Highlights
-
Stroke mortality declined substantially in British Columbia from 2002–2022, mainly due to improved survival, while incidence rose among adults aged 35–64.
-
Pre-admission stroke deaths were higher in females, especially those aged 85+.
-
Sociodemographic disparities persisted, with higher rates and fatalities in lower-income and less urban areas.
Introduction
Stroke remains a leading cause of mortality, ranking as the third leading cause of death globally in 2021 Reference Naghavi, Ong and Aali1 and the fourth in Canada in 2023. 2 Age-standardized stroke mortality rates have fallen in many countries over the past three decades, including Canada. Reference Kamal, Lindsay, Côté, Fang, Kapral and Hill3,Reference Feigin, Stark and Johnson4
Improvements in primary and secondary stroke prevention measures have been effective in reducing stroke incidence. Reference Lackland, Roccella and Deutsch5 Furthermore, stroke outcomes are improving thanks to advancements in stroke care delivery, including increased use of neuroimaging, refined risk stratification and the expanded use and access to reperfusion therapies like intravenous (IV) thrombolysis and endovascular thrombectomy. Reference Lackland, Roccella and Deutsch5 These advancements, alongside better systems of care, have likely played a key role in reducing stroke-related mortality. It is critical to analyze whether declining stroke mortality is attributable to a decline in disease incidence (indicating effective prevention) or to a decline in disease fatality (indicating improved treatment), because these drivers have distinct implications for public health policy and practice and are crucial to identify future intervention objectives.
While overall declining rates of stroke mortality reflect a major improvement in population health, there remain important disparities that will need to be addressed to deliver more equitable care. Stroke outcomes vary by geographic location in both the USA and Canada. Reference Flynn, Vaughan and Casper6–Reference Kapral, Hall and Gozdyra8 Inequalities across race/ethnicity and socioeconomic status with regard to negative stroke outcomes are persistent in high-income countries. Reference Prust, Forman and Ovbiagele9 An understanding of the true burden of stroke stratified by social determinants of disease is vital for developing targeted interventions to improve stroke outcomes for equity-deserving populations.
Using health administrative data, our study describes temporal trends in stroke event rate, case fatality and mortality in British Columbia (BC), Canada, from 2002 to 2022 and examines differences in temporal trends by sex, age, urban/rural geography and neighborhood income. We also quantify the relative contributions of stroke event rates and case fatality rates to the overall reduction in mortality, stratified by demographics and social determinants of disease.
Method
Setting, data sources and ethics
This retrospective cohort study examined temporal trends in stroke events, case fatality and stroke mortality rates across BC over a 21-year time period. BC is the westernmost province of Canada with a population of 5.5 million in 2023. BC’s publicly funded universal healthcare system generates comprehensive, population-based administrative data.
We accessed data from Population Data BC (PopData), a data platform that supports interdisciplinary research on factors influencing human health, well-being and development. The following datasets were used in this study: Discharge Abstract Database (DAD), Vital Statistics (VS), consolidation files and PharmaNet.
The DAD contains data on discharges, transfers and deaths of in-patients and day-surgery patients from acute care hospitals in BC, while VS data captures all registered deaths in the province. The consolidation files include basic demographics (age, sex), census geo-codes for residence and healthcare registration data. It primarily relies on Medical Services Plan Registration and Premium Billing files, supplemented by other sources, such as census data, to identify unregistered individuals receiving healthcare services. PharmaNet, used in the sensitivity analysis, is an online, real-time system in BC that records prescriptions dispensed at community and outpatient hospital pharmacies. Although datasets are de-identified, they can be linked using unique system identifiers across these administrative sources.
As this work falls under BC Center for Disease Control’s mandate to inform public health policy, ethics board approval was not required.
Study population and definitions
Residents of BC aged 35–110 years who experienced a stroke event between January 1, 2002, and December 31, 2022, were included in the study. Stroke events were defined using a methodology similar to prior myocardial infarction studies and included two mutually exclusive components: stroke admissions and pre-admission stroke deaths (Figure 1). Reference Smolina, Wright, Rayner and Goldacre10,Reference Camacho, Nedkoff and Wright11 A stroke admission was defined as a hospital discharge record with a primary diagnosis of stroke (International Classification of Diseases, 10th Revision (ICD-10) codes H34.1 and I60–I64 excluding I62 and I63.6) and included patients discharged alive after a hospital stay longer than one day, as well as those who died in hospital. Hospital records occurring within 30 days of a stroke admission were considered part of the same stroke event. A pre-admission stroke death was defined as a death in which stroke (ICD-10 codes H34.1 and I60–I64 excluding I62 and I63.6) was recorded as the underlying cause on the death certificate, with no hospitalization for stroke in the preceding 30 days. Individuals who died in the emergency department (ED) without being admitted may be included in this category.
Figure 1. Definition of stroke events.
