Global, regional, and national burden of traumatic brain injury and spinal cord injury, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016

1Traumatic brain injury (TBI) and spinal cord injury (SCI) are increasingly considered to be important global health priorities. These injuries not only cause health loss and disability for individuals and their families, but also represent a burden to health-care systems and economies through lost productivity and high health-care costs. Given that the injuries that lead to TBI and SCI are frequently preventable, there is also value in measuring the extent to which different causes of injury lead to TBI or SCI to help to understand the effect that injury-prevention programmes could have.

2Many epidemiological studies have been limited by difficulties in comprehensively measuring the incidence of cross-injury sequelae such as TBI and SCI, and have instead focused on the incidence of the causes of injury, such as falls, road injuries, and interpersonal violence. As a result, few comprehensive epidemiological assessments have been done across all sources of injury, despite increasing dialogue about the long-term neuropsychological consequences of concussions in young people and professional athletes playing sports and about the risk of TBI from head trauma in bicycle crashes and other causes of injury., Epidemiological studies that have focused specifically on TBI and SCI without estimation of all potential causes of injury have identified substantial burdens, but are often limited by relying on locations where incidence data were available without adopting modelling strategies for estimation of the burden in locations where data were sparse.,,,,,, Epidemiological assessments have been done in low-income and low-middle-income countries but typically have been limited by poor availability of data.,, Few studies have reported age-standardised incidence rates, which would enable comparison between countries with different populations, and the studies that have reported such data showed that the incidence rates of TBI and SCI vary substantially between countries., These studies have not measured the relative disability caused by different injuries over time; such data are important because, whereas injuries such as fractures might be disabling only in the short term, conditions such as cognitive impairment from TBI or paraplegia from SCI can leave patients with lifelong health loss. In general, measurement of the burden of TBI and SCI in greater geographical and demographic detail—and over time—is of substantial value.

3Evidence before this study Previous epidemiological studies of the incidence and outcomes of traumatic brain injury (TBI) and spinal cord injury (SCI) have been limited by focusing on certain subpopulations, including only select injuries, or by providing estimates only for areas of the world with accessible data. Previous Global Burden of Diseases, Injuries, and Risk Factors (GBD) studies have reported the burden of injury by cause of injury, such as self-harm, road injuries, and falls, but have not reported results by nature of injury sustained as a result of those causes, including TBI and SCI. To date, no studies have systematically measured the burden of TBI and SCI globally for all countries, ages, and sexes through recent years and from all causes of injury. To identify sources of injury data that could inform an assessment of non-fatal burden from TBI and SCI, we used results from the GBD 2016 injuries estimation process, which included systematic reviews of injury incidence data for all causes of injury that were initially done for GBD 2010 and updated as new data and literature studies became available in GBD 2013, GBD 2015, and GBD 2016. Inclusion criteria for the systematic reviews were representative, population-based surveys; reporting of injuries incidence; and clinical records from general hospitals, outpatient primary care facilities, and health insurance claims when such data were available with injury diagnosis codes. In this study, we updated a previous review of injuries data done for the World Bank that contributed to GBD 2010, GBD 2013, and GBD 2015 by searching the Global Health Data Exchange for surveys, hospital datasets, and literature studies in any language that were tagged as having injury-related data up to Dec 31, 2016. Added value of this study In this study, we used for the first time the GBD 2016 framework to report estimates of the global, regional, and national burden in terms of incidence, prevalence, and years of life lived with disability of TBI and SCI for 195 countries and territories. We have provided these estimates globally, by region, and by Socio-demographic Index quintiles in 2016, as well as the percentage change since 1990. We also provide estimates of the proportions of TBI and SCI caused by different causes of injury for each geographical region in 2016. Although epidemiological assessments that focus on particular populations have been done, no other studies of TBI or SCI have provided estimates in this level of detail for all countries derived from a standardised, systematic approach. We were able to measure uncertainty in our estimates by using the uncertainty propogation methods used throughout the GBD study. Implications of all the available evidence Our estimates suggest that TBI and SCI are severely disabling injuries. The global burden of TBI increased significantly between 1990 and 2016, whereas that of SCI has not changed significantly over time in terms of age-standardised incidence and prevalence. Age-standardised incidence and prevalence of TBI and SCI were high in central Europe, eastern Europe, and central Asia; the incidence and prevalence of SCI were high in North America and western Europe. Addressing the global burden of these conditions requires improved efforts to decrease the causes of SCI and TBI (eg, fall-prevention strategies, reducing alcohol overuse, and improving road safety, all of which could help to prevent injuries or decrease injury severity) and improved access to, and quality of, medical and social care (which could improve survival and reduce morbidity). People with TBI or SCI can have other medical conditions that require close supervision and might benefit from rehabilitation and medical care to reduce disability. Hence, although injury prevention efforts are key, health-care systems should also anticipate a growing burden from caring for people with TBI and SCI. These conditions could necessitate special focus within health-care systems, because they can be medically complex and burdensome for patients, clinicians, and families. In the future, development of improved methods for surveillance of TBI and SCI will be important, particularly in low-income settings, as will development of methods to identify patients with TBI who do not seek medical care.

