Burden of Neurological Disorders Across the US From 1990-2017

Key Points Question What is the current burden of neurological disorders in the US by states, and what are the temporal trends (from 1990 to 2017)? Findings Systematic analysis of the Global Burden of Disease study shows that, in 2017, the 3 most burdensome neurological disorders in the US were stroke, Alzheimer disease and other dementias, and migraine. The burden of individual neurological disorders varied moderately to widely by states (a 1.2-fold to 7.5-fold difference), and the absolute numbers of incident, prevalent, and fatal cases and disability-adjusted life-years of neurological disorders (except for traumatic brain injury incidence; spinal cord injury prevalence; meningitis prevalence, deaths, and disability-adjusted life-years; and encephalitis disability-adjusted life-years) across all US states increased from 1990 to 2017. Meaning A large and increasing number of people have various neurological disorders in the US, with significant variation in the burden of and trends in neurological disorders across the US states, and the reasons for these geographic variations need to be explored further.


Methods
Details on the methodology of the GBD study overall 15 and in association with the burden of neurological disorders have been published elsewhere. [5][6][7][8][9][10][11][12][13][14] In brief, using statistical modeling techniques, the GBD study allows use of all available epidemiological data (both published and unpublished) and routinely collected data (eg, vital registration, hospitalizations, medical claims) within and just outside of the country, region, or state of interest to provide the most accurate estimates of the disease burden, even for regions or states with no accurate epidemiological data. 15 The burden estimates are presented in terms of age-adjusted annual rates per 100 000 people and numbers of prevalence, incidence, disabilityadjusted life-years (DALYs), and deaths by cause (in millions), with corresponding 95% uncertainty intervals (UIs), and their trends or changes from 1990 to 2017.
Nonfatal estimates were obtained from systematic reviews, surveys, administrative health records, registries, and disease surveillance systems. We used available surveys and databases of claims information for US private and public insurance schemes, corrected for bias in health service encounters, as described elsewhere. 2,16,17 Data sources used for quantifying nonfatal outcomes are available online in the GBD results tool. 18 Fatal estimates were obtained from vital registration data (death records from the National Center for Health Statistics and population counts from the US Census Bureau). The single cause of death was determined using the

International Classification of Diseases, Ninth Revision, Ninth Revision, Clinical Modification; International Statistical Classification of Diseases and Related Health Problems, Tenth
Revision; and International Statistical Classification of Diseases, Tenth Revision, Clinical Modification, with redistributions from less precise codes (such as ill-defined disease [R99] or injuries of undetermined intent [Y21-Y33]) to more specific diseases and injuries using regression methods of the subset of vital registration data for which we have multiple causes of death information. Under International Classification of Diseases rules, injury deaths are classified by the cause of injury (eg, fall or road injury) and not the nature of injury, such as TBI and spinal cord injuries (SCI). Thus, we report on incidence, prevalence, and years lived with disability for these causes, but not mortality. Causes of death data were analyzed using the Cause of Death Ensemble model (CODEm, 19 with corrections for changes in coding practices for underlying causes of death as explained in detail elswhere 20 ), and nonfatal data were analyzed using DisMod-MR version 2.1 (World Health Organization), 21 a bayesian meta-regression tool that adjusts data points for variations in study methods among different data sources and enforces consistency between prevalence, incidence, and mortality. For each disease and injury, years lived with disability for mutually exclusive sequelae (ie, disabling outcomes) were quantified as the product of prevalence and a weighting for severity (with the GBD disability weights). Values for disability weights have been derived from population surveys in 9 countries and an open-access internet survey. 22 To account for co-occurrence of disease and injury outcomes, years lived with disability were corrected for comorbidity, assuming a multiplicative rather than additive function of disability weights. 23,24 To allow comparison with the GBD estimates on the global burden of neurological disorders, 3 14 neurological disorders that are quantified as part of GBD are included in this report.
These are stroke, AD and other dementias, PD, idiopathic epilepsy (ie, epilepsy that is not secondary to any of the other GBD causes), multiple sclerosis (MS), motor neuron disease (MND), migraine, tension-type headache (TTH), TBI, SCI, brain and other central nervous system (CNS) cancers, meningitis, encephalitis, and tetanus. While the US-wide age-standardized incidence, prevalence, mortality, and DALY rates of most neurological disorders declined or remained flat from 1990 through 2017 (Table 2), the absolute number of incident cases, prevalent cases, mortality, and DALYs increased, except for meningitis and encephalitis (Table 1). In 2017, the 5 highest incidence rates were for TTH (  to TBI incident and SCI prevalent cases, where small, nonsignificant reductions were observed). The largest increases were observed for (in order of increase) PD, MND, AD and other dementias, brain and other CNS cancers, and stroke. Figure 2 and Figure 3 show age-standardized incidence, prevalence, mortality, and DALY rates of the selected neurological disorders in the US states in 2017. Age-standardized prevalence, incidence, mortality, and DALY rates for specific neurological disorders in the US states and changes in the rates from 1990 to 2017 and the US states' ranking based on the 2017 estimates are described and shown in eTables 1 through 13 in the Supplement. The different metrics of age-standardized rates varied between the US states from a 1.2-fold difference for TTH to 7.5fold for tetanus. Southeastern states and Arkansas had a relatively higher burden for stroke, northern states had a relatively higher burden of MS, and eastern states had higher rates of PD, idiopathic epilepsy, migraine, TTH, and meningitis, encephalitis, and tetanus (eFigure, eAppendix, and eTables 1 through 12 in the Supplement).

