Twenty Years of Progress Toward Understanding the Stroke Belt

See related articles, p 708, 711, 719, 729, 736

The Stroke Belt is a region of higher stroke mortality in the Southeastern United States that has persisted since at least 1940. Two decades ago (1999), potential contributors to this disparity were reviewed. This report serves to describe the subsequent progress in understanding the magnitude and contributors to the Stroke Belt.The Stroke Belt is defined based on higher stroke mortality from all strokes in the Southeastern United States compared with other regions. In 1968, the age-adjusted stroke mortality rate in the Stroke Belt was 582/100 000 compared with 433/100 000 for the rest of the United States. National declines in stroke mortality reduced these rates, and by 2016, the stroke mortality rates were 126/100 000 and 99/100 000, respectively. These changes reduced the absolute magnitude of the Stroke Belt disparity from 149/100 000 to 26/100 000 but the relative excess by only 34% to 27%. Higher mortality can be the product of higher incidence, higher case-fatality, or both. Higher stroke mortality in the Stroke Belt appears to be primarily related to a higher stroke incidence, with higher stroke case-fatality playing a smaller role. The Stroke Belt also includes a higher proportion of rural residents relative to other regions, and nationally, residents of rural regions have higher stroke incidence compared with residents living in more urban areas. The data suggest that contributors to the Stroke Belt may include larger proportion of blacks and residents with higher prevalence of traditional stroke risk factors, higher prevalence of inflammation and infection, and lower socioeconomic status. Environmental exposures and lifestyle choices perhaps play a lesser role. While substantial progress has been made to advance the understanding of the contributors to the Stroke Belt, much work remains to better understand this disparity.

Definition, Scale, and Persistence of the Stroke Belt

The higher stroke mortality rates in the Southeastern United States were first noted in 1965 summarizing stroke mortality back to 1949 to 1951, concluding the highest rates are found in the South Central and South Atlantic States and the lowest rates in the Southwestern and Mountain States.¹ By the mid 1970s, this region was referred to as the Stroke Belt,² although the source of this term is unknown.³ The Centers for Disease Control reports stroke mortality at the county level (Figure 1), where the higher stroke mortality rates in the Southeastern United States are shown to be 2 to 4× higher than in other regions.⁴ Because the majority of reports have described differences in any stroke, this report will not be able to describe geographic differences in infarction and hemorrhage.

• Download figure
• Download PowerPoint

The Stroke Belt is most commonly defined at the state level; however, the specific states included are ill-defined. In the 1980s, the National Heart Lung and Blood Institute defined the Stroke Belt as states with age-adjusted stroke mortality at least 10% above the national rate, which in 1980 included Alabama, Arkansas, Georgia, Indiana, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia.⁵ This approach has the advantage of being objective, but the disadvantage that as the deaths from stroke vary from year-to-year, the states included will also vary. For example, applying this definition during 1999 to 2016, a total of 24 states met the 10% threshold in at least 1 year, with many states sporadically included, for example, Delaware in 2016; Idaho in 2001, 2003, 2005, and 2006; Missouri between 2006 and 2014, and Washington between 1999 and 2002. Other investigators have employed alternative definitions, such as selecting a group of southern states that include many counties with high stroke mortality. Investigators in the national cohort study, REGARDS (Reasons for Geographic and Racial Differences in Stroke), include the states of North Carolina, South Carolina, Georgia, Tennessee, Alabama, Mississippi and Arkansas, and Louisiana.⁶ With the exception of Louisiana for a single year (2000), all 8 of these states met the 10% above the national average threshold for every year between 1999 and 2016. Finally, hot spot counties with high stroke morality have been identified using small area spatial statistics, with some of these counties falling outside of these states traditionally defined as the Stroke Belt.⁷ Unless otherwise stated, we will define this 8-state region as the Stroke Belt.

Between 1968 and 2016, age-adjusted stroke mortality for ages 45+ years has declined a remarkable 77% from 455.5/100 000 to 104.1/100 000.⁸ The geographic disparity, however, has persisted (Figure 2A), with the age-adjusted stroke mortality ratio for the Stroke Belt relative to rest of nation decreasing slightly from above 1.3 to slightly below (Figure 2B). Specifically, in 1968, the stroke mortality rate in the Stroke Belt was 582/100 000 compared with 433/100 000 for the rest of the United States, while in 2016, stroke mortality rates were 126/100 000 and 99/100 000, respectively. These changes reduced the absolute magnitude of the Stroke Belt from 149/100 000 to 27/100 000; however, the decline in the relative excess was only from a 34% to 27%. Recently, the national decline in stroke has plateaued or reversed, with the Southeastern Census Region showing a 4.2% increase in stroke deaths between 2006 and 2013 (with no change in other Census Regions), with 14 of 17 states in the region showing a stall or reversal in stroke deaths.⁹ Hence, the relative increased stroke mortality in the Stroke Belt persists despite the national decline in stroke mortality.

