The GBD study is a comprehensive and methodological analysis of global diseases. Using the estimates from this study, the present global, regional, and national disease burdens can be compared and assessed (9). The research adheres to the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) (10). The research, conducted by the Institute for Health Metrics and Evaluation (IHME) at the University of Washington, utilizes anonymized data, which eliminates the requirement for informed consent. The stroke data analyzed in this study were sourced from GBD 2021, which offers the most recent epidemiological estimates of 371 diseases and injuries across 21 GBD regions and 204 countries and territories from 1990 to 2021. All data are freely available at the Global Health Data Exchange¹ (11), with comprehensive information on the data, methodologies, and statistical modeling available in previous reports (5, 6). The relevant data were anonymous and publicly accessible. Our Institutional Review Board approved the waiver of informed consent for this study (approval number: EUMC 2025-01-05). This manuscript was produced as a part of the GBD Collaborator Network and is in accordance with the GBD Protocol.

We assessed the overall stroke burden encompassing both total stroke events and specific stroke subtypes, including ischemic stroke, intracerebral hemorrhage (ICH), and subarachnoid hemorrhage (SAH). Stroke cases within the GBD framework were identified according to the World Health Organization (WHO) criteria as acute neurological deficits of vascular origin lasting >24 h or resulting in death, confirmed through clinical diagnosis and neuroimaging. Incident stroke was classified as the first clinically diagnosed stroke event, as determined by physician assessment and imaging, which is consistent with WHO standards. Ischemic stroke includes vascular occlusive events that cause cerebral infarction and excludes transient ischemic attacks. Hemorrhagic strokes comprise spontaneous (nontraumatic) events categorized as ICH, defined by intraparenchymal bleeding, and SAH, characterized by bleeding within the subarachnoid space, both verified through neuroimaging (4, 6). Case ascertainment for ischemic and hemorrhagic strokes utilized validated International Classification of Diseases (ICD-9 and ICD-10) diagnostic codes (Supplementary methods), with stroke burden estimates reflecting only the first-ever stroke occurrence in an individual’s lifetime.

For the GBD study, systematic reviews from various data sources were analyzed along with data from the National Health Interview Survey, the National Health and Nutrition Examination Survey in the US, health system administrative data such as the Healthcare Cost and Utilization Project (HCUP), and additional national claims data (12). Studies or datasets with small sample sizes (<150 participants), review articles, non-population-based studies, or those lacking clear subpopulation definitions were excluded by the IHME. Methods for gathering data on nonfatal outcomes and mortality have been detailed in previous studies (5, 11). Fatal estimates were obtained from vital registration and surveillance data. Detailed search terms for the systematic analysis are demonstrated in Supplementary methods (5, 11). To investigate the burden of stroke, prevalence, incidence, disability-adjusted life years (DALYs), years of life lost (YLLs), years lived with disability (YLDs), and mortality were estimated. DALYs are a measure of health loss in a population. One DALY represents the loss of 1 year of full health. DALYs were calculated by adding YLLs to YLDs. Detailed methods for determining the risk factors for GBD are demonstrated in Supplementary methods.

Bayesian meta-regression methods were implemented in the MR-BRT software (IHME) to systematically account for known methodological biases across studies. Adjustments were specifically applied to studies derived exclusively from administrative data sources, such as the HCUP, those combining first-ever and recurrent stroke events, studies reporting only aggregated data for stroke subtypes, and studies in which prehospital stroke deaths were omitted from stroke definitions or measurement protocols (13, 14). DisMod-MR, an epidemiologic state-transition modeling software, was used to estimate the stroke incidence and prevalence. Stroke mortality, including subtype-specific mortality, was derived from vital registration data classified using ICD coding standards. Although verbal autopsy data are generally incorporated into the GBD methodology for aggregated stroke mortality estimation, these data were excluded from subtype-specific estimates because of limited stroke-related medical history details. In the specific context of the US, verbal autopsy data are omitted entirely because of the availability of detailed vital registration records. Statistical modeling methods have been employed to improve comparability among diverse mortality data sources. Population-attributable fractions were independently computed for each risk factor based on exposure levels, relative risk values derived from meta-analyses, and theoretical minimum risk benchmarks tailored for each specific risk-outcome combination. To address comorbidities, simulations involving 40,000 hypothetical individuals per demographic subgroup (defined by age, sex, country, and year) were performed, assigning each individual a probability of developing conditions reflective of the observed prevalence rates. The reported 95% uncertainty intervals (UIs) for all estimates were derived from 1,000 posterior draws from the Bayesian models and expressed as intervals spanning the 2.5th to 97.5th percentiles of the resulting distributions (13, 14).