Changes in the Burden From 1990 to 2017
From 1990 to 2017, the direction of trends in changes in the age-standardized incidence, prevalence, mortality, and DALY rates across US states mirrored those for the country as a whole. Over that period of time, there were very few changes in any of the states in the rates of migraine and TTH.

Discussion
To our knowledge, this is the first comprehensive report on incidence, prevalence, mortality, and DALY estimates and their trends from 1990 to 2017 for 14 neurological disorders for the individual 50 US states and the country overall. The study showed reductions in the age-adjusted rates of most burden metrics of stroke, AD and other dementias, TBI, SCI, meningitis, and encephalitis, but increasing numbers of people affected by various neurological disorders in the US, with a significant (up to 5-fold) variation in the burden of and trends in particular neurological disorders across the US states. Falling rates of stroke, AD and other dementias, TBI, SCI, meningitis, and encephalitis might suggest that primary prevention of these disorders are beginning to show an influence, while between-state variations may be associated with differences in the case ascertainment, as well as access to health care; racial/ethnic, genetic, and socioeconomic diversity; quality and comprehensiveness of preventive strategies; and risk factor distribution. For dementia, improving educational levels of cohort reaching the age groups at greatest risk of disease may also be contributing to a modest decline over time. 25 While globally, 6 neurological disorders (AD and other dementias, PD, epilepsy, MS, MND, and headache disorders) in 2017 constituted 4.4% (95% UI, 3.7%-5.3%) 26 of total DALYs and ranked ninth among the leading causes of DALYs, in the US, they constituted 6.7% (95% UI, 6.0%-7.6%) and ranked fifth. In 2017, deaths from these neurological disorders were ranked fifth   3 Given the association of prevalent disorders, such as AD, stroke, PD, and MND with older age, the higher rank of neurological disorders in the US can be explained by the longer life expectancy in the US compared with the world overall. In addition, it may also be attributable to better case ascertainment in the US, including improved diagnosis, surveillance or reporting, and health care access. The exact causes for significant between-state variations in the age-standardized rates of some neurological disorders are unknown, but these may be associated with betweenstate differences in the completeness of medical examinations and accuracy of diagnosis, completeness and accuracy of data from medical claims, and referral patterns to specialized centers (eg, seeking medical advice for brain tumors outside of the place of residence), despite corrections for these measurement biases during data analysis. The highest between-state variations were observed for tetanus, but the number of tetanus cases was very low, and thus large relative variation reflected variation by just 1 or a few cases. However, our data on between-state variations in the burden from neurological disorders concur with previous GBD observations of large differences in the burden of disease among US states 2 and significant geographical variations in the burden of neurological disorders in the world. 3 For example, for stroke, there is a strong increasing gradient from north to south, while for MS, the gradient is decreasing from north to south. Both gradients might have been expected based on our understanding of the epidemiology of both diseases. [27][28][29] While confirming previous observations of the so-called stroke belt mortality in the southeastern United States, unlike previous estimates that identified North Carolina, South Carolina, and Georgia with a higher stroke mortality rate than the other states of the stroke belt, 30,31 our 2017 GBD stroke mortality estimates show that the stroke belt is now in Alabama, Mississippi, and South Carolina. Our dementia mortality estimates also showed a higher percentage change (increase) in the stroke belt states compared with the other states, likely attributable to the recipro-  cal association between stroke and dementia. 32 These findings can be used to set research priorities (eg, identifying causes of between-state variations in the burden of neurological disorders). These data will also allow health care professionals and policy makers on national and state levels to allocate resources (eg, number of hospital beds, specialists, services) and give priorities for improving care for people with some major neurological disorders in each US state.
The finding that the 3 most burdensome neurological disorders in the US in terms of the absolute numbers of DALYs are stroke, AD and other dementias, and migraine is in line with ranking of age-standardized DALY rates for neurological disorders found in other high-income countries in 2016. 3 There were diverse changes in the rates of all neurological disorders in the country overall (eg, decreased incidence rates of stroke, AD and other dementias, TBI, SCI, and meningitis, and increased incidence rates of PD, MS, MND, and brain and other nervous system cancers) and across all US states from 1990 to 2017. However, the absolute number of people who are affected by noncommunicable neurological disorders over that period has increased substantially and is likely to continue increasing because of aging of the US population and population growth. Unfavorable trends in some lifestyle factors (eg, overweight, fasting plasma glucose level) 2 in the US are also likely additional contributing factors for the increasing burden from some neurological disorders (eg, stroke, dementia). 33 These developments are consistent with the global trends 3 and support the call to action to reduce the burden of neurological disorders in the United States, as outlined by Gooch et al. 1 Although we understand that individuals may have more than 1 neurological condition, assuming no overlap between stroke, TBI, SCI, brain and other CNS cancers, meningitis, encephalitis, tetanus, and other neurological disorders (AD and other dementias, PD, idiopathic epilepsy, MS, MND, migraine, TTH, and other neurological disorders), it can be estimated that in 2017 more than 200 million Americans (60% of the population) were afflicted by at least 1 neurological disorder, ranging from TTH and migraine to stroke and dementia. This is twice the estimate by Gooch et al 1 and 9 times the estimate by Borlongan et al. 34 The greater than previously reported overall prevalence of neurological disorders in the US may be explained by at least 3 factors. First, we included many disorders in this study that Gooch et al 1 did not, such as TTH (the most prevalent of all neurological disorders analyzed in the GBD study), brain and other CNS cancers, tetanus, meningitis, and encephalitis, although they 1 did include chronic low back pain, which we did not. The selection of neurological disorders included in this analysis was largely based on 2 factors: (1) the GBD study currently provides estimates of the burden of only these specific neurological disorders, and (2) these neurological disorders are considered by the GBD study as conditions for which neurologists play a particular important role in care and diagnosis (as opposed to, for example, low back pain, for which the role of a neurologist is less dominant). Second, Gooch et al 1 based their analysis on previously published articles only, while the GBD study analyses also included administrative data and modeling, thus allowing estimates for US states with no epidemiological data on a particu-lar neurological disorder. Differences in the methodologies may explain differences in the estimates of stroke prevalence (7.8 million in the GBD study vs 6.8 million in the Gooch et al study 1 ), AD and other dementias (2.9 million vs 5.3 million), epilepsy (1.4 million vs 2.8 million), TBI (2.1 million vs 1.4 million), and SCI (2.2 million vs 0.3 million). The large difference in SCI cases may reflect a weakness in GBD methods, which rely on sparse data on the association between the incidence of a cause of injury (eg, falls and road injury) and the nature of the injury (eg, fall or SCI), which should be examined in future rounds of GBD. Our GBD estimates of PD (0.6 million), migraine (69 million), MS (0.4 million), and MND (38 000) are close to those reported by Gooch et al (0.6 million, 72 million, 0.4 million, and 18 000-30 000, respectively). 1 Third, the Gooch et al 1 study was based on articles published prior to 2017, with many of them referring to the data collected 5 to 10 years ago; thus, ongoing aging of the population and population growth may partly explain the greater prevalence of selected neurological disorders in our study.