4Evidence before this study

5Previous epidemiological studies of the incidence and outcomes of traumatic brain injury (TBI) and spinal cord injury (SCI) have been limited by focusing on certain subpopulations, including only select injuries, or by providing estimates only for areas of the world with accessible data. Previous Global Burden of Diseases, Injuries, and Risk Factors (GBD) studies have reported the burden of injury by cause of injury, such as self-harm, road injuries, and falls, but have not reported results by nature of injury sustained as a result of those causes, including TBI and SCI. To date, no studies have systematically measured the burden of TBI and SCI globally for all countries, ages, and sexes through recent years and from all causes of injury. To identify sources of injury data that could inform an assessment of non-fatal burden from TBI and SCI, we used results from the GBD 2016 injuries estimation process, which included systematic reviews of injury incidence data for all causes of injury that were initially done for GBD 2010 and updated as new data and literature studies became available in GBD 2013, GBD 2015, and GBD 2016. Inclusion criteria for the systematic reviews were representative, population-based surveys; reporting of injuries incidence; and clinical records from general hospitals, outpatient primary care facilities, and health insurance claims when such data were available with injury diagnosis codes. In this study, we updated a previous review of injuries data done for the World Bank that contributed to GBD 2010, GBD 2013, and GBD 2015 by searching the Global Health Data Exchange for surveys, hospital datasets, and literature studies in any language that were tagged as having injury-related data up to Dec 31, 2016.

6Added value of this study

7In this study, we used for the first time the GBD 2016 framework to report estimates of the global, regional, and national burden in terms of incidence, prevalence, and years of life lived with disability of TBI and SCI for 195 countries and territories. We have provided these estimates globally, by region, and by Socio-demographic Index quintiles in 2016, as well as the percentage change since 1990. We also provide estimates of the proportions of TBI and SCI caused by different causes of injury for each geographical region in 2016. Although epidemiological assessments that focus on particular populations have been done, no other studies of TBI or SCI have provided estimates in this level of detail for all countries derived from a standardised, systematic approach. We were able to measure uncertainty in our estimates by using the uncertainty propogation methods used throughout the GBD study.

8Implications of all the available evidence

9Our estimates suggest that TBI and SCI are severely disabling injuries. The global burden of TBI increased significantly between 1990 and 2016, whereas that of SCI has not changed significantly over time in terms of age-standardised incidence and prevalence. Age-standardised incidence and prevalence of TBI and SCI were high in central Europe, eastern Europe, and central Asia; the incidence and prevalence of SCI were high in North America and western Europe. Addressing the global burden of these conditions requires improved efforts to decrease the causes of SCI and TBI (eg, fall-prevention strategies, reducing alcohol overuse, and improving road safety, all of which could help to prevent injuries or decrease injury severity) and improved access to, and quality of, medical and social care (which could improve survival and reduce morbidity). People with TBI or SCI can have other medical conditions that require close supervision and might benefit from rehabilitation and medical care to reduce disability. Hence, although injury prevention efforts are key, health-care systems should also anticipate a growing burden from caring for people with TBI and SCI. These conditions could necessitate special focus within health-care systems, because they can be medically complex and burdensome for patients, clinicians, and families. In the future, development of improved methods for surveillance of TBI and SCI will be important, particularly in low-income settings, as will development of methods to identify patients with TBI who do not seek medical care.