In 2021, the US experienced approximately 0.41 million incident strokes (95% UI, 0.36–0.47), of which ischemic strokes accounted for 0.31 million cases (75.6%; 95% UI, 0.26–0.37). The incidence of ICH and SAH were 0.07 million (95% UI, 0.06–0.08) and 0.03 million (95% UI, 0.03–0.03), respectively. The number of prevalent stroke cases was 6.3 million (95% UI, 5.89–6.76), with 3.07 million occurring among men (48.7%; 95% UI, 2.84–3.46) and 3.23 million among women (51.3%; 95% UI, 3.04–3.46). Prevalent cases of ICH and SAH totaled 0.75 million (95% UI, 0.68–0.84) and 0.45 million (95% UI, 0.41–0.50), respectively. Stroke accounted for approximately 0.19 million deaths (95% UI, 0.16–0.21), comprising 0.08 million deaths among men (95% UI, 0.07–0.09) and 0.11 million among women (95% UI, 0.09–0.12). DALYs lost owing to stroke amounted to 3.91 million (95% UI, 3.53–4.23), with 1.84 million DALYs among men (95% UI, 1.71–1.97) and 2.07 million among women (95% UI, 1.81–2.25) (Table 1).

From 1990 to 2021, the burden of stroke in the US, as measured by the absolute number of incident cases, prevalent cases, DALYs, YLDs, and YLLs, showed an overall increasing trend (Table 1 and Supplementary Figure S1). However, the numbers of stroke incidents, deaths, DALYs, and YLLs exhibited a transient decline during the intermediate period within this interval. For nearly all evaluated measures, including total stroke, ischemic stroke, ICH, and SAH, the magnitude of the increase was consistently higher in men than in women, except for YLL, which was attributed to ischemic stroke. Specifically, between 1990 and 2021, the ischemic stroke incidence increased by 10.18% (95% UI, 3.78–17.9%), while the incidence of ICH and SAH increased by 28.01% (95% UI, 20.77–36.94%) and 49.72% (95% UI, 42.2–60.09%), respectively. The stroke prevalence exhibited substantial growth, with the prevalence of ischemic stroke increasing by 65.66% (95% UI, 57.32–74.33%), ICH by 78.31% (95% UI, 68.53–87.17%), and SAH by 70.58% (95% UI, 62.72–78.91%). Stroke-related mortality increased by 16.1% (95% UI, 9.02–20.34%) for ischemic stroke, while mortality due to ICH and SAH increased significantly by 62.22% (95% UI, 53.51–68.64%) and 59.91% (95% UI, 50.98–66.93%), respectively (Table 1).

From 1990 to 2021, the age-standardized incidence, prevalence, mortality, DALYs, YLDs, and YLLs of stroke decreased or remained stable, with changes typically greater in men than in women (Table 2). Considerable declines were observed in overall stroke incidence, mortality, DALYs, and YLLs, whereas changes in the prevalence and YLDs were relatively minor. Notably, the prevalence exhibited varying trends across stroke subtypes during this period (Figure 1). Specifically, ischemic stroke prevalence declined by 3.44% (95% UI, −8.02 to 1.19%), whereas ICH and SAH prevalence increased by 6.34% (95% UI, 1.80–10.77%) and 6.03% (95% UI, 1.06–11.19%), respectively. Additionally, the reductions in the incidence and mortality were more pronounced for ischemic stroke than for hemorrhagic stroke (Table 2). While the incidence of ischemic stroke and ICH generally decreased from 1990 to 2021, the incidence of SAH demonstrated an increasing trend in recent years. Mortality, DALYs, and YLLs associated with hemorrhagic stroke peaked around 2000 (Figure 1 and Supplementary Figure S3).

The percentage change in age-standardized DALYs due to stroke ranged from −11.72% (95% UI, −22.31 to 0.24%) in Mississippi to −48.86% (95% UI, −54.91 to 42.54%) in the District of Columbia (Supplementary Tables S1, S2). The stroke prevalence decreased most substantially in individuals aged ≥75 years (−2.76%; 95% UI, −9.06 to 4.79%), followed by the age group of 50–74 years (−0.29%; 95% UI, −5.60 to 5.51%), whereas the prevalence increased by 2.46% (95% UI, −1.58 to 6.43%) among those aged 15–49 years. Alaska and Nevada experienced the greatest increases across all stroke measures. The number of incident cases, deaths, and DALYs decreased in most eastern US states in the youngest age group (15–49 years). Arkansas showed the highest increase in age-standardized prevalence and YLD rates among those aged 15–49 years, whereas the District of Columbia experienced the largest reductions in death and DALYs for the same age group (Figure 2). The trends in each stroke subtype according to geographic location are described in Supplementary Tables S3, S4.

Among the various risk factors, metabolic risks accounted for the largest proportion of the stroke burden, followed by behavioral and environmental/occupational risks (Supplementary Figures S4, S5). Deaths attributed to household air pollution from solid fuels showed the largest decrease, declining by 87.13% (95% UI, −99.99 to 90.15%), while high temperature demonstrated the greatest increase of 10.10% (95% UI, −146.43 to 148.37%) (Supplementary Table S6). The DALYs associated with high systolic blood pressure declined by 43.71% (95% UI, −49.17 to 39.30%) for ischemic stroke, 27.35% (95% UI, −33.89 to 21.36%) for ICH, and 29.14% (95% UI, −35.27 to 22.86%) for SAH (Table 3). The age-standardized rates and percentage changes in stroke are described in Supplementary Tables S7, S8.