Introduction

Traumatic brain injury (TBI) constitutes a major health and socioeconomic problem throughout the world [6, 9]. It is prevalent in both low- and high-income countries and affects people of all ages. TBI is called the ‘silent epidemic’ because problems resulting from TBI are often not immediately visible, and TBI patients are not very vociferous. The term ‘silent’ further reflects the common underestimation of the actual incidence and that society is often unaware of the impact of TBI [14]. Epidemiological studies of TBI are essential to the targeted prevention and effective treatment of brain-injured patients.

Epidemiological studies are, however, often confounded by a general lack of clear definitions for TBI. A clear, concise definition of TBI is essential in the attempt to understand the epidemiology.

‘Traumatic brain injury’ has replaced the former term ‘head injury’ as it better captures the importance of the ‘brain’ [28]. TBI was recently defined as: ‘An alteration in brain function, or other evidence of brain pathology, caused by an external force’ [23].

Tagliaferri et al. [38] conducted a systematic review on the epidemiology of TBI in Europe in 2006. In their review they analysed 23 studies published between 1980 and 2003. An aggregated (i.e. fatal plus hospitalised) incidence rate of 235 cases per 100,000 people per year, an average mortality rate of 15 per 100,000 people per year and a case fatality rate of 2.7 % were calculated.

In the past decade, new insights into the epidemiology of TBI have emerged. Epidemiological patterns appear to be changing with an increasing incidence of TBI in the elderly. Various reports claim that mortality in TBI is decreasing [8, 15]. The purpose of this systematic review is to provide a contemporary overview of epidemiology of TBI in Europe with a specific focus on epidemiological patterns and on the methodological quality of epidemiologic studies.

Methods

A search was conducted in the PubMed electronic database using the following search-terms: epidemiology, incidence, brain injur*, head injur* and Europe. Reference lists of review studies and articles included in the review were screened for titles that included the key terms.

Inclusion criteria

Studies were included if they met the following inclusion criteria: (1) published in English in the period 1990–2014 with a full text available; (2) original study; (3) predominantly focusing on the epidemiology of TBI; (4) predominantly focusing on TBI, not on the more general head injury; (5) focusing on the population as a whole, not only on a specific subgroup (e.g. cyclists, rugby players, children, etc.); (6) study period at least 1 year; (7) only including data from 1990 or later; (8) not only focusing on mild TBI; (9) if multiple publications used the same study population, the most recent report was used, as it generally addressed a larger population.

Data extraction

Relevant papers were selected by screening the titles (first step), abstracts (second step) and entire articles (third step), retrieved through the database searches. During each step the title, abstract or entire article was screened to ensure that it met the inclusion criteria. This screening was conducted independently by two researchers (W.P. and R.v.d.B.). Extracted data included source population, study period, study group size, case ascertainment, case criteria, incidence, age distribution, sex distribution, mortality and most frequent cause of TBI.

Methodological quality

Characteristics and methodological quality of selected studies were evaluated with a particular focus on study design, case ascertainment, case definition, patient population and the description of the methodology. We based the evaluation of methodological quality on five elements of the STROBE checklist [39] which were most relevant to the quality of reported incidence and mortality rates: study design, setting, participants, data sources/measurement and study size.

Data and statistical analysis

Data are reported as in the original manuscripts. For calculation of an overall incidence rate in the meta-analysis, we used random effects modelling to address heterogeneity between the studies. Heterogeneity was expressed by the τ2 and I 2 statistics. Tau-squared represents the estimate of the between-study variants in a random effects meta-analysis. A τ2 > 1 suggests the presence of substantial statistical heterogeneity. I 2 represents the percentage of the total variation across studies due to heterogeneity [11]. Comprehensive Meta-Analysis (CMA) software was used for the calculations.

Results

The PubMed search identified 743 articles; 109 duplicates were removed, resulting in 634 potentially relevant citations (see ESM 1). Following the screening of titles, abstracts and entire articles, a total of 28 articles were retained for inclusion in this systematic review (Fig. 1).

