Predictive Factors for Fatality After Traumatic Brain Injury Among Road Trac Crash Victims in Addis Ababa City, Ethiopia

Methods This study aimed to examine the predictive factors of short-term mortality after severe brain injury due to a road trac crash. The study was done on a prospective cohort of 242 severely brain-injured patients selected using cluster sampling in Addis Ababa City hospitals. The study was conducted from February 2018 to November 2019. Data were collected from brain-injured patients using a questionnaire and recorded ndings within the rst 24 hours of admission, Survival Analysis was used for statistical analysis. Ethical clearance was obtained from the Addis Ababa University, College of Health Sciences Institutional Review Board (IRB). Condentiality of information about injured patients was maintained. In this study, Survival Analysis was employed. Descriptive analysis of survival data utilized nonparametric methods to compare the survival function of two or more groups. The Kaplan-Meier estimate (product-limit-estimate) of the survival function was employed. The Log-rank test was utilized to test whether observed differences in survival experience between/among groups were signicant or not. The multivariable model used a semi-parametric regression model known as the Proportional Hazard Regression (PHR) model to estimate the hazard of death.


Abstract
Background Traumatic brain injury (TBI) is one of the common preventable causes of mortality and disability among road tra c victims worldwide, most especially in low-and middle-income countries, including Ethiopia.
Objective to determine risk factors of mortality after traumatic brain injury due to road tra c crash.

Methods
This study aimed to examine the predictive factors of short-term mortality after severe brain injury due to a road tra c crash. The study was done on a prospective cohort of 242 severely brain-injured patients selected using cluster sampling in Addis Ababa City hospitals. The study was conducted from February 2018 to November 2019. Data were collected from brain-injured patients using a questionnaire and recorded ndings within the rst 24 hours of admission, Survival Analysis was used for statistical analysis. Ethical clearance was obtained from the Addis Ababa University, College of Health Sciences Institutional Review Board (IRB). Con dentiality of information about injured patients was maintained.

Results
In this study, the death rate was 73(30.2%). The majority of TBI patients accounting for, 186(81%) were men. The median age of TBI patients was 29 years. The hazard for those patients with subnormal body temperature was 1.64 times that of normal temperature (AHR: 1.64; CI: 2.14-10.29). The estimated fatality hazard ratio for patients who experienced Glasgow Coma Scale (GCS)below six was 5.61 times higher compared to GCS six to eight (CI:3.1-10.24).

