Traumatic brain injury in Brazil: an epidemiological study and systematic review of the literature

ABSTRACT Background: Traumatic brain injury (TBI) is a serious public health problem worldwide. Although TBI is common in developing countries, there are few epidemiological studies. Objective: To investigate the sociodemographic and clinical features of patients with TBI at the Hospital João XXIII, a public reference center for trauma in Belo Horizonte, Brazil, and to systematically review the available literature on TBI in Brazil. Methods: Clinical and sociodemographic data were collected from electronic medical records for the entire month of July 2016. The literature on epidemiology of TBI in Brazil was systematically reviewed using MeSH/DeCS descriptors in the PubMed and Lilacs databases. Results: Most patients admitted with TBI were male and under 60 years of age. Mild TBI was the most prevalent form and the most common cause of TBI was falls. A Glasgow Coma Scale score below 12, neuroimaging changes on computer tomography, and presence of any medical conditions were significantly associated with longer hospital stay. Brazilian studies showed that TBI affected mainly men and young adults. In addition, mild TBI was the most common TBI severity reported and the most common causes were motor vehicle accidents and falls. Conclusions: Overall, the profile of TBI in this center reflects the data from other Brazilian studies.


INTRODUCTION
Traumatic brain injury (TBI) is defined as an injury caused by external force to the head that results in an anatomical lesion or functional impairment of cranial or encephalic structures. TBI is the leading cause of morbidity and mortality in polytrauma patients and is one of the main causes of death in individuals under 45 years of age 1-3 . TBI can have a variety of causes, from falls to car accidents.
Because of its medical and socioeconomic burden, TBI is a major public health problem worldwide. In the United States, 2.8 million emergency department visits were due to TBI and approximately 124,000 of the most severe cases develop longterm impairment 4, 5 . In a single North American state, the annual direct medical cost of TBI was estimated at $95 million, or $1.67 million per 100,000 people 6 . Although lifetime costs for patients with TBI vary according to their demographic characteristics, the costs in Canadian dollars (CAD) for nonfatal cases was estimated at $2,318 for males and $2,200 for females 6 . In Europe, TBI accounted for 37% of all injury-related deaths and was estimated to cost a total of €22,907 million in 2010 7,8 . Limited demographic and socioeconomic information on TBI is available from developing countries 9 .
Although TBI is widespread in Brazil and seems to have an economic and social impact, there are very few epidemiological studies 10,11 . A previous study reported that 40% of deaths in patients aged 5 to 9 years in Brazil are due to TBI and that for every patient who dies, there are at least another three more patients with long-term sequelae 12 . In addition, the annual cost of hospitalizations due to TBI has been estimated at approximately R$ 156,300,000 (US$ 70,960,000) 11 . Unfortunately, these estimates may not reflect the actual Brazilian reality, due in part to a high rate of unreported cases associated with immediate death and the absence of a nearby emergency unit 11,13,14 . Furthermore, reliable quantification of the impact caused by TBI is usually not accurate because measurements are not standardized and data collection on the incidence and outcome of brain injury is incomplete. Therefore, clinical-epidemiological studies are urgently needed to systematically investigate TBI in Brazil.
The current study aimed to investigate sociodemographic and clinical characteristics of patients admitted to João XXIII Hospital with TBI and to identify factors that may influence TBI morbidity and mortality. Also, the epidemiological data available on TBI in Brazil was systematically review.

Original report
This was an observational study conducted at the Joao XXIII Hospital. This hospital is the main trauma center in the Metropolitan region of Belo Horizonte, the third largest metropolitan region in Brazil with more than five million inhabitants. The study was approved by the Human Research Ethics Committee of the Federal University of Minas Gerais (COEP-UFMG).
All records of patients admitted to the Emergency Department of the Joao XXIII Hospital within one month (July 2016) were evaluated using a structured protocol to obtain sociodemographic and clinical information. The sociodemographic data included: (i) sex, (ii) ethnicity, (iii) marital status, (iv) place of residence (Belo Horizonte, metropolitan area, rural area), and (v) educational level. Clinical variables included TBI features (Glasgow Coma Scale Score [GCS], CT neuroimaging changes, hemodynamic instability, and ventilatory support) and hospital outcome. The following premorbid variables were also recorded: (i) clinical comorbidities (any medical conditions that were either secondary to the TBI or that the patient already had on admission) and (ii) alcohol or illicit drug use (assessed via medical record). The causes, severity, and type of TBI were also recorded. Neuroimaging results were included when available.
Statistical analyses were conducted with Statistical Package for the Social Sciences (SPSS) software, version 17.0. Chi-squared analyses were performed to determine statistically significant frequencies of specific events in subgroups. Binary logistic regression using a backward elimination approach was performed to determine which variables were significantly associated with a longer hospital stay, defined as more than 24 hours, as opposed to patients discharged within 24 hours after hospital admission. At the João XXIII Hospital, patients whose state of consciousness remained stable for 24 hours were discharged. The following variables were included in the initial model: age, sex, GCS score (greater than or equal to 13 or less than 12), comorbidity (presence or absence), neuroimaging changes in computed tomography, and alcohol and drug use. Stepwise backward selection was performed automatically using the SPSS software, version 17.0 (SPSS Inc., Chicago, IL, USA), and exclusion testing was done with the likelihood ratio based on the conditional parameter estimates. The goodness of fit of the logistic regression model was assessed using the Hosmer-Lemeshow test and a Receiver Operating Characteristic (ROC) curve.