10The Global Burden of Diseases, Injuries, and Risk Factors (GBD) study is the product of a global research collaboration that quantifies the effects of hundreds of diseases, injuries, and risk factors around the world, producing annual estimates of all-cause mortality, causes of death, non-fatal health outcomes, and risk factors. Within the GBD framework, estimates for TBI and SCI burden have not previously been available as reported results. Instead, these nature-of-injury codes were incorporated as part of the analytic process that computed disability and results were ultimately provided only by cause (eg, falls) rather than by nature of injury (eg, TBI). Here, we describe an approach for estimation of nature-specific non-fatal burden estimates for all injuries, and report the incidence, prevalence, and years of life lived with disability (YLDs) for TBI and SCI, as well as the proportion of TBI and SCI caused by different injuries by region.

11Our approach to measuring TBI and SCI was developed within the GBD 2016 study framework. In the GBD 2016 study, standardised analytic methods were used to estimate incidence, prevalence, and YLDs by age, sex, cause, year, and location. The study was an attempt to use all accessible information about disease and injury occurrence, clinical course, and severity that passed a set of inclusion criteria. The comparability of data was optimised by adjusting for different case definitions, enforcing consistency between data for prevalence, incidence, and cause of death estimates, and predicting estimates for locations with sparse data by borrowing information from other locations and covariates. These methods, data, and criteria are described in more detail in other GBD 2016 reports.,,,,

12Detailed elements of the GBD methods for measurement of TBI and SCI (including case definitions and severity definitions), a flowchart for our TBI and SCI estimation, and overall GBD study methods are in appendix 1. The measurement of TBI and SCI burden had two key deviations from the standard GBD framework. First, the GBD cause hierarchy categorised both TBI and SCI as being a nature of injury as opposed to a cause of injury—ie, these conditions previously had been measured as consequences of causes of injury. For example, a cause, such as a fall, could lead to SCI. Historically falls have been measured and reported but the actual nature of injury (eg, TBI, ankle fracture) that occurred because of the fall has not been directly reported. This aspect of the GBD study design was consistent across other natures of injuries. Second, estimation of TBI and SCI deviated from the GBD study framework in terms of the measures that were reported for the conditions, because we do not estimate death from TBI or SCI. Although TBI and SCI can lead to death, they were not considered causes of death in the GBD 2016 framework. Instead, the cause of injury (eg, falls) that led to a nature of injury such as TBI was considered the cause of death. For example, an individual who had a fall, sustained a TBI, and then died while in hospital after the injury would be considered to have had a death caused by a fall and an incident TBI. In this study, we estimated the non-fatal burden and therefore report incidence, prevalence, and YLDs, but not cause-specific mortality or years of life lost.

13The process for estimation of incidence, prevalence, and YLDs was as follows. First, the incidence of 29 different causes of injury (appendix 1) were modelled with DisMod-MR 2.1, a meta-regression tool that was used extensively throughout the GBD study. These cause-of-injury models measured the incidence of each cause of injury that required medical care, which included patients who were admitted or seen in an outpatient clinic and received a diagnosis code for a given cause of injury. Receiving an injury diagnosis code did not preclude the possibility of death in the hospital or after discharge. Each of these cause models used an array of data types, including surveillance studies, literature studies, hospital discharge records, and emergency department records. The details of these models have previously been described in more detail. Although we do not estimate death from TBI or SCI in this study, our modelling strategy also included cause-specific mortality rates from the cause of death ensemble model to inform incidence estimates for cause-of-injury models such as road injuries in data-sparse areas using estimates from data-rich areas. The outputs from these models were estimates of inpatient (admitted) and outpatient incidence rates of causes of injury and were specific for location, sex, age, and year. The outpatient incidence of each cause was derived from the inpatient incidence on the basis of a regression coefficient for outpatient incidence that was extracted from DisMod-MR 2.1 incidence models in locations that had both inpatient and outpatient data.