Conclusion
In conclusion, there was high early mortality of patients (30.2%) in Ethiopia. Being men, young and lower GCS were associated with higher mortality hazards. Hence, optimum advanced neuro-surgical prehospital care programs are urgently needed.
Background Traumatic brain injury (TBI) constitutes a considerable portion of the global injury burden caused primarily by road injuries and leads to high levels of disability and mortality. Traumatic Brain Injury [TBI] is de ned as an alteration in brain function or anatomic structure induced by mechanical force such as a motor vehicle crash that caused temporary or permanent neurological dysfunction [1]. TBI is also de ned by both the initial primary injury and subsequent secondary injuries. Primary damage occurs when TBI happens due to initial impact, while secondary brain damage ensues the development of mass lesions which frequently exacerbates by systemic insults as a consequence of intrinsic pathophysiologic mechanisms [2]. TBI can be diagnosed and estimated by clinical examinations of the level of consciousness using scoring systems such as the Glasgow Coma Scale(GCS), Computed Tomography (CT) imaging, and the assessment of other vital parameters such as intracranial pressure [2,3].
The pathophysiology of severe TBI can be viewed as a two-step process [1,2]. The rst step is primary injury, in which there is a wide range of TBI pathologies, such as focal contusions, and space-occupying intra-extradural hematomas, and diffuse axonal injury [4]. Diffuse axonal injury (DAI), also known as shear injury or traumatic axonal injury, refers to intracranial injury caused by rapid and sustained deceleration or acceleration of the brain [4,5]. The second step is the secondary injury that occurs as a consequence of the primary injury [2][3][4]. It is especially because most of its consequences, especially cognitive impairments are not obvious [4]. It is also estimated that the majority of patients with severe TBI have other body site traumatic injuries [2]. A study showed that TBI plus one additional injury was the most common injury [3]. Studies revealed that an increased mortality rate following TBI [5][6][7][8]. Several factors contributed to TBI mortality. Studies revealed that being young and male is an important risk factor for TBI [6,[8][9][10]. Glasgow coma scale score and presence of hematoma on CT as independently signi cant predictors of survival. The studies also showed factors related to a poor outcome such as age > 38 years, Glasgow Coma Scale score < 8, subdural hematoma, and development of secondary systemic insults (respiratory, circulatory, and metabolic) [5,8].
It was stated that the brain depends on the mean arterial pressure for perfusion, making maintenance of cerebral perfusion and oxygenation crucial in the setting of TBI as the brain is so vulnerable to ischemic injury [7]. A study in Scandinavia revealed that arterial hypotension and hypothermia were pre-hospital risk factors for mortality [11]. Signi cant morbidity and mortality occur when a patient experiences both hypoxia and hypotension, even when it is merely a brief instance of each [2,12]. Patients with hypotension that is not corrected in the eld had a worse outcome than those whose hypotension was corrected in the eld (during scene and transport time) and at the emergency department [13,14].
Hypothermia is related to uncontrolled bleeding in the prehospital and emergency settings [14]. TBI patient mortality increases with the magnitude of hemodynamic instability, and mainly the extent of the injury with bleeding [15] Thus, neurological monitoring of these patients is essential to limit episodes of decreased cerebral perfusion pressure [16].
Detecting traumatic brain injury (TBI) patients with a high risk of mortality is important to maximize the resource for trauma care [17]. Studies show that when the effective post-crash response was practiced, secondary injury is potentially preventable and represents endpoints for goal-directed resuscitation using evidence-based research input [2,18]. Once a crash has happened there should be fast, easy, and adequate medical care to increase the chance of survival and to limit the physical consequences [5].
Glasgow Coma Scale (GCS) measure is a signi cant and reliable indicator of the severity of TBI, and it can help to identify improvement or deterioration in neurological status through repeated measurements during emergency management [6]. Prediction of the patient outcome can be useful as an aid to clinical decision making, to explore possible biological mechanisms, and as part of the clinical audit process [5].
In Ethiopia, the study regarding the extent of early mortality after severe traumatic brain injury due to road tra c crashes is limited. Therefore, the purpose of this study was to determine predictive factors of fatality after traumatic brain injury due to road tra c crashes and identify the extent of early mortality.

Study design, period and settings
We conducted institutional base prospective cohort study, with follow-up of severe TBI patients from the time 24 hours after crash brain injury admission up to 30 days or earlier death. Minilik II is also believed to be the rst in importing two cars in Addis Ababa and introduced car technology in the City for the rst time in 1907 [20]. In Addis Ababa, there were 28,361 crashes of which 4,433 of them harming humans during 2019 [20]. This study predicts the risk of mortality after severe Traumatic Brain injury among vehicle crash victims in Addis Ababa City.