The selected ICD-10 codes were informed by literature Reference Hall, Mondor, Porter, Fang and Kapral12–Reference Botly, Lindsay and Mulvagh17 demonstrating high positive predictive values for stroke identification and were aligned with those used in the BC chronic disease surveillance system. 18 In our analysis, we focused on overall stroke events and, in one supplementary analysis, categorized them as hemorrhagic (I60, I61) or ischemic (H34.1, I63 [excluding I63.6] and I64). I64 (unspecified stroke) was included in the ischemic stroke definition, consistent with the BC Chronic Disease Registry and Canadian Institutes of Health Information (CIHI) guidance. According to CIHI, I64 is most likely ischemic stroke because this code is assigned when hemorrhagic stroke has been ruled out, but the specific stroke type remains unspecified. Reference Ganesh, Stang and McAlister19 Furthermore, in the Hall et al. study, including I64 in the ischemic stroke category increased sensitivity from 66.5% to 79.6%, Reference Hall, Mondor, Porter, Fang and Kapral12 and other studies have reported similarly minimal impact on positive predictive value. Reference Woodfield and Grant20,Reference McCormick, Bhole, Lacaille and Avina-Zubieta21 We also investigated this in our stroke hospitalization data, and we found a marked improvement in diagnostic specificity over time, with I64 decreasing from 40% of stroke diagnoses in 2002 to 1% in 2022, alongside an increase in ischemic stroke (I63) from 41% to 81%. Meanwhile, the proportion of hemorrhagic strokes remained stable over time, with subarachnoid hemorrhage (I60) accounting for 5%–7% and intracerebral hemorrhage (I61) for 12%–13% of cases annually. These trends suggest that many cases previously coded as I64 were reclassified as ischemic strokes, supporting the reliability of classifying I64 as ischemic stroke.
Stroke admissions were further classified as non-fatal stroke admissions if no death occurred within 30 days of admission and fatal stroke admissions if death from any cause occurred within 30 days. Among fatal stroke admissions, deaths were further categorized as either due to stroke, when stroke was listed as the underlying cause of death, or due to another cause, when a non-stroke cause was recorded despite a recent stroke admission. The combination of pre-admission stroke deaths and fatal stroke admissions constituted fatal stroke events, which were used as the numerator for calculating case fatality, with total stroke events as the denominator. Stroke-related deaths, defined as deaths where stroke was listed as the underlying cause, included both pre-admission stroke deaths and fatal stroke admissions due to stroke, which were used as the numerator for calculating stroke mortality, with population as the denominator. All definitions are summarized in Figure 1 and Supplementary Table S1.
We excluded records with invalid dates, missing age or missing sex (0.2% of records). We applied a one-day criterion for hospital stay to avoid counting suspected stroke in patients who were discharged home without confirmation (2.6% of records).
Statistical analysis
Using the BC population registered for Medical Services Plan at the mid-point of the calendar year (July 2) as the denominator, we calculated annual event and mortality rates per 100,000 population of BC. The total rates were age-standardized in five-year age groups based on Statistics Canada’s 2011 postcensal population estimates. Reference Martel and Chagnon22 We estimated the 95% confidence intervals (CIs) assuming the observed event count followed a Poisson distribution.
We calculated the 30-day case fatality rate as the proportion of stroke events resulting in death, using all fatal events as the numerator and the total number of stroke events as the denominator. Case fatality rates were directly age standardized using 5-year age groups and the age distribution of this study’s 21-year (2002–2022) stroke event cohort as the standard population. All rates were stratified by age group and sex. We conducted a supplementary analysis stratifying stroke outcomes by type (ischemic vs. hemorrhagic), with corresponding rates presented in the supplementary materials. As a sensitivity analysis for the stroke-type assessment, we restricted the cohort to hospital admissions, where stroke coding is more accurate and I64 use is lower (40% overall vs. 80% in pre-admission stroke death) and has improved over time. This allowed us to assess whether stroke trends were consistent in a higher-quality subset.
We assessed health disparities across population subgroups defined by geographic location (rurality) and income. Using residential postal codes, study participants were linked to area-level census data. To assess geographic differences, we analyzed stroke events using a standardized 7-level urban/rural classification. Due to small case numbers, the last three categories in the original classification (rural hub, rural and remote) were combined into a single rural/remote group, resulting in five groups: (1) metropolitan, (2) large urban, (3) medium urban, (4) small urban and (5) rural/remote. Each patient’s postal code was mapped to one of 218 Community Health Service Areas (CHSAs) for 2014–2018 data or 231 for 2019–2022 data, which were then classified into one of the five urban/rural categories. The urban/rural classification was applied starting in 2014 as CHSAs in BC were defined using data from the 2016 census. Consequently, this classification is unsuitable for data prior to 2014.
For the income analysis, we used neighborhood income quintiles (NIQs) generated by PopData. NIQs provide a stable indicator of socioeconomic status, less affected by annual fluctuations than individual-level income. 23 These quintiles were created using census data to rank dissemination areas by average pre-tax income within census metropolitan areas, dividing the population into quintiles to represent community-specific income categories. Stroke event, case fatality and mortality rates were analyzed by income level and degree of urbanization using five-year average rates (2018–2022) in order to simplify interpretability and visualization. We also conducted a sensitivity analysis by excluding the COVID-19 years (2020 and 2021) to assess whether their inclusion significantly affected the results.