14Clinical record data that coded for both cause and nature of injury were used to estimate the proportion of each cause that resulted in each nature of injury. If an injury cause resulted in more than one nature of injury, the most severe was chosen on the basis of a mixed-effects regression model that estimated the disability experienced by an injured individual adjusted for age, sex, and never-injured status, with country and individual random effects. Because SCI was associated with higher disability than TBI (appendix 1), SCI was chosen if both conditions occurred as a result of the same injury. We used this method after finding in a previous GBD study that statistically assigning multiple injury categories to a single individual was difficult because of a sparsity of data. This process and the severity rankings are described in more detail in appendix 1. These proportions were calculated for each external cause-of-injury–nature-of-injury (cause–nature) combination, such that the proportions of all natures of injury for a given cause of injury sum to 1 because of a Dirichlet regression. The output from this step was incidence for each cause–nature combination.

15From the incidence estimates for each cause–nature combination, we separately modelled short-term and long-term estimates using proportions of individuals expected to experience short-term versus long-term disability (the cutoff for long-term disability was 1 year). The proportions estimated to experience permanent health loss generally increased with age and were different for TBI and SCI (appendix 1). The short-term prevalence estimates were then calculated on the basis of average duration of a short-term case, whereas the long-term estimates were considered to be permanent and underwent comorbidity adjustment as described previously. Cause–nature incidence rates were converted to prevalence with the differential equation solver used in DisMod-MR 2.1. This solver reconciled the incidence rates from the previous steps with standardised mortality ratios derived from literature studies to estimate prevalence, because people with long-term disability due to TBI and SCI die at a higher rate than the background mortality in the population. The final output from this step was prevalence of each cause–nature combination for each location, year, age, and sex combination.

16YLDs were then calculated by multiplying the prevalence by the disability weight. Measurement has been described in more detail previously, but in summary, disability weights were measured through population and internet surveys on the basis of lay descriptions of disabling conditions. For example, the disability weight for short-term mild TBI and for short-term moderate or severe TBI were 0·110 (95% uncertainty interval [UI] 0·074–0·158) and 0·214 (0·141–0·297), respectively, meaning that the affected people experienced health losses of 11·0% and 21·4%, respectively, compared with a person in full health. All disability weights for different severities of TBI and SCI are provided in appendix 1.

17After estimation of YLDs, the prevalence, incidence, and YLDs for TBI and SCI were then summed across all causes to estimate the all-injury prevalence, incidence, and YLDs for TBI and SCI separately. Uncertainty was propagated throughout this process by maintaining distributions of 1000 draws for each estimation stage (including percentage change over time). We use the 25th and 975th sorted values in the draw distributions as the upper and lower UIs for mean estimates and for percentage change, whereby change was judged to be significant if the lower and upper UIs did not overlap zero. This process is consistent with management of uncertainty throughout the GBD study framework.

18We grouped countries into quintiles on the basis of their 2016 Socio-demographic Index (SDI) value, which is a composite measure of development derived from income per person, educational attainment, and total fertility rate. Additionally, we measured the most common causes of TBI and SCI separately in terms of the original cause of injury that led to the disability. Finally, we measured the proportion of TBI that was mild versus the proportion that was moderate or severe and the proportion of SCI that occurred at the neck versus below the neck and present these values at the global level. Analyses were done in Python (version 2.7), Stata (version 13.1), and R (version 3.3). Statistical code used for this study will be made available upon publication of this Article via the Institute for Health Metrics and Evaluation. This study complies with the Guidelines for Accurate and Transparent Health Estimates Report (GATHER) recommendations (appendix 1).

19The funder of the study had no role in study design, data collection, data analysis, data interpretation, or the writing of the report. All authors had full access to the data in the study and had final responsibility for the decision to submit for publication.

20We used incidence data for every cause of injury and every GBD region. The number of sources by injury and by region are in appendix 1. Incidence, prevalence, and YLD estimates for every cause of injury by age, sex, and location for 1990–2016 are available through an online results tool.