Source population
The source population was all charts of severe traumatic brain injured patents due to vehicle crash victims that visited the tertiary trauma care hospitals in Addis Ababa City. Study population Victims of road tra c crashes who were diagnosed with severe traumatic brain injury based on Computed Tomography scan nding recorded by examining medical doctors were the study population. The study included every individual TBI patient with an age of greater or equal to 18 years old, having severe Traumatic Brain Injury (TBI) due to road tra c crash with a measured intracranial lesion Abbreviated Injury Score (AIS) ≥ 4 on the base of the Computer Tomography (CT) scan ndings recorded within the rst 24 hours of admission [18]. TBI patients with un-recordable vital signs, like pulse, blood pressure that indicates pre-hospital physiological status, died during the pre-hospital time, had no signs of brain trauma, and were referred or transferred to other hospitals were excluded.
Sample size determination and Sampling procedure The study initiated when individual diagnosed for severe TBI and admitted in the tertiary hospitals but have not experienced the death event at the time of ascertainment, and prospectively followed to observe the death event. The participants were ether experience the mortality event or censored at the 30th day of follow up to exit the study.
Participants in this mixed cohort were recruited based on brain injury severity status in which all those ful lled severe TBI diagnostic criteria and admitted for it included. We ascertain risk factors such as age, sex, GCS, type of brain injury and TB severity status at the time of enrollment, and then prospectively followed the individuals with the severe TBI to observe the death events of interest. The study initiated when individual diagnosed for severe TBI and admitted in the tertiary hospitals but have not experienced the death event at the time of ascertainment, and prospectively followed to observe the death event. The participants were ether experience the mortality event or censored at the 30th day of follow up to exit the study.
Participants in this mixed cohort were recruited based on brain injury severity status in which all those ful lled severe TBI diagnostic criteria and admitted for it included. We ascertain risk factors such as age, sex, GCS, type of brain injury and TB severity status at the time of enrollment, and then prospectively followed the individuals with the severe TBI to observe the death events of interest. The total sample size of eligible Severe Traumatic Brain Injury patients reached to 242 samples during the study period.
All tertiary care hospitals in Addis Ababa that provide trauma care service were identi ed. All hospitals with functional computed Tomography scan were listed. The randomly selected three hospitals (Minilic II Hospital, Tikur Anbessa Hospital and AbaT Hospital) that provide service to severe brain injury patients were recruited. The study subjects were recruited based on eligible criteria during the study period from the selected hospitals emergency departments when the injured patient admitted within 24 hours crash injury event.

Data collection instrument and techniques Variables
Dependent variable Mortality at 30 days or earlier The response in this research was the "survival time/time for death". It was de ned as the number of days from the date of crash injury for which the patient arrived at the Emergency Department of Hospitals assessed and followed up for 30 days for survival or death outcome. The survival data studied here were "right-censored". Independent variable Socio-demographic characteristics; age, sex, religion, marital status, Clinical factors: traumatic brain injury type on the rst day of crash occurrence, GCS, hypoxia, hypotension, and hypothermia.
Operational de nitions Initial neuro-physiological variables Glasgow Coma Scale (GCS);is a neurological scale aimigto provide a reliable, objective way of recording the conscious state of a person. unconscious patients were de ned as presenting a GCS <9.

Result of the descriptive analysis
Socio-demographic characteristics Prehospital arterial hypotension was de ned as systolic blood pressure <90 mmHg measured at time point within 24 hours of hospital admission including hospital arrival value.
Prehospital hypoxia was de ned as body oxygenation with a pulse oximeter oxygen saturation <90% measured at any time point within 24 hours of hospital admission, including hospital arrival value. Prehospital hypothermia was de ned as body temperature with a thermometer measurement ≤ 35.0°C measured at any time point within 24 hours of hospital admission including hospital arrival value. Fatality is de ned as a severe traumatic injury patient's death during 30 days or earlier after the crash. The data collection tool included a structured questionnaire responded to by severely injured patients or relatives. Besides, patient record data with measured traumatic brain injury severity and the vital sign was collected by using a checklist adapted from another study [5,6,21]. The Survival data for this study were obtained from selected brain-injured patients collected by emergency medical doctors and emergency MSc nurses at emergency departments. The data collectors and supervisors with the principal investigator have followed up the injured patients from the rst day of crash injury assessment to 30 days for the patient's survival or death.