A further sensitivity analysis was conducted using individual-level income categories from the BC Ministry of Health’s Fair PharmaCare and Income Assistance Plan C, accessed through PharmaNet. Fair PharmaCare categories are divided into income quintiles based on net family income from two years prior. Plan C recipients, who have an annual income of less than 30,000 CAD and receive full coverage for eligible prescription costs, were classified in the lowest-income quintile. Linkage to PharmaNet data revealed that 57% of the study cohort was enrolled in Fair PharmaCare or Plan C, while 37% were assigned to other non-income-based PharmaCare plans, primarily condition- or status-based (e.g., plans for HIV patients, psychiatric medications, cystic fibrosis patients or those residing in long-term care). The remaining 6% were unmatched due to the absence of any dispensation within two years of their stroke event. Consequently, only 57% of the cohort was included in the individual-level income analysis. Given the limitations of this analysis and challenges in estimating population denominators by income for event and mortality rates, case fatality rates by individual income are presented in the supplementary data as a sensitivity analysis.
We estimated the average annual percentage changes in event and mortality rates using Poisson regression. In the overall analysis, the natural logarithm of the annual rate was used as the dependent variable, while five-year age groups, calendar year and NIQ served as independent variables. NIQ was excluded from models when its removal improved model fit. Furthermore, negative binomial models were used if overdispersion was present. For case fatality, we used a generalized linear model with a binomial distribution and a log link function. The average annual percentage change was calculated from the β coefficient for the calendar year using the formula 100×[exp(β)-1]. All models were stratified by sex, which is what is typically recorded in administrative datasets. Information on gender was not available. In a sensitivity analysis, we stratified the analysis into two time periods, 2002–2010 and 2011–2022, based on observed shifts in case fatality. Coincidentally, this period marks the gradual uptake of organized stroke care, including increased thrombolysis use following the European Cooperative Acute Stroke Study (ECASS) III, expanded stroke unit implementation and early adoption of endovascular thrombectomy. Results for each period are presented in the supplementary materials.
The relative contributions of event and case fatality rates to the decline in stroke mortality were calculated using methods from previous studies. Reference Smolina, Wright, Rayner and Goldacre10,Reference Camacho, Nedkoff and Wright11 In brief, the average annual change in mortality rate equals the sum of the average annual changes in event and case fatality rates, expressed as percentages. This is based on the formulas: mortality rate (M) = event rate (E) × case fatality rate (C) in each year, and M’/M = E’/E + C’/C, where M’, E’ and C’ represent the average annual change for each rate in absolute terms. Expressing the average annual change in percentage terms gives ΔM = ΔE + ΔC. Next, the change in mortality rate was divided by the change in event rates and case fatality, respectively, to calculate the relative contributions of each. The sum of the relative contributions of case fatality and event rates equals 100%. Data reading and cleaning were performed in SAS version 9.4, while statistical analysis was conducted in R version 4.3.1.
Results
Stroke events and demographic characteristics
Between January 1, 2002, and December 31, 2022, there were 123,075 stroke events in BC, of which 83% were ischemic strokes compared to hemorrhagic strokes. Table 1 summarizes the characteristics of these events by sex. The mean age of the cohort was 77 years (SD = 13), with females being older on average (79, SD = 13) compared to males (74, SD = 13). Twice as many females were aged 85 years or older compared to males (42% vs. 24%). Sex differences in stroke events were observed in stroke-related admissions and deaths, with females being more likely than males to experience a pre-admission death.
Table 1. Descriptive characteristics of stroke events in British Columbia by sex, 2002–2022
€ The total stroke events are not simply the sum of stroke admissions and fatal stroke, as some individuals may have experienced both within the dataset. Instead, stroke events are defined as the sum of stroke admissions and pre-admission stroke deaths.
B Proportions are calculated as percentages of total stroke events.
αFatal stroke admissions and pre-admission stroke deaths are included under both stroke events and fatal stroke events to indicate their proportion of total stroke events as well as of fatal stroke events.
¥ Fatal strokes include stroke-related hospital admissions resulting in death within 30 days, as well as pre-admission stroke deaths without a stroke hospital admission in the preceding 30 days.
∑ Proportions are calculated as percentages of total fatal stroke events.
§ Stroke-related deaths are a subset of fatal stroke events in which stroke was identified as the primary cause of death on the death certificate. Therefore, this includes both pre-admission stroke deaths and deaths due to stroke following hospital admission for stroke.
µIschemic stroke: International Classification of Diseases, 10th Revision (ICD-10) codes I63 (excluding I63.6) and I64; hemorrhagic stroke: ICD-10 codes I60 and I61.
Total stroke events in the sample were comprised of 79% stroke admissions and 21% pre-admission deaths due to stroke. In females, these proportions were 75% admissions and 25% pre-admission deaths, whereas in males, they were 83% admissions and 17% pre-admission deaths (Table 1). The sex difference in pre-admission deaths was most pronounced in those aged 85 years and older: 40% of stroke events in females occurred as pre-admission deaths, compared with 33% in males (Table S2). Between 2002 and 2022, these proportions among those 85 years and older fluctuated between 36%–46% in females and 29%–37% in males (Figure S1).