Fig. 1
figure 1

Flow diagram of the literature search and selection of articles

Study characteristics

Eight reports were of national populations (Austria, Finland, Germany [2×], Norway, Scotland, Spain and Netherlands). One study compared the epidemiology of TBI between regions of different European countries [22]. Nineteen focused on regions, counties or provinces of one European country. Altogether we found data from sixteen different countries: Norway, Sweden, Netherlands, Italy, Germany, Greece, Finland, France, Austria, Slovak Republic, Croatia, Macedonia, Bosnia, Poland and Scotland.

Fifteen out of the 28 studies had a study period of exactly 1 year, five studies [14, 18, 25, 33, 35] had a study period of 10 years or more. The number of included patients ranged from 247 [12] to 280,000 [7], the size of the total source population from 83,900 [24] to 82,037,100 [34]. Nine studies did not report their source population size. Characteristics of the included studies and results of quality assessment are presented in Table 1.

Table 1 Study characteristics and quality assessment

Methodological quality and incidence

A total of 19 studies met the five selected STROBE criteria. Nine studies did not meet all five criteria, of which two failed on two criteria and a further 2 on three criteria (Table 1). Table 2 summarises details of inclusion criteria, case definitions, severity assessment and reported/calculated incidence rates per year of the selected studies. A large variation was found in inclusion criteria, case ascertainment and case definitions. Eight studies were based on hospital admissions, six on emergency department admissions and four on a combination of both. Other sources used for case ascertainment were death certificates, ICU admissions, hospital discharges, pre-hospital emergencies, or a combination of these. We also found large differences in the case criteria that were used in the studies. Seven studies used ICD-10 codes to define TBI, seven used ICD-9 codes and another two used both. Five studies used the GCS. Other tools that were used to define TBI, were Head Injury Severity Scale (HISS), Abbreviated Injury Scale (AIS) or clinical symptoms. Twenty-one out of 28 studies provided information on the severity distribution of TBI. The severity of TBI was measured by the GCS score in 12 out of these 21 studies. Other methods that have been used to measure the TBI severity were AIS head score, HISS score, or ICD codes. Eight out of 21 studies focused on severe or moderate-to-severe TBIs. In studies that provide complete information on all TBI severities (n = 12; [2, 3, 5, 7, 10, 12, 24, 27, 31, 34, 36, 37]), we see that the percentage of mild TBIs varies between 71 % [24] and 97.5 % [3].

Table 2 Inclusion criteria and incidence rate

These differences make it difficult to compare the incidence. Six out of 28 studies did not report an incidence rate. Out of the remaining 22 studies, five focused on severe or moderate-to-severe TBI [1, 1820, 26]. The other 17 studies focused on patients with all TBI severities. The incidences of these 17 studies displayed a large variation: Pérez et al. (2011) [25] reported an incidence rate of 47.3 per 105 population per year in Spain in 2000–2009, while Andersson et al. (2003) [3] reported a rate of 546 per 105 population per year in Western Sweden in 1992–1993. Including only the studies that focus on patients with severe TBI (n = 4), a range of incidence is reported from 4.1 per 105 population in Norway [1] to 17.3 per 105 population in Aquitaine, France [19]. Fig. 2 illustrates this wide variation of reported incidence rates. We note that studies concentrating on severe TBI [1, 1820] cluster to the left (low incidence) and those including all injuries to the right (higher incidence).

Fig. 2
figure 2

Reported incidence rates for TBI. Rates are expressed per 100,000 population. Each study is marked by an open circle; the size of the blue centre is proportional to the size of the population under study

A meta-analysis of the 17 studies focusing on patients with all TBI severities was performed. Figure 3 shows the large variation of these incidences and a substantial degree of heterogeneity was confirmed on statistical evaluation (I 2 = 99.9 %; Z = 6.687). An overall incidence rate of 262 (CI, 185–339) per 100,000 per year for admitted TBI patients was derived.

Fig. 3
figure 3

Forest plot of incidence rate per study sorted by year of publication. The forest plot represents the meta-analysis on 17 studies focusing on patients with all TBI severities. A random effects model was applied. Incidence rates are denoted by the black boxes and the 95 % CIs by the horizontal lines. The overall incidence rate is represented by the black diamond, where the diamond width correspondents to the 95 % CIs. Heterogeneity is substantial: τ2 = 17650.3; χ2 = 72801.5, df = 16 (p = 0.000); I 2 = 99.9 %

Epidemiological patterns: age, sex and cause of TBI

Table 3 presents demographic data of the study populations. In assessing the age distribution, we must note that some studies only include adults in their study population. With this caveat in mind, we see that, in general, TBI is more prevalent among people aged <25 years and among people >75 years. In three studies [14, 18, 26] an increase is seen in the elderly percentile or the mean age over the years of the study.

