Factors associated with survival in adult patients with traumatic arrest: a retrospective cohort study from US trauma centers

Background Traumatic arrests increasingly affect young adults worldwide with low reported survival rates. This study examines factors associated with survival (to hospital discharge) in traumatic arrests transported to US trauma centers. Methods This retrospective cohort study used the US National Trauma Databank 2015 dataset and included patients who presented to trauma centers with “no signs of life”. Univariate and bivariate analyses were done. Factors associated with survival were identified using multivariate regression analyses. Results The study included 5980 patients with traumatic arrests. Only 664 patients (11.1%) survived to hospital discharge. Patients were predominantly in age group 16–64 (84.6%), were mostly males (77.8%) and white (55.1%). Most were admitted to Level I (55.5%) or Level II trauma centers (31.6%). Injuries were mostly blunt (56.7%) or penetrating (39.3%). The median of the injury severity score (ISS) was 19 (interquartile range [IQR]: 9–30). Factors associated with decreased survival included: Age group ≥ 65 (Ref: 16–24), male gender, self-inflicted and other or undetermined types of injuries (Ref: assault), injuries to head and neck, injuries to torso and ISS ≥ 16 (Ref: < 16) and ED thoracotomy. While factors associated with increased survival included: All injury mechanisms (with the exception of motor vehicle transportation) (Ref: firearm), injuries to extremities or spine and back and all methods of coverage (Ref: self-pay). Conclusion Patients with traumatic arrests have poor outcomes with only 11.1% surviving to hospital discharge. Factors associated with survival in traumatic arrests were identified. These findings are important for devising injury prevention strategies and help guide trauma management protocols to improve outcomes in traumatic arrests. Level of evidence Level III.


Level of evidence: Level III.
Keywords: Traumatic arrest, Survival, Outcome, Injury, Resuscitation Background Traumatic arrests (TA) or traumatic deaths are increasingly affecting the young population worldwide. According to a 2014 report by the World Health Organization (WHO), more than 5 million individuals die from injuries each year which accounts for approximately 9% of global deaths [1]. Main causes of fatal injuries are road traffic injuries, homicide and suicide [1]. TA has also increased in the US by 22.8% between the years 2000 and 2010, affecting mainly young adults [2]. As a result, traumatic fatalities compete with cancer and heart related mortality in the young population [2].
Prior studies that examined traumatic arrests have reported very low survival rates ranging from 1.5 to 12.5% [3][4][5]. Guidelines to withhold resuscitative measures specific for traumatic arrests have also been proposed in previous studies and by major organizations such as National Association of Emergency Medical Services Physicians and the American College of Surgeons Committee on Trauma (NAEMSP/ACS-COT) because of the futility of resuscitative measures for this condition [6,7]. Survival has however been increasing with the improvement of medical interventions as well as the establishment of trauma systems [8,9].
With evolving trauma care, improved survival is expected and factors associated with survival might change. This study used the 2015 dataset from the US National Trauma Databank (1) to examine characteristics and outcomes of patients suffering from traumatic arrests and (2) to identify factors associated with survival in traumatic arrests victims who were treated in US trauma centers.

Study design
This retrospective cohort study used the 2015 public release dataset of the National Trauma Data Bank (collected in 2015 and released in 2017).
An Institutional Review Board Exemption was obtained at the American University Of Beirut to use this de-identified dataset.
The variables included in the database include patient related information (demographic, coverage, outcome), event related information (injury severity, characteristics, diagnosis, pre-hospital, ED and hospital information) in addition to other information related to quality assurance and process of care measures. The injury severity score was categorized into two groups (≤ 15, ≥ 16) to classify all patients as having minor or major trauma [15].

Study population
The 2015 NTDB data set contains 917,865 patients of which 8026 presented to the ED with "no signs of life" defined as "having none of the following: organized EKG activity, pupillary responses, spontaneous respiratory attempts or movement, and unassisted blood pressure" [16]. These patients were assumed to have CPR in progress on admission to the ED as per the NTDB dictionary and were considered to have sustained a traumatic arrest. Only adult patients were included and they were defined similarly to previous studies as those aged 16 years and above [17]. Patients were excluded if age was not recorded (n = 443), if they were transferred from other hospitals (recorded as inter-facility transfer; n = 521) and if outcome was unknown (ED discharge disposition recorded as: not known/recorded (n = 280), not applicable (n = 92), home with services (n = 9), home without services (n = 173), left against medical advice (n = 9), transferred to another hospital (n = 97), other (jail, institution care facility, mental health; n = 6) or if hospital discharge disposition was recorded as unknown; n = 1).