In total, 43,593 fatal stroke events were identified in the cohort, among which 60% were pre-admission stroke deaths without stroke admission in the prior 30 days, 25% were deaths due to stroke after admission, and 15% were deaths due to other reasons after a stroke admission (Table 1). Fatal strokes made up 40% of all stroke events in females compared to 31% in males, driven by a higher relative proportion of pre-admission stroke deaths (versus post-admission deaths) in females (62% vs. 56% of fatal strokes) (Table 1). This difference again was most pronounced in those aged 85 and older, where pre-admission stroke deaths comprised 70% of fatal strokes in females and 64% in males (Table S2). Between 2002 and 2022, the proportion of pre-admission deaths among fatal strokes in those aged 85 and older ranged from 65% to 74% in females and from 60% to 69% in males (Figure S2).
Almost half of all stroke events occurred among individuals in the two lowest-income quintiles, and 42% resided in Metropolitan areas (Table 1).
Trends in stroke event rates
Overall trends in event rates
From 2002 to 2022, the age-standardized stroke event rate declined for both sexes, with consistently higher rates observed in males throughout the study period (Figure 2A). The difference between sexes was primarily attributed to higher stroke admission rates in males rather than pre-admission stroke death (Figure S3 A). Stroke event rate declined by 33% in females (from 207.7 to 139.9 per 100,000 population) and by 25% in males (from 247.6 to 186.5 per 100,000 population) (Figure 2A, Table S3). Event rates for both ischemic and hemorrhagic strokes declined over time, with a greater reduction seen in ischemic strokes, particularly among females (Figure S4 A). Ischemic stroke rates dropped by 36% in females (from 171.2 to 110.0 per 100,000 population) and 25% in males (from 206.7 to 154.9). Hemorrhagic stroke rates declined by 16% in females (36.4–30.0) and 23% in males (40.9–31.6). These results were consistent in the sensitivity analysis restricted to hospital-admitted stroke cases.
Figure 2. Age-standardized stroke event (A) 30-day case fatality (B) and mortality (C) rates among individuals aged 35+ years, British Columbia, 2002–2022.
Temporal trends in stroke event rates varied between age groups: older adults (65+ years) experienced consistent declines, whereas rates increased among younger adults (35–64 years) (Figure 3A, Table S3). The sharpest declines between 2002 and 2022 were in females aged 75–84 years (–45%) and males aged 75+ years (−38%), while the largest increases occurred in those aged 35–54 years, rising by 14% in females and 27% in males (Figure 3A, Table S3). Across all age groups, stroke event rates were consistently higher in males, except in the 85+ group, where rates were higher in females or comparable between sexes (Figure 3A).
Figure 3. Three-year centered moving average of stroke event rate (A), 30-day case fatality rate (B) and mortality rate (C) by age group among individuals aged 35+ years, British Columbia, 2002–2022. A centered moving average was used, with the three-year average spanning from 2003 (first center year) to 2021 (last).
Annual change in event rates
Over the 21-year period, the annual average event rate reduction was greater in females (−1.8% per year, 95% CI: −2.0% to −1.6%) than in males (−0.9% per year, 95% CI: −1.1% to −0.7%), with a more pronounced decline among adults aged 75 years and older (Table S3, Figure 4A). The reduction was more significant for pre-admission stroke death events (females: −3.6% [−3.8% to −3.3%]; males: −2.8% [−3.1% to −2.4%]) compared to hospital admissions (females: −1.3% [−1.5% to −1.1%]; males: −0.6% [−0.8% to −0.4%], data not shown). The decline was steeper from 2002 to 2010 (females: −2.8% [−3.4% to −2.2%]; males: −2.2% [−2.8% to −1.6%]) compared to the period from 2011 to 2022 (females: −1.4% [−1.7% to −1.0%]; males: −0.4% [−0.8% to 0.0%]), with variations observed across age (Figure S5 A).
Figure 4. Average annual changes in event rates and 30-day case fatality rates among individuals aged 35+ years in British Columbia. Note that overall changes in stroke mortality rates equal the sum of changes in event rates and case fatality rates. Data are presented by age group (A) and neighborhood income quintile (B) for the period 2002–2022 and by urban/rural classification (C) for the period 2014–2022. Neighborhood income is categorized into quintiles, with the 1st quintile representing the lowest income and the 5th quintile the highest.
Trends in 30-day case fatality rates
Overall trends in case fatality rates
From 2002 to 2022, age-standardized case fatality rates declined in both sexes, with consistently slightly higher rates in females (Figure 2B). This sex difference was primarily driven by higher fatality rates from pre-admission stroke deaths in females (Figure S3 B). Case fatality declined by 22% in females (from 39.5 to 30.8 per 100 stroke events) and by 15% in males (from 37.4 to 31.9 per 100 stroke events) (Figure 2B, Table S3). Case fatality rates were higher for hemorrhagic compared to ischemic stroke and declined over time for both stroke types, with a greater reduction in females than in males (Figure S4 B). Ischemic stroke rates decreased by 20% in females (from 36.2 to 29.0 per 100 stroke events) and by 13% in males (from 34.1 to 29.5). Hemorrhagic stroke rates also declined by 21% in females (from 51.1 to 40.3) and by 13% in males (from 52.5 to 45.9). These patterns remained consistent in the sensitivity analysis limited to hospital-admitted stroke cases.