21Table 1 shows the incidence and prevalence of TBI in terms of all-age counts, age-standardised rates (per 100 000 population), and percentage change in age-standardised rates between 1990 and 2016. Table 2 shows the same information for SCI. YLDs from TBI and SCI in terms of all-age counts, age-standardised rates, and total percentage change are in appendix 2, which also includes these estimates by age and sex, and for 1990. Between 1990 and 2016, age-standardised incidence rates significantly increased by 3·6% (95% UI 1·8 to 5·5) for TBI and decreased non-significantly by −3·6% (−7·4 to 4·0) for SCI, leading to age-standardised incidence rates of 369 (331 to 412) per 100 000 for TBI and 13 (11 to 16) per 100 000 for SCI (Table 1, Table 2).

22Figure 1, Figure 2 show the age-standardised incidence by country for 2016 for TBI and SCI, respectively. Central Europe, eastern Europe, and central Asia had substantially higher incidence rates of TBI than the rest of the world; at the regional level, the age-standardised incidence rate was highest in central Europe, at 857 (95% UI 750–988) per 100 000 (table 1). Syria had the highest age-standardised incidence rate of TBI of any country, with 1322 (95% UI 481–2779) cases per 100 000. Slovenia (1092 [938–1294] per 100 000) and the Czech Republic (1022 [885–1191] per 100 000) had the next highest age-standardised incidence rates. The incidence rates for SCI were highest in the high SDI regions high-income North America (26 [20–33] per 100 000) and Western Europe (26 [20–33] per 100 000; table 2). However, at a country level, Syria had the highest age-standardised incidence rate of SCI (136 [25–441] per 100 000), followed by Yemen (42 [15–111] per 100 000), Iraq (37 [13–105] per 100 000), and Afghanistan (37 [11–101]; table 2).

23In terms of individuals living with disability from these conditions in 2016, TBI had a global age-standardised prevalence of 759 (95% UI 731–788) per 100 000 (table 1), and SCI had a global age-standardised prevalence of 368 (340–409) per 100 000 (table 2). These estimates corresponded to 55 million (53–58) individuals with TBI and 27 million (25–30) with SCI (for unrounded estimates see table 1). From 1990 to 2016, the age-standardised prevalence of TBI increased by 8·4% (95% UI 7·7 to 9·2; table 1), whereas that of SCI decreased non-significantly by −0·2% (−2·1 to 2·7; table 2). Age-standardised prevalence of TBI was high in the super-region of cental Europe, eastern Europe, and central Asia at 1539 (1464–1614) per 100 000, representing roughly 7·5 million prevalent cases (7·1–7·9). Age-standardised prevalence for SCI was highest in high SDI regions—specifically western Europe (854 [780–945 per 100 000) and high-income Asia Pacific (821 [747–907] per 100 000; table 2).

24TBI and SCI caused 8·1 million (95% UI 6·0–10·4) and 9·5 million (6·7–12·4) YLDs, respectively, in 2016. The age-standardised YLD rates were 111 (82–141) per 100 000 for TBI and 130 (90–170) per 100 000 for SCI (appendix 2). The global age-standardised YLD rates per 100 000 population for TBI increased by 8·5% (7·6–9·3) from 1990 to 2016 and those for SCI decreased by 10·0% (7·0–13·3) from 1990 to 2016. At the country level, for TBI, the distribution of YLDs was similar to those of incidence and prevalence. Specifically, countries in central Europe, eastern Europe, and central Asia had the highest age-standardised YLD rates, with country-specific rates ranging from 135 (99–175) per 100 000 in Tajikistan to 335 (241–421) per 100 000 in Slovenia. For SCI, the high-income super-region had the highest age-standardised YLD rates (229 [159–303] per 100 000). Within these locations, Finland (287 [197–381] per 100 000), Ireland (283 [192–373] per 100 000), and Israel (282 [181–396] per 100 000) had the highest age-standardised YLD rates.

25Figure 3 shows the global age-specific and sex-specific incidence rates per 100 000 for minor TBI, moderate or severe TBI, spinal cord lesions at the neck, and spinal cord lesions below the neck for 2016. For TBI, these figures show divergent patterns between males and females that start in teenage years and extend to ages 50–60 years (figure 3). At older ages (ie, older than 60 years), the sex-specific incidence rates in males and females is similar (figure 3). The incidence is more similar between the sexes for both subtypes of SCI than for TBI, although men have higher incidences than women of spinal cord lesions at the neck level at ages 20–40 years (figure 3).