Statistical analysis
Data were analyzed using the Statistical Package for the Social Sciences (SPSS, version 24). Descriptive analysis was initially performed. Categorical variables were presented using frequencies and percentages whereas the continuous variables were summarized using the mean ± standard deviation (SD), median and interquartile range (IQR).
Recoding was done for some variables due to the low frequencies of their categories. For example, the nature of injury which was originally composed of 12 categories (amputations, blood vessels, burns, crush, dislocation, fractures, internal organ, nerves, open wounds, sprains and strains, system wide and late effects, unspecified) was grouped into 7 categories with other nature of injury created by combining the following categories together: amputations, burns, crush, nerves, sprains and strains, system wide and late effects.
Depending on the cell count, Pearson's Chi-Square or Fishers' exact tests were used to compare the proportions of all categorical variables in terms of the outcome variable (survived: yes/no). The descriptive analysis revealed that more than 10% of the following variables (ethnicity: 11.0%, whether patient used alcohol: 13.5%, whether patient used drug: 11.1%) were categorized as being not known/not recorded. To provide accurate estimates, missing data were handled through multiple imputations. Five imputed datasets were generated. An automatic imputation method was used. This option automatically chooses an imputation method based on a scan of the data set.
A multivariate logistic regression using a forward selection procedure was conducted to find the best model that explains the association between hospital survival and all patients' demographic and clinical characteristics. More specifically, a multivariate analysis was performed by taking into consideration all factors deemed to be clinically or statistically significant [age, gender, race, hospital teaching status, geographic region for the hospital, comorbidity, ICD-9-CM mechanism of injury E-Code, indication of the type (nature) of trauma produced by an injury, injury intentionality as defined by the CDC injury intentionality matrix, location where injury occurred, whether patient used alcohol, whether patient used drug, the patient's primary method of payment, mode of transportation, the injury severity score reflecting the patient's injuries directly submitted by the facility regardless of the method of calculation, ICD-9 body region as defined by the Barell injury diagnosis matrix (blood vessels, dislocation, fractures, internal organ, open wounds, unspecified, other), nature of injury as defined by the Barell injury diagnosis matrix (extremities, head and neck, spine and back, torso, unclassifiable by site), ED Thoracotomy procedure. p value of ≤ 0.05 was used to denote statistical significance. The c-statistic of Fig. 1 Study population the regression model (area under the curve = 0.934; p value < 0.001; 95% CI 0.924-0.943) indicated an outstanding discrimination between the two categories of the outcome (died/survived). There was no multicollinearity problem in the regression model since the variance inflation factors of all independent variables were less than 10; these ranged from 1.023 to 4.257.