The greatest decline in case fatality occurred in young adults (35–54 years), with decreases of 53% in females and 54% in males (Figure 3B, Table S3). The smallest changes were observed in those aged 85+ years (−13% in females, −2% in males).
Annual change in case fatality rates
Between 2002 and 2022, the annual case fatality reduction was slightly greater in females (−2.7% per year, 95% CI: −3.0% to −2.4%) than in males (−2.5% per year, 95% CI: −2.8% to −2.3%), with a more pronounced decline among younger adults (ages 35–54 years) (Table S3, Figure 4A). Declines in case fatality rates were driven by both reductions in hospital admissions (females: −2.1% [−2.5% to −1.8%]; males: −2.7% [−3.2% to −2.4%]), as well as decreases in pre-admission stroke deaths (females: −1.9% [−2.2% to −1.6%]; males: −1.3% [−1.6% to −0.9%]) (data not shown). In contrast to stroke event rates, case fatality rates declined more during 2011–2022 (females: −2.4% [−3.1% to −1.8%]; males: −1.6% [−2.3% to −0.9%]) compared to 2002–2010 (females: −0.6% [−1.7% to 0.4%]; males: −1.3% [−2.4% to −0.2%]), with variations observed across age (Figure S5 A).
Trends in mortality rates
Overall trends in mortality rates
Between 2002 and 2022, the age-standardized mortality rate declined in both sexes, with consistently higher rates in males (Figure 2C). Mortality declined by 53% in females (from 71.6 per 100,000 population to 33.6) and by 43% in males (from 71.9 to 40.8) (Figure 2C, Table S3). Mortality rates for both ischemic and hemorrhagic strokes declined over time, with a greater reduction seen in ischemic strokes (Figure S4 C). Ischemic stroke rates dropped by 56% in females (from 57.8 to 25.7 per 100,000 population) and 44% in males (from 56.6 to 31.9). Hemorrhagic stroke rates declined by 43% in females (13.8 to 7.9) and 42% in males (15.3 to 8.8). These findings remained consistent in the sensitivity analysis restricted to hospital-admitted stroke cases.
The smallest relative decline was observed in the 55–64 age group, while females aged 85 and older had consistently higher mortality rates than males of the same age group (Figure 3C, Table S3).
Annual change in mortality rates
The annual mortality rate reduction was slightly greater in females (−3.8% per year, 95% CI: −4.1% to −3.6%) compared to males (−3.2% per year, 95% CI: −3.5% to −2.9%). Reductions were more substantial among older adults (75+ years), whereas young adults (35–64 years) showed smaller declines (Table S3).
Determinants of mortality trend
From 2002 to 2022, reductions in case fatality rates contributed relatively more to mortality declines than reductions in event rates. This effect was stronger in males (74% of mortality decline) than in females (60%) (Figure 4A, Table S3). Among those under 65 years, mortality reductions were entirely due to case fatality declines, which offset rising event rates. Before 2010, mortality reductions were mainly driven by lower event rates (82% in females, 63% in males). After 2010, declines in case fatality became the primary driver of mortality reduction (accounting for 64% in females and 80% in males), with variations across age groups (Figure S5 A).
Sociodemographic disparities in stroke outcomes
From 2018 to 2022, the 5-year average age-standardized stroke event rates were highest among individuals living in lower-income neighborhoods, ranging from 178.4 per 100,000 in females and 232.5 in males in the lowest-income quintile to 133.8 in females and 152.0 in males in the highest quintile (Figure 5A). Rates were also higher in medium to small urban areas relative to larger urban areas or remote/rural areas (Figure 5A). Across all income and urbanicity groups, males consistently had higher stroke event rates than females. These patterns remained unchanged when excluding the COVID-19 years (2020–2021; data not shown).
Figure 5. Five-year average age-standardized stroke event (A), 30-day case fatality (B) and mortality rates (C) by neighborhood income quintile and urban/rural classifications, stratified by sex among individuals aged 35+ years, British Columbia, 2018–2022. Neighborhood income is categorized into quintiles, with the 1st quintile representing the lowest income and the 5th quintile the highest.
Over the full study period, declines in event rates were more pronounced in higher-income neighborhoods. Among females, annual declines were 2.7% (95% CI: −3.1% to −2.4%) in the highest income quintile versus 1.2% (95% CI: −1.5% to −0.2%) in the lowest; for males, the corresponding declines were 1.5% (95% CI: −1.8% to −1.1%) and 0.7% (95% CI: −1.0% to −0.4%) (Figure 4B; Table S4).
Between 2014 and 2022, the most significant reductions in event rate occurred in metropolitan areas: −2.2% (95% CI: −2.9% to −1.3%) per year among females and −1.2% (95% CI: −1.8% to −0.5%) among males. Declines in other urban and less urban areas were smaller or not statistically significant (Figure 4C; Table S5).
The five-year average case fatality rates from 2018 to 2022 were slightly higher in lower-income areas, with a more pronounced gradient among males (Figure 5B). Sensitivity analysis using individual-level income showed a steeper gradient for both sexes (Figure S6). Case fatality was significantly lower in metropolitan areas, and excluding pandemic years did not affect results (Figure 5B).