Table 3 Epidemiological patterns: age and sex

Mean age varies strongly: Styrke et al. (2007) [36] reported a mean age of 22 years, while Mauritz et al. (2008) [22] reported a mean age of 49 years. The latter study, however, included only severe TBI cases. The variation in mean age probably reflects different case ascertainment and inclusion criteria. In most cases, the mean age in females was higher than the mean age in males.

In all 28 studies, there was a male predominance: the male-to-female ratio ranged from 1.2:1.0 [24] to 4.6:1.0 [22].

In 13 out of 26 studies that provided data on the mechanisms of injury, falls were the most frequent cause of TBI. Road traffic accidents (RTAs) were reported as the most frequent cause of TBI in 11 studies. Table 4 shows the most frequent causes of TBI in the study period and TBI severity. In 8 out of 13 studies that include data from before 2000, RTAs are reported as the main cause of TBI. Falls were dominant in the remaining five studies. Only 2 out of 12 studies that include solely data from 2000 or later report RTA as the main cause of the brain injury. In eight studies, falls were dominant. Thus, over time a clear shift can be seen in terms of leading cause of TBI, namely from RTAs to falls.

Table 4 Most frequent cause of TBI in the study period and TBI severity

Within the studies that focus mainly on more severe TBI, RTA as a cause of injury remains dominant. In this category of studies (moderate-to-severe and severe TBI only), RTA remains the leading cause in six out of eight studies.

A clear correlation was also found between age and mechanism of injury. Falls are most common in two age groups: the elderly and children. In contrast, RTAs are the most frequent cause in the age group of young adults. Also notable is the geographical spread of the mechanisms of injury: Scandinavian countries reported mainly falls, while other countries reported more RTAs.

Mortality rate and case fatality rate

Nine studies reported data on mortality rates (ESM 2). As with the incidence rates, a large variation was found in the mortality rates: from 3.0 per 105 inhabitants per year in Hannover and Münster (Germany) [27] to 18.3 per 105 inhabitants per year in Finland and Romagna (Italy) [14, 31]. This variation can largely be explained by differences in case ascertainment and case definitions. Overall, an average mortality rate of 10.5/100,000 was calculated, but interpretation should be with caution due to the heterogeneity of studies.

The case fatality rate (CFR) expresses disease-specific mortality (e.g. TBI). However, the specificity of the rate is influenced by the inclusion of patients who have died from systemic injuries or non-brain comorbidity. Distinction is made between in-hospital CFR (only in-hospital deaths) and overall CFR (in-hospital and out-of-hospital deaths). CFR is highly dependent on the severity of TBI and age of TBI patients: CFR of TBI in general ranges from 0.9 per 100 patients to 7.6 per 100 patients, while CFR of severe TBI ranges from 29 to 55 per 100 patients. None of the included studies provide information on the difference between CFR in mild TBI compared to severe TBI.

Discussion

In recent decades, substantial research has been conducted on the epidemiology of TBI in Europe. However, a full profit cannot be taken of this potential because data have not been collected in a uniform way [16]. This review illustrates the great variability, previously reported by Maas et al. (2011) [16], that exists in data collection and coding of variables in TBI studies. Differences in case ascertainment and case definition confound comparisons between and analysis across different studies. A general consensus on choice and coding of variables for TBI studies is needed in order to acquire the exact epidemiological evolution of TBI. This is currently facilitated by the (common data elements, CDEsFootnote 1) In context of the epidemiology, the following categories are of great importance: participant/subject characteristics; participant and family history; injury/disease related events. In general, many reports have focused on participant/subject characteristics, but fewer on the other two categories. The CDEs represent a major advance towards standardisation, which is highly relevant both from a scientific point of view and from the perspective of cost-efficiency, as this will obviate repeated development of case report forms for new studies [16].

