patients with a total of 319 perforating hollow viscus injuries, 25 (0.07%) of which were blunt gastric injuries. When compared with the small bowel or colon injuries, the blunt gastric injury group had a higher Injury Severity Score (22 versus 17, P = 0.04), more patients with a chest Abbreviated Injury Score greater than 2 (36% versus 12%, P < 0.01), and a shorter interval from injury to laparotomy (221 versus 366 minutes, P = 0.017). Multivariate analysis identified five independent risk factors for mortality: age older than 55 years, head Abbreviated Injury Score greater than 2, chest Abbreviated Injury Score greater than 2, the presence of hypotension on admission, and Glasgow Coma Scale 8 or less. The results of this study suggest that mortality in patients with blunt hollow viscus injuries can be attributed to concurrent head and chest injuries, but not the specific hollow viscus organ that is injured.