26The proportion of causes leading to TBI and SCI by region are shown in figure 4. In general, falls were the main cause of TBI. In some regions, such as central Europe, more than 50% of the age-standardised incidence of TBI was caused by falls; in other regions, such as Oceania, falls were still the predominant cause but accounted for a smaller proportion of the age-standardised incidence (figure 4). In addition to having high age-standardised incidence, prevalence, and YLDs attributable to TBI, central and eastern Europe also had the highest incidence of TBI caused by falls. The second most common cause of TBI in most regions was motor vehicle road injuries (figure 4A). The main cause of SCI in most regions was also falls, which accounted for more than 50% of age-standardised incidence in nine different GBD regions (figure 4). Conflict and terrorism was the most common cause in North Africa and the Middle East in 2016 (figure 4B).

27This study, in which we used the GBD framework to estimate the non-fatal burden of TBI and SCI, is to our knowledge the first effort to quantify the burden of these conditions at global, regional, and national levels for all ages and sexes, and over time, from 1990 to 2016. Globally, these conditions cause non-fatal health loss that is distributed across various levels of income, geographies, and the lifespan, and represent a substantial proportion of global injury burden that could be averted through injury prevention and safety measures.

28We identified an increase in global age-standardised incidence, prevalence, and YLDs of TBI between 1990 and 2016. This increase probably reflects the increasing rates of falls and road injuries over this period, which could in turn be due to increased use of motor vehicles, unsafe road conditions, and, in some areas, increased rates of alcohol consumption or unsafe infrastructure.,, By contrast, we noted no significant change in the age-standardised incidence or prevalence of SCI, although with global population growth, the absolute number of people living with the effects of SCI is expected to increase. The increasing global incidence of both TBI and SCI starting approximately at age 70 years also shows the importance of preventive measures for injuries through all years of life—particularly in the context of an ageing global population—and of adequate access to acute medical care resources such as emergency medical services and emergency department care.

29Regional patterns differed between TBI and SCI. The highest incidence rates of TBI were in central Europe, eastern Europe, and central Asia, whereas the highest incidence rates of SCI were in high-income North America, western Europe, and high-income Asia Pacific. Conflict-affected countries in the Middle East—ie, Syria, Yemen, and Iraq—and Afghanistan had the highest country-specific incidence of SCI, and Syria also had the highest incidence of TBI. Rates of TBI and SCI were lower in some low SDI countries in regions such as sub-Saharan Africa, which generally corresponded with the geographical patterns of falls and road injuries in those regions as reported in GBD 2016.,, These variations in the underlying causes of TBI and SCI probably explain much of the geographical variation in the incidence of TBI and SCI. Access to health-care resources could also explain some geographical variation. For example, the higher prevalence of SCI in North America and western Europe could be related to survival bias in high SDI areas, whereby medical services have led to successful resuscitation in injury victims who otherwise would have died without receiving a TBI or SCI diagnosis code. The high rates of TBI in central Europe, eastern Europe, and central Asia generally correspond with the high all-injury rate estimated in those regions in GBD 2016.,,

30Our findings show that, globally, falls and road injuries were the most important cause of non-fatal cases of both TBI and SCI, reflecting the findings for all 328 diseases and injuries from GBD 2016, in which falls were the tenth leading cause of age-standardised YLDs from 1990 to 2016. This burden of falls was particularly evident in our study for central Europe, eastern Europe, and central Asia, where falls were the second most common cause of disability in 1990 and the third most common cause in 2016. Although the context in which a fall occurred could not be established in this study because of a lack of International Classification of Diseases (ICD) coding detail, falls can be preventable irrespective of where they occur. Falls leading to SCI have been associated with alcohol use in countries such as Estonia, so risk factor profiles across countries could explain some geographical patterns in this study. Road injuries were also important causes of these conditions, suggesting that achievement of Sustainable Development Goal 3.6 (“By 2020, halve the number of global deaths and injuries from road traffic accidents”) could reduce the burden of conditions such as TBI and SCI that can result from road injuries.