Limitations
General limitations of the GBD study discussed elsewhere 3 fully apply to this report and may account for differences between these data and those collected by other governmental or disease-specific organizations. Specifically, for this report, we were not able to provide burden estimates for all neurological disorders combined, or burden estimates by age and sex, because these estimates will be the subjects of separate reports. In addition, some very prevalent neurological disorders (eg, restless leg syndrome and peripheral neuropathy are not currently estimated by GBD) and some common disorders (eg, low back and neck pain) are not regarded as neurological disorders in this article, although they are at least partly within the realm of neurology. Inclusion of these currently unaccounted disorders would increase the estimates of the burden of neurological disorders. Other limitations specific to this report include (1) difficulty in determining death from dementia because coding practices have changed by orders of magnitude over the last 30 years; (2) lack of data to quantify headaches in the US; and (3) accuracy of medical claims and hospitalization data and potential inaccuracies in the adjustments of administrative data to those reported in epidemiological studies. Because GBD study estimates are updated annually, the current limitations can be addressed, and with an increasing amount of data being added year to year, a more accurate picture of the burden of neurological disorders in the US can be made.

Conclusions
In summary, this report showed that, while there were reductions in the rates of most burden metrics of stroke, Alzheimer and other dementias, TBI, SCI, meningitis, and encephalitis, there was a large and increasing number of people affected by various neurological disorders in the US, with a significant variation of the burden of and trends in particular neurological disorders across the US states. The reasons for geographic variations among different US states need to be explored further. Health care professionals and policy makers at the national and state levels can use the information reported in this article to advance their health care planning and resource allocation, including research funding, to prevent and reduce the morbidity and mortality of neurological disorders.