Over the full study period, declines in case fatality were also greater in higher-income groups (−3.3% annually in both sexes vs. −2.0% in the lowest-income group) (Figure 4B; Table S4). Between 2014 and 2022, females experienced significant reductions in large (−2.8%, 95% CI: −5.1 to −0.5) and medium urban areas (−4.2%, 95% CI: −6.6 to −1.7). Among males, reductions were only significant in medium urban areas. Notably, case fatality increased among males in small urban and rural/remote areas, though not significantly (Figure 4C; Table S5).
From 2018 to 2022, the 5-year average age-standardized stroke mortality rate was highest among individuals living in lower-income neighborhoods, 45.4 per 100,000 in females and 53.8 in males, and in medium to small urban areas (Figure 5C). In medium urban areas, mortality rates were 43.5 in females and 49.8 in males, while in small urban areas, rates were 49.9 in females and 49.5 in males (Figure 5C). Annual declines in mortality mirrored patterns observed in case fatality (Tables S4 and S5).
In terms of mortality reductions, declines in case fatality contributed more than reductions in event rates across all income levels (Figure 4B; Table S4). Notably, after 2010, case fatality became the primary driver of declining mortality, whereas before 2010, reductions in event rates played a more dominant role (Figure S5 B). Patterns by urbanicity also varied, with case fatality contributing more substantially to mortality reductions among females across all area types (Figure 4C; Table S5).
Discussion
This population-based study in BC shows a significant decline in stroke mortality from 2002 to 2022. Stroke event rates and case fatality rates both declined during the same period. While the study did not directly assess the impact of treatments, the increasing contribution of reduced case fatality to mortality decline reflects improved survival over time, which, in turn, could be a result of changes in acute stroke care, treatment and/or changes in diagnostic practices, among other factors. However, these trends were not observed equally across the population, with an overall increase in stroke event rates and smaller reductions in mortality rates among younger adults, as well as a higher burden in those living in lower-income and non-metropolitan areas.
We found an increasing temporal trend in stroke events among young adults. Similar patterns have been observed for stroke mortality in the USA, Reference Ariss, Minhas and Lang24–Reference Hall, Vaughan, Ritchey, Schieb and Casper26 Japan Reference Ohya, Matsuo and Sato27 England Reference Li, Scott and Rothwell28 and other high-income countries, Reference Feigin, Stark and Johnson4,Reference Scott, Li and Rothwell29 as well as for stroke-related hospitalizations in Canada between 2007 and 2017, Reference Botly, Lindsay and Mulvagh17 and for both hospitalizations and ED visits in Ontario between 2003 and 2017. Reference Joundi, Smith, Yu, Rashid, Fang and Kapral16 Several factors have been suggested to explain the rising incidence of stroke among younger adults. Studies from Japan Reference Ohya, Matsuo and Sato27 and Spain Reference Soto-Cámara, González-Bernal, González-Santos, Aguilar-Parra, Trigueros and López-Liria30 report that younger stroke patients are more often affected by lifestyle-related risk factors such as smoking and alcohol use, as well as non-atherosclerotic causes, while older adults more commonly present with atrial fibrillation and atherosclerosis. In the USA, increasing stroke incidence in younger adults has been linked to a concurrent increased prevalence of stroke risk factors such as hypertension and diabetes. Reference George, Tong and Bowman31 The rising incidence of stroke among young adults in England has been suggested to reflect the undertreatment of vascular risk factors, the underestimation of risk by current prediction tools and the influence of occupational and psychosocial stressors, rather than the impact of healthcare access or improved diagnostics. Reference Li, Scott and Rothwell28 Results from the Canadian Community Health Survey for BC suggest that trends in smoking, alcohol consumption, physical activity and fruit and vegetable intake among younger adults follow a similar pattern to those in older age groups. 32 Thus, other emerging contributors may be involved, including changes in diet, Reference Lane, Gamage and Du33 increased intake of ultra-processed food, Reference Lane, Gamage and Du33,Reference Nardocci, Polsky and Moubarac34 insulin resistance Reference Li, Chi, Wang, Setrerrahmane, Xie and Xu35 and altered sleep patterns. Reference Zheng, Annis and Master36,Reference Mc Carthy, Yusuf and Judge37 Furthermore, factors more specific to younger adults, such as migraines, oral contraceptive use, pregnancy and the postpartum period, patent foramen ovale and recreational drug use, may also play a role. Reference Bukhari, Yaghi and Bashir38
We found that declines in case fatality rates were mainly driven by decreases in deaths during/following hospital admissions, with no significant differences observed between females and males. The steepest decline in case fatality among hospitalized stroke patients occurred around 2007 and 2011, coinciding with significant system changes that facilitated greater uptake of acute reperfusion therapies around 2008 across North America. These included the widespread adoption of IV thrombolysis in emergency rooms following the publication of the National Institute of Neurological Disorders and Stroke in 1995 39 and the ECASS III trial in 2008, Reference Hacke, Kaste and Bluhmki40 trends that were reflected globally. Reference Olavarría Vónica, Hoffmeister, Vidal, Brunser, Hoppe and Lavados41,Reference Scherf, Limburg, Wimmers, Middelkoop and Lingsma42 After 2015, additional system reorganizations improved access to endovascular therapy for mechanical thrombectomy, spurred by a series of landmark trials demonstrating substantial benefits for patients with large vessel occlusions in the anterior circulation. Reference Goyal, Menon and van Zwam43 The advancement in treatment since about 2008 may explain why our stratified analysis by year showed a greater relative contribution of case fatality to mortality reduction compared to event rate after 2010. However, it is also important to consider that increased use of MRI over this period may have led to the detection of more minor strokes, which could also contribute to the observed decline in case fatality by including milder cases with better prognosis. Reference Whiteley, MacRaild and Wang44,Reference Tedyanto, Tini and Pramana45 The smaller decline in case fatality among pre-admission stroke deaths likely reflects ongoing accessibility challenges, consistent with the higher fatality observed in less urban and remote areas, as well as the older age profile of those who die before hospital admission.