31Our estimates for TBI incidence diverged from estimates in other published literature. Our study relied on cause-of-injury models that by design estimate the incidence of injuries requiring medical care. A limitation of this approach is that some people with TBI, particularly mild TBI, might not seek medical attention after injury and are thus not captured in the analysis, which could lead to underestimation of the global burden of TBI., In a study done in New Zealand, in which proactive screening methods were used to contact people after an accident to their upper body (including use of broad ICD-10 codes [S00–09] in addition to community-based case-ascertainment sources to identify individuals not seeking medical treatment), the incidence of TBI was 790 per 100 000 (substantially higher than that in our study), and approximately 30% of people with mild TBI did not seek medical attention soon after their TBI. However, this study was done in only one country, and the findings can probably not be generalised to the global population. However, the findings of that study emphasise the need for other international studies to use a comprehensive community-based approach for case ascertainment to increase the accuracy of GBD estimates.

32In general, our study had similar limitations to other GBD studies, but with the added complexity and uncertainty of measuring TBI and SCI within other injury estimates, which has not been done previously in the GBD framework. In terms of TBI-specific and SCI-specific limitations, we used medical record data extensively in our modelling process, which might not be representative of the entire population. This point is pertinent because most of the dual-coded clinical data that was used in the derivation of cause–nature proportions was from high-income countries. Additionally, the derivation of the incidence coefficient that adjusts for injuries receiving outpatient care was based on limited data. These factors could have introduced selection bias, which was addressed to some extent by incorporation of income and health-care access in our modelling process. However, by relying on medical care records, we potentially did not include people with mild TBI who did not seek medical care, which therefore could be a source of detection bias leading to underestimation, although we addressed this issue by using cause-of-injury incidence models for all injuries requiring medical care, followed by a Dirichlet-based modelling approach of cause–nature combinations.,, An additional limitation stems from the studies examining how TBI and SCI can occur together. A proportion of people can experience an SCI from a traumatic event and also experience TBI, and because of the disability-ranking approach that we used in our cause–nature proportion analysis, these patients would be assigned SCI as their nature of injury. Experiencing both SCI and TBI can also complicate recovery, and presence of non-brain injuries in people with TBI can affect survival, although estimates of the cumulative effect are outside the scope of this analysis. The ICD codes used to identify SCI cases also include some injuries that do not necessarily lead to paraplegia or tetraplegia, and some such injuries, such as spinal cord contusions, can improve over time. Additionally, emerging evidence about long-term deficits such as dementia, stroke, and increased risk of engagement in antisocial behaviour linked to TBI were not included in our disability computation.,,, The long-term neurological and psychological sequelae of TBI are poorly understood, and the epidemiological, neuropathological, and psychiatric analyses intended to understand the resultant disabilities will be important to incorporate in future assessments. Similarly, our analysis does not capture cohort effects over time, a limitation that can be addressed in future GBD studies. Overall, the long-term sequelae due to TBI and SCI suggest that further work in terms of measurement of long-term disability is needed to measure the effect of these conditions more accurately, and to ensure that the disability weights accurately reflect the health loss observed in clinical practice and experienced by individuals; such further work could influence future research into disability-weight measurement via health loss surveys. The limitations we describe also show how more research is needed, particularly in low-income areas of the world, to collect comprehensive injury data. Focusing of resources on injury epidemiology data could improve the accuracy of future epidemiological assessments of TBI and SCI.

33In conclusion, the age-standardised incidence, prevalence, and YLD of TBI are increasing globally, whereas age-standardised rates of SCI have not changed (although the number of individuals with SCI is likely to be increasing globally). In view of the expense and complexity of managing patients with TBI and SCI, ministries of health, medical systems, and social support infrastructure should focus on development and improvement of injury-prevention strategies, although maintenance of short-term and long-term care pathways to mitigate health loss and improve outcomes among patients with TBI and SCI is also crucial. Finally, measurement of the burden of these conditions could be improved with the establishment of registry systems for patients with TBI and SCI worldwide, which could help to facilitate further research and intervention efforts and improve the accuracy of future epidemiological assessments of these important conditions.