Data Quality Assurance
The training was given to data collectors and supervisors about the objectives of the study and how to collect data in detail.
The questionnaire was developed from previous studies [5,6,[10][11][12], translated from English to Amharic and again to English by language experts to ensure consistency. To ensure the quality of data, a pre-test of the data collection tool was done in Zewditu Hospital. The supervisors and investigator facilitated the data collection process. The principal investigator worked with data collectors to assure the trustworthiness of the data and to minimize inter-observer bias. Every day, the collected data were checked for completeness of data including vital signs, level of consciousness, Computed Tomography scan nding, missing value, and consistency using the checklist. Double data entry was done to ensure accuracy.
Data processing and analysis Data were coded and entered into the computer using STATA 14 statistical software package. Outliers were cleaned and validated data were prepared for analyses. In this study, Survival Analysis was employed. Descriptive analysis of survival data utilized nonparametric methods to compare the survival function of two or more groups. The Kaplan-Meier estimate (product-limit-estimate) of the survival function was employed. The Log-rank test was utilized to test whether observed differences in survival experience between/among groups were signi cant or not. The multivariable model used a semiparametric regression model known as the Proportional Hazard Regression (PHR) model to estimate the hazard of death.
In this study 242 severe Traumatic Brain Injured (TBI) patients were observed. There were 73(30.2%) deaths. Of those, 65(89%) of deaths occurred among patients with subnormal body temperature and 8(11%) among normal temperature. Ninety-four percent of TBI patients with normal body temperature and 50% of TBI patients with subnormal temperature survived for 20 days. The hazard of death during the rst 10 days for TBI patients with subnormal temperature and normal body temperature was 40% and 12% respectively. Overall Survival and hazard functions are shown in gure 1. Table 1 indicates that seventy percent of TBI patients were right-censored and the remaining uncensored. TBI patients lived for an average of 24 days (95% CI: 22.69-25.40 days). The incident rate was 0.365 per person-30 days. The majority of TBI patients accounting for, 186(81%) were men. The incidence of death for age groups above fty years old were 16(52%). The incidence of death for married and unmarried were 47(30.1%) and 26(30.2%) respectively. The incidence of death for Muslims and Orthodox were 16(28.1%) and 44(31.0%) respectively.      Table 3 Shows results based on the log-rank test. The p-values show differences in survival experience between two or more levels of predictors. The predictors that manifested differences in levels of survival function are indicated as follows.

Results of Multivariate analysis
The cox model procedure that includes model selection, tests, diagnosis, and t con rmed showed no problems concerning the interaction of main effects and cofounding. Therefore, the results in Table 4 are based on the bi-variable and multi-variable analyses. It should be pointed out that variables with p-values below 0.05 were considered statistically signi cant. Based on bi-variable cox regression, the estimated HRs for those age groups greater or equal to 50 are associated with a poorer prognosis compared to the 19 to 29 age group (CHR: 2. 18; CI: 2.21-3.96), the multi-variate model is not statistically signi cant. The result of the multivariate analysis revealed that the hazard of death for those patients with subnormal body temperature is 1.64 times that of normal temperature (AHR: 1.64; CI: 2.14-10.29).