Consistent with previous studies, Reference Joundi, Smith, Yu, Rashid, Fang and Kapral16 event rates for both ischemic and hemorrhagic strokes declined over time. Ischemic strokes, being more common, contributed more to the overall burden in event and mortality rates, whereas hemorrhagic strokes were associated with higher fatality, a pattern consistent with our findings and previous studies. Reference Feigin, Lawes, Bennett, Barker-Collo and Parag46,Reference Grysiewicz, Thomas and Pandey47 Case fatality declined similarly for both stroke types, with approximately 20%–21% reduction in females and 13% in males. This is likely from a combination of advances in reperfusion therapy for ischemic stroke and improvements in acute intracerebral hemorrhage management, such as better blood pressure control, reversal of blood thinners and enhanced neurocritical care. Reference Hemphill, Greenberg and Anderson48,Reference Qureshi, Mendelow and Hanley49 System-wide changes like stroke unit implementation and standardized protocols likely benefited both types, with hemorrhagic stroke patients gaining more from advances in intensive care. 50 Although stroke severity score data were unavailable in our study, it is a key determinant of mortality outcomes and likely influences the differences observed in case fatality trends between stroke types. Reference Qureshi, Mendelow and Hanley49,Reference Donnan, Fisher, Macleod and Davis51 Future studies incorporating clinical severity measures are needed to better understand these relationships and further explain stroke fatality trends.
Although males had higher age-specific stroke event and mortality rates similar to findings in the literature, Reference Ariss, Minhas and Lang24 females exhibited slightly higher case fatality rates Reference Appelros and Åsberg52 and had higher event and mortality rates in the 85+ age group. This may be due, at least in part, to their greater longevity and higher burden of stroke risk factors with age, such as atrial fibrillation, diabetes, dyslipidemia, hypertension and obesity. Reference Silva, Lima and Camargo53 Additionally, many female stroke survivors live alone and rely on external care, which can lead to delays in care-seeking, reduced access to timely treatment, lower quality of life and increased risk of fatal outcomes. Reference Appelros and Åsberg52,Reference Bushnell, Howard and Lisabeth54,Reference Reeves, Bushnell and Howard55 There may also be opportunities for bias in EDs, where females could be under-admitted or their symptoms overlooked, leading to missed diagnoses. Reference Foerch, Misselwitz, Humpich, Steinmetz, Neumann-Haefelin and Sitzer56 Our disaggregated analysis confirms that the higher case fatality in females was largely driven by pre-admission stroke deaths, particularly in the oldest age group. Consistent with this, US national data have also reported a high proportion of pre-transport stroke deaths, especially among women. 57 These findings highlight the need for targeted strategies to improve stroke recognition, diagnosis and emergency care for older women, who may be especially vulnerable to delays in acute management. Furthermore, differences in stroke presentation, treatment eligibility and access to reperfusion therapies such as thrombolysis and endovascular thrombectomy may also contribute to these disparities. Reference Reeves, Bushnell and Howard55,Reference Haast, Gustafson and Kiliaan58 Disentangling biological (sex-based) from social (gender-based) determinants is essential to inform equitable stroke care. Reference Appelros and Åsberg52 Future research should identify actionable causes underlying the persistent gap in stroke fatality between sexes.
Our study identified sex- and age-specific differences in stroke mortality drivers, with declining case fatality rates contributing more than event rates to mortality reduction. However, the influence of case fatality diminished with age. While older populations benefited from both reduced event rates and lower case fatality, younger individuals experienced a less pronounced decline in mortality, largely due to the rising incidence of stroke. This trend underscores the urgent need for enhanced preventive measures tailored to younger demographics. Additionally, evidence shows that half of all subsequent stroke events occur after the first year following a transient ischemic attack or minor stroke, Reference Khan and Yogendrakumar59 highlighting the importance of long-term monitoring and secondary prevention strategies, particularly crucial for younger individuals who may be under-monitored in routine clinical follow-up.