Discussion
Research involving traumatic arrests is rare. This study examined outcomes of patients with traumatic arrests treated in US trauma centers and identified factors associated with survival in this population using the largest trauma database in the US. These factors have not been previously described in the literature.
The overall survival to hospital discharge among patients with traumatic arrests was 11.1%. This rate is slightly lower than the rate of 12.5% reported by Ahmed et al. [5] but higher than other survival rates reported in earlier studies [13,18]. Young, white and male individuals continue to be mostly affected by traumatic arrests [1,5,11]. Improved survival in traumatic arrests presenting to U.S hospitals has been previously attributed to the increasing frequency of emergency interventions such as ED thoracotomy and other procedures [19]. Variation in survival rates of patients with traumatic arrest is however mostly related to differences in study sample selections with most studies including in the denominator arrests that are not transported to hospitals [20]. Our study included only patients who were transported to a trauma center which might have overestimated the survival rate since patients declared dead on scene are not usually included in the NTDB registry.
There were several factors associated with increased survival in patients with traumatic arrests. Demographic factors positively associated with survival included age group (16-64) (compared to age ≥ 65) and female gender. This differs from previous studies [5,18,21] where no similar associations were reported. Patients in the younger age are usually expected to have better odds of survival because of lower comorbidities. Female gender was significantly associated with survival in this population. A previous study did not identify significant association between female gender and outcomes in severely injured patients [22]. Age stratification was however done in that study to account for hormonal status. This finding in our study needs further examination with the potential role of other unmeasured confounders such as role of hormones based on age category (pre vs postmenopausal) and obesity etc.
Several injuries related characteristics that are relevant to treating physicians were also significantly associated with survival.
Type of trauma (blunt vs penetrating), which is mainly used for classification in trauma studies, was not significantly associated with survival similar to other previous studies [5,18,23]. This is in contrast with other studies that reported better outcomes with either penetrating [24] or with blunt injuries [25]. The sample used for our study was however heterogeneous and not restricted to a specific type which might explain this difference.
Additional classification by mechanisms of injuries revealed that most mechanisms were also positively associated with survival when compared to injury by firearm. While fall-related injuries were previously shown to be associated with increased survival when compared to MVT in blunt traumatic arrests [24,26] this study is the first to examine all available mechanisms in traumatic arrests. The case fatality rate of firearm is usually much higher than any other mechanism of injury [27,28] and is related to several factors including number of entrance wounds, range and site of entrance wounds and intentionality [22,29]. The study findings highlight the need for better understanding of firearm related injuries and for developing preventive measures targeted to improve survival in this population.
Injury body region was also found to be significantly associated with survival: Injuries to vital locations such as head and neck, or torso were associated with poor outcomes. This was expected because of the risk of bleeding and hemorrhagic shock from damage to vital organs [30] and is in line with the ATLS approach to management of trauma patients by prioritizing management according to life-threatening injuries [31]. Other factors were also found to be positively associated with survival such as specific types of injuries (fractures, internal organ damage) in addition to alcohol or drug use. These are more likely related to reporting of such data elements in patients who survive after the initial resuscitation measures. Such patients are expected to have a more detailed documentation of minor injuries or better description of other elements contributing to the injury event.   As expected an injury severity score (ISS < 16) was associated with higher survival in traumatic arrest patients. This finding is in line with previous studies [5,13,18] and validates the need to incorporate ISS in outcome predictive models in not only trauma patients but also in traumatic arrests to avoid futility of extreme measures in resuscitation. ED thoracotomy was negatively associated with survival in our patient population. This is expected since the procedure is highly time dependent and considered as a last resort [19]. Thus, if not used in the appropriate population i.e. penetrating chest injuries, ED thoracotomy may be ineffective [9], as such its success rate misrepresented, especially in a heterogeneous population such as ours.
Our study also identified that hospital location in the South was associated with increased survival (reference Northeast) for patients with traumatic arrests. While improved outcomes in trauma patients have been previously linked to geographic clustering of trauma centers which are primarily located in the Northern area [32] and where patients benefit from the greatest access to trauma Level I and II centers within 45 and 60 min [33], our study did not identify such association. In our study sample, more Others is the combination of the following variables: Drowning/submersion and Fire/flame and Hot object/substance and Machinery and Pedal cyclist, other and Pedestrian, other and Natural/environmental, Bites and stings and Natural/environmental, Other and Overexertion and Poisoning and Suffocation hospitals were located in the South (47.2%) than in the Northeast (13.9%). Patients in the South were more likely to survive as compared to those in the Northeast (17.1% vs. 13.4%) and additional stratification by mortality status revealed that disproportionate distribution of the participant hospitals may be responsible for this apparent survival differences between the two regions (Only two patients with firearm injury survived to hospital discharge in the Northeast as compared to 41 patients in the South). Further research should examine closely the impact of hospital region on survival in trauma patients.
Financial coverage status was also significantly associated with survival. When compared to self-payment or uninsured, all other methods of coverage were positively associated with survival. The available literature reports contradicting findings on financial coverage and association with survival in trauma patients. Greene et al. concluded that insurance status was a predictor of outcome with uninsured patients being at higher risk of death in both blunt and penetrating trauma [34] while Lober et al. [21] noted better survival in patients with no insurance coverage attributing this finding to potentially larger proportion of healthy patients in the uninsured group. In this study of traumatic arrests any insurance status (compared to uninsured) was associated with improved outcomes. Further research is needed to clarify the reasons for this disparity such as examining resources utilization including but not limited to access to surgical procedures.
This study has potential limitations related to its retrospective nature and to availability of data reported in the database. The NTDB uses ICD-9 coding for the data retrieved from different hospitals and is like other national databases subject to coding variations and errors. This study did not include traumatic arrests that were not transported to a trauma center which might have led to overestimation of the survival rate in this group of patients. The NTDB also uses "convenience samples" from disproportionate number of large to small hospitals that contribute to the database. This unequal sample size across regions and the lack of weighting should be taken into considerations when comparing outcomes across different US regions. Another limitation is related to absence of NTDB of important clinical factors (no-flow time, low-flow time) which have been shown to be important predictors of survival. Adding these data elements to trauma registries can improve future trauma studies examining outcomes.
Despite these limitations, the NTDB is the largest database for trauma in the United States of America collecting data from over 900 hospitals and the study findings can be generalized to hospitals in the US and to other settings with similar trauma systems. This study also is comprehensive in examining a heterogeneous sample of traumatic arrests that is reflective of Traumatic arrests treated in different trauma centers in the US. The findings of this study are also important for physicians managing trauma patients in the US and in other areas of the world and familiarity with the identified predictors can allow for better assessment of prognosis when treating traumatic arrests.

Conclusion
Patients with traumatic arrests continue to have poor outcomes with only 11.1% surviving to hospital discharge. Several factors were identified to be positively associated with survival in this population. Survival is higher for younger age group, female gender and with any type of insurance coverage. Patients with traumatic arrests from firearm mechanisms experience poor survival. These findings are important for future studies examining closely different predictors, and for devising injury prevention strategies. More evidence is needed to guide trauma management protocols and initiation of resuscitation to improve outcomes further in traumatic arrests.