Discussion
This study nds out that patients with severe traumatic brain injury have a high mortality rate (30.2%), especially, the men and the younger were the victims. The rate of mortality reported in our study was higher than the study conducted in New York and Tunisia, which were 25%, and 26.9% respectively [2,8], and lower than the studies conducted in Uganda [6], South Africa [8]. But our study involved only shortterm mortality. Our study revealed that female Traumatic Brain Injury (TBI) patients had on average similar survival times to those of men. The majority of severely brain-injured patients, accounting for 80% were men which is similar to other studies conducted in Paris [1], USA [5], Tunisia [8], and Uganda [6,9]. The possible explanation might be since men's vehicle users are younger and energetic to exert forceful driving that could lead to a crash. Our study showed that the mean age of severe traumatic brain injury was 36 years which is lower than the study conducted in the USA [5] and higher than the study conducted in Uganda [6] and similar to studies conducted in Uganda and Paris [1,9]. This difference might be explained by the fact that in our setting, younger victims were mostly exposed to crashes which are dangerous to lead to severe brain injury.
A previous study showed that primary injury has a wide range of TBI pathologies, such as diffuse axonal injury, focal contusions, and space-occupying intra-and extradural hematomas [2]. That study was similar to our nding which revealed that 38%deaths were due to hemorrhagic cerebral contusion. That might be due to not wearing protective devices.
This study nds out that the most fatal skull fracture was basal skull fracture and multiple skull fracture. Moreover, patients with TBI have often committed other body site traumatic injuries. In this study, the traumatic brain injury patient who undergoes deep comma had the least survival. The estimated hazard ratio for those patients who experienced deep comma was more likely to die compared to comma patients. This is in agreement with other studies that indicate lower on-scene GCS associated with higher mortality [7,14,15]. This might be probably due to worsening of primary impact as a result of lack of advanced life support care for damaged brain tissue as well as lack of anesthetist who secures airway of victims in the prehospital setting. In the current study, hypotension resulted in a major fatality, accounting for 51.5% of deaths. Based on bi-variable cox regression, the estimated HRs for those who were hypotensive was 2.83(CI:1.78-4.48). This nding was similar to another study [2]. The majority of deoxygenated severe traumatic patients (51.7%) died. Those patients having deoxygenated had lived eight days' shorter survival time on average than those having oxygenated. Based on bi-variable cox regression, the estimated hazard ratio for those who were deoxygenated were 3.6 times more likely to die compared to oxygenated (CI: 2.37-6.16). This nding was in agreement with previous ndings that stated hypoxemia as a prehospital risk factor for the early fatality of severely brain-injured patients [14,15]. This might suggest that pre-hospital care lacks immediate airway management with an anesthetist to prevent deoxygenation.
Our study found out that those patients who had traumatic brain injury plus other site injuries were twofold more likely to die compared to those who had isolated traumatic brain injury. That might be probably due to the limitation of pre-hospital neurosurgical care integrated with other body injury care on scene and during transportation.
The combination of hypotension and hypoxia occurring before arrival at the pre-hospital time is associated with a signi cant increase in death compared with either physiologic insult alone. This study was inconsistent with another study [12[. This might be due to a lack of immediate advanced neurosurgical care at the scene and transportation time to manage the primary impact of the brain tissue.
Limitation of the study This study may have some limitations: First, the data were collected where the crash injury victims have been in physical and psychological stress situations that can limit recalling the exact measure of scene and transport time values. Second, the study mainly focused on pre-hospital care of TBI, but prehospital data were not originally documented on the scene and during transportation for this study, but the data were veri ed by the supervisor and facilitator to reduce biases. Third, the study did not consider the previous status of the patient's physiological status like hypovolemia and pathological status like comorbidity. Fourth, collecting data at the time of hospital arrival may not be inclusive to collect the overall situation of scene time and transport time characteristics of crash injured victims. Lastly, this study did not consider long-term mortality after 30 days of TBI and neurological de cit as a consequence of TBI, especially cognitive impairments that are not obvious during the study period.

Conclusion
In conclusion, there was high early mortality of TBI patients (30.2%) in Ethiopia. The lower GCS was associated with a higher mortality hazard, which may lead to a higher proportion of TBI patients with poor outcomes. Post-injury events, such as hypoxemia, hypotension, and intracranial hypertension may also initiate the pathophysiological mechanisms of secondary brain injury. This demonstrates the need for early correction of the alarming situation in the pre-hospital setting to lower the risk of secondary brain injury. Trauma patient mortality increases with the magnitude of hemodynamic instability that require appropriate skilled practice programs on pre-hospital TBI management are urgently needed.
Further study long-term outcome TBI that includes long-term mortality and disability to large extent should be considered.