This study has shown that hospitalized trauma deaths constitute only 13.5% of all trauma deaths in Abu Dhabi. Eighty-two per cent of the deaths were in young males, and 85% occurred in either the ED or the ICU. Motor vehicle collisions, causing mainly head injuries or injuries to two body regions, and fall from height were the two predominant mechanisms in the patients who died. The in-hospital mortality rate of 1.5% is less than the 4.7% reported in the United States [13] and 8.3% from the UK [14]. The low death rate in this study reflects our registry criteria, which allows for the inclusion of trauma patients with low ISS. However, our median age of 33 years is low compared with the median of 48 years reported from Australia [15] and 64.5 years from UK [14]. This indicates greater years of life lost (YLL) in our setting.

In the suggested trimodal distribution of trauma deaths, over 50% of deaths occur in the first few hours before the patients reach the hospitals [16]. The 86% prehospital death rate in our study is more than the reported range of 30–70% [17]. Furthermore, most of the deaths in the ED were in patients with no vital signs on arrival. These patients could arguably be classified as prehospital deaths, thus further increasing the proportion of those who died in the prehospital setting. The high prehospital death rate may be attributed to the severity of the primary injury or prolonged transfer times. Increased injury severity rate, defined as the ratio of fatalities and injuries per 1000 road traffic accidents was reported in our setting [18]. The current study did not look at the ambulance on-scene and transfer times. However, the geographical distribution of the trauma centers and the land mass of the Abu Dhabi Emirate, which represents 87% of the UAE as shown in Fig. 1, may suggest that delayed transfer time is a contributory factor. This is mitigated by the network of good roads and the interventions provided by our EMS, which are limited to providing airway support and hemorrhage control on the scene. There is a major challenge in preventing the high prehospital deaths in our setting. Pfeifer et al. reported that only 4.9–11.3% of prehospital trauma deaths are preventable. They identified delayed transportation, early treatment, and medical errors as the three main contributors to high prehospital death rates [19]. Helicopters can reduce delays in transfer of trauma patients over long distances, and they are an effective component of a trauma system [20]. Only a small proportion of trauma patients are transferred to the hospital by helicopters in our setting. Intervention measures must focus on these areas to reduce our high prehospital death rates.

Our study has shown two definite in-hospital peaks; one in the ED, which was immediate, and a second in the ICU which occurred over several days. Gunst et al. described a bimodal distribution of deaths, where the first peak occurred in the prehospital phase, followed by a second peak several days later [21]. More recent studies have shown an early peak occurring within hours, followed by a gradual decrease in deaths over time without any peak [22]. Rather than looking at trauma deaths as a function of time [23] because we do not have the prehospital death data, we chose instead to look at the locations in the hospital where the deaths have occurred. This may help to identify where in-hospital intervention measures and resource allocation can be most effective. Our results show the ED (33.3%) and the ICU (51.7%) as the two areas where most of the deaths have occurred and where intervention measures must be directed.

In this study, head injuries accounted for over 45% of the deaths while hemorrhagic deaths (injuries to the abdomen, pelvis and, to some extent, the chest) constituted only 10%. This indicates that head injury remains a major problem in our setting. Earlier deaths from head injuries may have resulted from hypoxia and airway compromise, while the later deaths are most likely caused by severe traumatic brain injuries. Advances in trauma care across the globe have demonstrated a reduction in-hospital trauma deaths due to hemorrhage but little change in death due to traumatic brain injury [24, 25]. Nonetheless, while primary brain injuries can only be mitigated by legislation and primary injury prevention measures, much can be done to prevent secondary brain injuries, such as targeted brain protection from hypoxia, hypercarbia, hypotension, and early neurosurgical intervention. This approach was highlighted in the World Society of Emergency Surgery consensus guidelines on the management of poly-trauma patients with severe head injuries [26]. Adopting these measures, and further development of the existing neuro-intensive care will improve survival in the ICU patients with head injury, which is close to 50% in this study.

About 7% of the reported deaths in this study occurred in the ward. These were mainly elderly patients with limb injuries and normal vital signs on arrival at the hospital. This reflects the population structure of the UAE in which most of the elderly are Emirati while most of the younger population are expatriates who tend to retire to their home countries. The median ED length of stay in these patients was almost double the overall median ED length of stay of two hours. Many studies have reported an increasing trend in trauma in the elderly [27,28,29,30]. Recognizing these high-risk patients will help focus on intervention measures such as preventing falls in the elderly and optimizing their care in the ED by reducing the ED length of stay and instituting measures to reduce injury complications.

There were significant differences in mortality rates in our trauma-receiving hospitals. Hospitals A and E each received over seven thousand trauma cases during the study period. This equates to more than 1000 trauma admissions per year. The two hospitals admitted close to 50% of all trauma cases but accounted for only 25.8% of the total deaths. In contrast, three other hospitals received between three to five thousand cases representing 40% of the total trauma admission but accounted for almost 70% of the total deaths. There are different factors that may explain this finding, including the annual number of treated trauma patients and the proportion of those with ISS > 15 (Additional file 1: Table S1). Rural hospitals lack angioembolization facilities, some lack neurosurgical care, and they vary in developing their own trauma team structure which depends on the local policies of each hospital. This was reflected in the significant variation of the time spent in the ED between the hospitals. Several studies have shown that the regionalization of trauma services leads to better patient outcomes by concentrating trauma resources and expertise in a few centers [31,32,33]. The optimal annual trauma admissions that are associated with improved survival are estimated to be greater than 240 patients/year for patients with ISS of more than 15. In our setting, only Hospital E is close to this standard with the remaining hospitals receiving less than half the recommended numbers. (Additional file 1: Table S1). High volume leads to better trauma care outcomes [32, 34]. A system of hubs and spokes with agreed procedures with the EMS to bypass smaller hospitals and transfer trauma patients directly to the major trauma centers will ensure adequate patient volume and the maintenance of skills and improved patients' outcome.

About 75% of those who died in the ward received no trauma code or only had a consultation. This suggests a significant level of under-triaging indicating that the patients were not considered sick and were therefore given a low priority. Furthermore, the median (IQR) age for the patients who died on the ward was 65.50. Appropriate triaging of elderly trauma patients is challenging. The American College of Surgeons Committee on Trauma (ACS-CoT) adult triage tool performs poorly in this age group. Many triage tools that modify the physiological thresholds have been developed to improve triage reliability. Still, elderly trauma patients continue to be under-triaged which impacts their clinical outcomes [35,36,37]. Adopting any of the new elderly trauma triaging tools may improve outcomes in our population.

Deaths from assaults are uncommon in our setting compared with other countries [38]. Only 7.1% of the deaths in this study resulted from assaults. This has implications for the management of patients who arrive at the ED with traumatic cardiac arrest. Majority of these patients are blunt trauma patients who had road traffic collisions or fall from height having a low survival rate [39]. Additional file 2: Table S2 summarizes the proposed recommendations to improve our trauma system.

Our study has several limitations. There was no prehospital data for on-scene and transfer time. In addition, in some cases, the exact cause of death is unknown, as UAE law does not permit routine post-mortem examination. It is also possible that some patients were treated in other small hospitals, but they were not captured in our study because there was no mortality recorded. Finally, we do not have the catchment population for each hospital as they overlap with each other. Accordingly, we combined hospitals in the same regions to calculate the annual incidence of hospitalized death.