Variability in case definitions and case ascertainment does not directly influence the methodological quality of individual studies. However, 9 of the 28 studies included in the review did not meet the quality criteria of the five selected elements of the STROBE checklist. We chose to evaluate the methodological quality of the studies included according to a pre-specified checklist with specific criteria, rather than allocating a subjective judgment. We considered the STROBE checklist [39] as the most appropriate tool, and selected five criteria of this checklist as being most relevant to the evaluation of epidemiological studies. However, despite the use of this checklist and pre-defined criteria, an element of subjective assessment remains. Tables 1 and 3 illustrate the need for improvement of methodological quality, as well as a great need for standardisation of studies and their reporting.

Unlike Tagliaferri et al. [38], who reported an average incidence and mortality rate in their review, we used the random effects model of meta-analysis to calculate an overall incidence rate. This model is better suited for the comparison of studies with a large heterogeneity. Based on the random effects model of meta-analysis, we found an overall incidence rate of 262 per 100,000 per year. For sake of comparison, we also calculated a simple average incidence rate. This average incidence rate (only including the 17 studies focusing on patients with all TBI severities) was about 275 per 100,000 population per year. After excluding the aberrant rates from Spain [25] and Western Sweden [3], the average rate was 326 per 100,000 population per year. This estimate differs greatly from the incidence rate of 235 per 105 population per year reported by Tagliaferri et al. [38] in 2006. This could indicate an increase in incidence of TBI in the past decade or an under-registration of TBI in period 1980–1990. The latter is the most likely explanation in high income countries, while an increase in true incidence of TBI has been described for middle and low-income countries [17].

It remains, however, difficult to calculate an average incidence and mortality rate since great variation can be found in the case definitions, inclusion criteria and methods used in the studies. For example, studies that are based on hospital and emergency department admissions will report a higher incidence rate than studies that are only based on one of these two. For this reason, it is important to interpret the average rates in a critical manner. For morality, we calculated an average rate of 10.53 per 105 per year. This rate is lower than the mortality rate of 15.4 found by Tagliaferri et al. [38] in 2006. Interpretation of this decrease should, however, be viewed with great caution, given the heterogeneity between studies and absence of possibilities to adjust for case mix. Table 5 shows a comparison between the review of Tagliaferri et al. [38] and the current review.

Table 5 Comparison with review of Tagliaferri et al. 2006 [38]

More definitive conclusions can be drawn on changing epidemiological patterns. In most of the studies, a peak is seen in the oldest age groups. Some studies even report an evolution of the mean age over the years. These findings confirm the shift, reported by Roozenbeek et al. (2013) [28], towards older age groups over recent decades, especially in high-income countries. In contrast to Tagliaferri et al. [38], who reported RTA as the most common event leading to TBI, we find falls to be the leading cause. Table 4 clearly shows the shift over time from RTAs to falls as the leading cause of TBI. RTA still remains the most frequent cause in the group of severe TBI. However, an interaction may exist with study period as most of the studies on severe TBI contain data from before 2000.

Falls are thus becoming a more and more important cause of TBI, mainly in the high-income regions of Europe. An additional finding is the strong correlation between age groups and mechanism of injury. In the majority of the studies, we found that falls are more common in the youngest and oldest age group. On the other hand, we found that RTAs are most common in young adults. These differences have important implications for targeting prevention campaigns.

Strengths and limitations

We used clear search terms and conducted a thorough and systematic literature search. We attempted to include all the relevant articles and to display the study characteristics and results in a clear manner. However, we should note that some studies may have been missed, e.g. if they did not meet the search terms or were not included in PubMed. The major limitations are inherent to the studies underpinning this review and mainly relate to the differences in case ascertainment and case definitions. Although we used the random effects model of meta-analysis to derive an overall incidence rate, the large degree of heterogeneity identified implies that interpretation should be with caution.

Conclusions

This review does not show any trend towards a decreasing incidence of TBI in Europe. The average mortality rate appears lower than in a previous review. Interpretation of data should, however, be with caution, given existing heterogeneity between reports and major differences in approaches to definitions and case ascertainment. In 2006, Tagliaferri et al. [38] identified a need for high-quality epidemiological studies and collaborative intra-European Union population-based studies. Our review confirms the need for generalised/standardised case definitions, case ascertainment and study methods. We further identify changes in epidemiological patterns with increasing age and identify falls as currently the most common cause of TBI in Europe. This has changed compared with previous studies in which RTAs were the more dominant cause. These changes in epidemiological patterns should inform better targeting of prevention campaigns.