Systematic review
A systematic search for TBI studies in Brazil was performed independently by two authors (JLVMB and ASM) in the PubMED and Lilacs databases using the MeSH/DeCS descriptors for *traumatic brain injury, *epidemiology, and *Brazil. The inclusion criteria were as follows: (i) studies evaluating sociodemographic and clinical information on TBI cases in Brazil, (ii) original articles, and (iii) articles in Portuguese, Spanish, or English.

RESULTS
In July 2016, 6,184 patients were admitted to the hospital, with 490 individuals diagnosed with TBI. These 490 individuals accounted for 7.92% of the total admissions during the research period. Four hundred seventy-seven records had enough information to determine clinical outcome by age, while 436 records contained all information required by our research protocol (data not shown).
Patients with TBI were divided into three groups based on their GCS score on hospital admission. Patients who had GSC scores of 13-15 on hospital admission were classified as "mild TBI". Patients with GSC scores of 9-12 and 3-8 were classified as "moderate TBI" and "severe TBI", respectively 11,13 . Patients with mild TBI accounted for the majority of TBIrelated admissions and comprised 87.4% of the total number of TBI cases. Moderate and severe TBI cases accounted for 5.5 and 7.1% of TBI cases, respectively.
Next, we analyzed the mechanisms involved in TBI. The mechanisms of TBI were differed significantly between the different severity categories of TBI. Unspecified fall and traffic accident were the most frequent mechanisms for mild and severe TBI, respectively (data not shown).
Male patients were the most affected by TBI across severity levels (p=0.022). We also analyzed the incidence of comorbidities, CT neuroimaging changes, hemodynamic instability, ventilatory support, and death across TBI severity levels ( Table 2). Severe TBI accounted for the majority of deaths (57.1%), whereas mild and moderate TBI accounted for 21.4% each. These deaths were related to TBI or TBI-associated injuries.
In multivariate analysis, CT neuroimaging changes, the presence of medical comorbidities, and a GCS score of 12 or less remained as significant factors associated with longer hospital stay (>24h). The results are presented in Table  3. The logistic regression model was significant [Hosmer-Lemeshow goodness of fit test (step 5): chi-square=3.177; p=0.204] and predicted variability yielded an area under the curve (AUC) of 0.819 in the ROC analysis ( Figure 1).
In our systematic review, we first identified 148 possible titles in the PubMED and Lilacs databases. Four articles were duplicates, and 114 studies were excluded after title/abstract screening. Of these 114 articles, we set aside one review for further reference screening. Thirty articles were fully analyzed, and 10 of these either did not meet our inclusion criteria or did not contain the required information. Two additional articles were identified in the references of review studies. Also, five additional articles were identified while reading the selected manuscripts, giving us a total of 27 eligible articles ( Figure 2).