Socioeconomic status is an important determinant of health, with higher mortality rates in lower-income countries and lower-income neighborhoods. Reference Taghdiri, Vyas and Kapral60,Reference Addo, Ayerbe and Mohan61 Prior studies assessing socioeconomic impact on stroke outcomes have found that under-resourced neighborhoods are less likely to receive acute reperfusion therapies. Reference Taghdiri, Vyas and Kapral60 Our study found that in BC, stroke mortality varied by neighborhood and personal income level, with consistently lower rates observed in higher-income quintiles across both sexes. Reductions in both case fatality and event rates were more pronounced among individuals in higher-income neighborhoods. These improvements may reflect better access to healthcare, enhanced monitoring and more effective acute stroke management in wealthier areas. We also found that over a five-year period, stroke outcomes for both sexes were consistently better in metropolitan and large urban areas, aligning with findings from Ontario. Reference Kapral, Austin and Jeyakumar62 This disparity may be attributed to differences in risk and protective factors and healthcare access/use. However, stroke mortality rates for males only declined significantly in medium urban areas, and other geographic regions did not show the same beneficial trend. These results suggest that despite BC’s universal healthcare system, disparities in stroke outcomes exist based on place of residence.
Our study has several strengths. By using provincial administrative data, it includes the entire BC population, including often underrepresented rural residents, providing a comprehensive 21-year population-based analysis of stroke events, case fatality and mortality. Additionally, combining hospital records with death records captures pre-admission stroke deaths that occur outside of hospitals, thus enabling a more complete characterization of the burden of acute and significant stroke events. Our study also has limitations. Some excluded stroke cases, such as those with hospital stays of one day or less or a secondary stroke diagnosis, may represent true stroke events. Our analysis did not include ED data, as these are not available at the population level. As a result, some milder stroke events may not be captured, and some cases classified as pre-admission stroke deaths may have had an ED encounter. Nonetheless, our focus on hospitalized strokes is supported by validation studies showing high specificity and sensitivity of hospital discharge data for stroke. Reference Kirkman, Mahattanakul, Gregson and Mendelow63–Reference Liu, Reeder, Shuaib and Mazagri65 While this approach may still underestimate overall stroke incidence, it aligns with Canadian stroke surveillance guidelines recommending hospital data as a primary source. 66 Moreover, US national data show that only ∼3% of stroke deaths occur in the ED, while nearly half occur before hospital transport, particularly among adults aged 85+. This supports our interpretation that most pre-admission stroke deaths in our study represent fatalities before transport rather than unrecorded ED events. 57 Many stroke deaths were coded as unspecified stroke (I64) in the VS data, limiting subtype stratification. However, literature supports including I64 as ischemic stroke, and our hospital-based sensitivity analysis showed similar trends, supporting the robustness of our findings. Additionally, the lack of ethnicity information in the data source limits the ability to explore inequities across ethnic groups. Specifically, this analysis does not stratify by Indigeneity, and as such, the results are not reflective of the situation for First Nations, Métis and Inuit Peoples and communities.
Conclusion
From 2002 to 2022 in BC, Canada, overall stroke mortality rates declined by approximately 53% in females and 43% in males. This decline was seen across both ischemic and hemorrhagic stroke subtypes and was driven by reduced event rates and, more notably, improved hospital-based case fatality. Our findings highlight the impact of advancements in treatment and secondary prevention on reducing mortality in recent years and underscore the ongoing need to strengthen upstream prevention strategies. There is a need for focused interventions to address ongoing inequities in the disease burden of stroke among younger adults, older females at greater risk of pre-admission stroke death and those living in lower-income and/or more rural areas. Further research is needed to understand the mechanisms driving the observed differences and develop targeted strategies to promote equitable stroke care and improve outcomes.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/cjn.2025.10483.
Acknowledgments
We would like to thank Sacha Arsenault, provincial director at Stroke Services BC (SSBC), for her valuable input in interpreting the study findings from a clinical stroke care perspective and for helping coordinate collaboration with SSBC.
Author contributions
SS, KS and JF contributed to conceptualization and methodology. SS led the programming and data analysis, with additional programming support from JF. LZ and AD contributed to the clinical aspects of study design and clinical interpretation. SS and MH drafted the initial manuscript. KS, JF, LZ, GF and AD reviewed, revised and edited the manuscript. All authors contributed to the interpretation of results and critically reviewed and approved the final manuscript.
Funding statement
This study was conducted under the BC Centre for Disease Control’s mandate to monitor population health and its determinants and inform public health policy. No financial support was received.
Competing interests
Solmaz Setayeshgar and Kate Smolina are employed by the Provincial Health Services Authority. Gillian Frosst is employed by Interior Health and seconded to the Provincial Health Services Authority. Lilly W. Zhou is a recipient of the VCHRI Mentored Clinician Scientist Award. All other authors declare no competing interests.
Disclaimer
Access to data provided by the Data Stewards is subject to approval but can be requested for research projects through the Data Stewards or their designated service providers. The following datasets were used in this study: DAD, VS, consolidation files and PharmaNet. You can find further information regarding these datasets by visiting the PopData project webpage at my.popdata.bc.ca/project_listings/16-218. All inferences, opinions and conclusions drawn in this publication are those of the author(s) and do not reflect the opinions or policies of the Data Steward(s).
Open access