DISCUSSION
In the present study, we evaluated the sociodemographic and clinical characteristics of patients with TBI admitted to a public reference trauma center in Minas  Gerais, Brazil. This is the first study to perform such evaluation in the state of Minas Gerais, specifically at one of the largest reference trauma centers in Brazil. It is worth highlighting the large number of patients admitted to this center in a short period of time. The hospital admitted almost 17 patients with TBI every day. Young men were most commonly affected, and unspecific falls were the most common cause of TBI. Overall, these findings are consistent with the results of other Brazilian studies, as shown in our systematic literature review 11, .
The higher vulnerability of men can be explained by sociocultural and behavioral factors, such as higher exposure to urban violence than women 25 . A European systematic review found a preponderance of men in 28 studies in which the male-to-female ratio ranged from 1.2:1.0 to 4.6:1.0 41 . Accordingly, men in the United States have higher ageadjusted rates of emergency department visits and deaths related to TBI 4 . In our sample, TBI occurred more frequently in young adults, with mean ages ranging from 22 to 49 years in different studies 25,41,42 .
In contrast to most Brazilian reports, the current study found that falls were the main cause of TBI, but not traffic accidents 29,43 . One of the largest epidemiological studies conducted in the Brazilian population found that falls were the most common TBI mechanism, similar to our findings 33 . Falls were also the most common cause of TBI in European countries and in the USA 4, 41 .
Approximately 19% of our sample reported having consumed alcohol prior to the traumatic event. Our results show that falls, followed by traffic accidents, were the main causes of TBI in patients under the influence of alcohol. Falls were also the main cause of TBI in patients under the influence of illicit drugs (mainly marijuana and crack), but here traffic accident was followed by physical aggression. It is known that the use of alcohol and illicit drugs favors the occurrence of risky situations 29 . In an American study, it was found that both alcohol and illicit drug use were common before a TBI 43 . In Brazil, it is still unclear what role alcohol and other drugs play in TBI 36 . Most of the studies included in our review did not evaluate alcohol status of patients, and those that did had missing data on such information 15 Table 5. Traumatic brain injury-related consequences, clinical comorbidities, length of hospital stay, and alcohol intake information in epidemiological studies on traumatic brain injury. Regarding the severity of TBI, as determined by the GCS, the majority of our sample was diagnosed as mild (87.4%). Mild TBI was also the most common severity level in the Brazilian studies examined, but studies differed in their sample composition. For example, Marinho et al. analyzed a group of 18-30-year-old individuals -an age group more prone to riskier situations and to moderate and severe TBI 29,31,43 . Faria et al. grouped severe and moderate TBI together and yet accounted for only 52% of the total cases 38 .

Reference
The clinical meaning of mild TBI should not be underestimated, as it has been associated with the development of cognitive and behavioral changes 44 . According to one scoping review, half of patients with a single episode of mild TBI develop long-term impairments in several cognitive domains, including executive functions, learning/memory, attention, processing speed, and language 45 . This review included heterogeneous studies using different cognitive batteries in mild TBI patients at different time points after the traumatic event, which may explain the high rate of cognitive deficits. For example, significant episodic memory deficits can already be observed in the acute phase of mild TBI 46 .
Neuroimaging is an important tool in establishing the prognosis for TBI. Seventy-six of 436 (17.4%) patients had early tomographic/neuroimaging TBI-related alterations. It is well known that the more severe the TBI, the more likely the patient is to have neuroimaging changes 47 . Our results confirm that more than half of the patients with moderate or severe TBI had cranial CT changes. Conversely, about 10% of patients with mild TBI had neuroimaging changes. Few of the Brazilian studies reviewed included their neuroimaging findings, as neuroimaging is not considered cost-effective due to the low rate of positive neuroimaging findings in mild TBI 48 .
The length of hospital stay was less than 24 hours in 73.6% of the cases, as most were mild TBI cases. Conversely, a GCS score of 12 or less on admission, as well as neuroimaging changes and medical comorbidity (i.e., both clinical and psychiatric conditions), were associated with a longer hospital stay. Similar to our results, Sorensen et al. found that lower GCS score and psychiatric comorbidity were significantly associated with delay in hospital discharge in patients with TBI 49 . The length of hospital stay in our systematic review varied widely, probably due to the heterogeneity of the sample and the different protocols for treatment and management of TBI in different clinical settings.
In the current study, 3.6% of our post-TBI patients died (n=18). Mortality rates should be interpreted with caution, considering the heterogeneity of epidemiological studies on TBI. For instance, Fernandes et al. found a mortality rate of 12.0% in a much larger sample that included over 400,000 records from a much longer time window 33 . In Europe, there is also a wide variation in post-trauma mortality rate, ranging from 3.0/10 5 inhabitants per year in both Hannover and Münster (Germany) to 18.3/10 5 per year in Finland and Italy 41 .
In the USA, about one third of all related deaths are diagnosed with TBI 50 .
There are limitations to the present study. Some variables (e.g., level of education) were not available for a significant percentage of patients, reflecting the challenges of clinical data collection in a busy trauma center, and thus preventing a more thorough analysis. Medical records also did not include categories of falls. We were only able to capture serious sequelae during hospitalization, which prevented us from exploring less severe complications, including cognitive, behavioral or motor symptoms, and the associated impact on patients' lives. In addition, the present study was conducted in a time window of one month within one year -which was one of the main reasons that led us to conduct a systematic review. From the literature review, we obtained an accurate snapshot of TBI epidemiology in one of the main trauma centers in one of the largest metropolitan regions of Brazil. We chose the month of July because of winter break -a time of year in which people are more exposed to risky situations (such as car travel) and, consequently, to TBIs.
Future studies with a comprehensive longitudinal evaluation of TBI beyond the acute phase are warranted. The investigation of regional specificities in TBI profile in other Brazilian regions and other developing countries could also provide meaningful clinical and epidemiological information. Only with robust evidence can optimal prevention and rehabilitation measures be implemented, influencing the outcome of this daunting problem.