• Download figure
• Download PowerPoint

Incidence Versus Case-Fatality, and Confounding With Rurality

There are 2 ways for a geographic region to have a higher stroke mortality rate. First is a larger number of residents having a stroke, that is, higher stroke incidence. Second, stroke patients from the region are more likely to die from the event, that is, higher case-fatality. If incidence is the driving force, then community-based efforts to reduce stroke events would be a promising pathway to reduce the disparity. Alternatively, if case-fatality is the driving force, then hospital-based efforts to improve stroke outcomes would be a promising pathway to reduce the disparity. A potential higher case-fatality could be a product of less access to care in the Stroke Belt, where there are 13% fewer physicians per capita than in the rest of the country (357/100 000 versus 408/100 000).¹⁰ This could be related to a lower likelihood of patients with stroke in the Stroke Belt being treated at a Primary Stroke Center, where 44% of the population in the Stroke Belt lives within 60 minutes of a Primary Stroke Center compared with 69% of non–Stroke Belt residents,¹¹ and where in REGARDS, 14.7% of 546 strokes occurring in the Stroke Belt were seen at a Primary Stroke Center, compared with 27.3% of 454 strokes occurring outside of the Stroke Belt.¹² This deficit potentially affects both preventative measures, as well as acute and chronic stroke care. There have been increases in the use of both thrombolytic and endovascular therapies; however, thrombolysis is not provided to over 90% of ischemic patients with stroke¹³ and 98% fail to receive endovascular treatment.¹⁴ While both of these therapies improve a patient’s functional outcome, they have either a null impact or only slight increases on case-fatality.^14,15 As such, it seems unlikely that regional differences in these acute stroke therapies are substantial contributors to the Stroke Belt.

In an analysis of whether incidence or case-fatality is the primary contributor for counties with high stroke mortality (including the Stroke Belt counties), counties were categorized into quartiles by race using smoothed age-adjusted stroke mortality data from Vital Statistics 2005 to 2010 (similar to Figure 1). Stroke incidence rate (per 100 000) among participants in REGARDS was 658 (95% CI, 588–735) for counties in the lowest stroke mortality quartile and increased monotonically to 844 (95% CI, 761–937) in the fourth, with a P value of 0.0023 for increasing trend and a hazard ratio (HR) of 1.29 (95% CI, 1.10–1.52) comparing the fourth to first quartiles.¹⁶ The case-fatality rate was 16.5 (95% CI, 12.3–21.4) in counties in the lowest stroke mortality quartile, 23.9 (95% CI, 19.0–29.3) in the second, 18.2 (95% CI, 13.8–23.2) in the third, and 23.8 (5% CI, 19.3–28.9) in the fourth, with P value of 0.058 for an increasing trend and an odds ratio (OR) of 1.71 (95% CI, 1.13–2.59) comparing the fourth to first quartiles. These data suggest an increasing stroke incidence in areas of higher stroke mortality but a less consistent pattern for increasing case-fatality in areas of higher stroke mortality.¹⁶ Hence, community-based (and potentially hospital-based) efforts are likely needed to reduce disparities in high stroke mortality areas.

The application of these findings to the Stroke Belt region is complicated by the observation that a large proportion of Stroke Belt residents live in rural areas. Stroke mortality increases with increasing rurality, with stroke mortality 31% higher in the most rural compared with the most urban areas.¹⁷ Classifying counties by the National Center for Health Statistics Urban-Rural Classification Scheme,¹⁸ in 2016, only 12% of Stroke Belt residents lived in large central metro counties and 32% lived in medium metro counties, compared with 34% and 19% of residents from non–Stroke Belt counties (Figure 3A).⁸ Hence, at least part of the higher stroke mortality in the Stroke Belt is attributable to the more rural nature of the region. Data from Centers for Disease Control⁸ examining the distribution of county level stroke mortality within National Center for Health Statistics rural-urban strata show higher stroke mortality rates in the Stroke Belt across the spectrum of rurality (Figure 3B).

• Download figure
• Download PowerPoint

That a portion of the excess stroke mortality in the Stroke Belt is attributable to its more rural composure raises the question of whether that higher mortality in rural regions is attributable to higher incidence or case-fatality in these regions. In REGARDS analyses similar to that above, compared with urban areas, the age-race-sex adjusted stroke incidence rate was 23% (95% CI, 1%–51%) higher in large rural areas and 30% (95% CI, 3%–62%) higher in the most rural regions (P_trend=0.0073).¹⁹ However, the pattern for case-fatality in rural areas was not significantly different from urban areas (P_trend=0.61).¹⁹ Hence, it appears that higher stroke mortality in rural regions is primarily attributable to higher stroke incidence in these regions.

Collectively, this suggests that the higher stroke mortality in the Stroke Belt is attributable to both a regional effect and the larger proportion of rural residents of the region. The high stroke mortality regions are strongly associated with higher incidence in these regions but perhaps also associated with higher case-fatality in these regions. In contrast, the higher stroke mortality in rural areas appears largely associated with higher stroke incidence, with little effect for case-fatality. Jointly, this suggests that the emphasis to understand the Stroke Belt should primarily focus on contributors to a higher stroke incidence in the region.

Contributors to the Stroke Belt

Two decades ago, we reviewed potential contributors to the Stroke Belt, rating them on a scale from likely to uninvestigated (Figure 4).²⁰ Since that report, there have been substantial advances assessing potential contributors to the Stroke Belt.

• Download figure
• Download PowerPoint

Proportion of Black Residents

In 2017, non-Hispanic blacks comprised 26% of residents of the Stroke Belt, compared with 10% of residents of the rest of the United States.²¹ With age-adjusted stroke mortality among blacks ≈40% higher than whites,^22,23 it is clear the higher proportion of blacks living in the Stroke Belt is a contributor to the Stroke Belt. However, the earlier report failed to quantify the black-white difference in the magnitude of the Stroke Belt. A subsequent analysis of 1997 to 2001 mortality data was restricted to 26 states, with sufficient numbers of blacks to produce stable estimates. Results showed that across 10-year age-strata, the average black-to-white stroke mortality ratios were 6% to 20% higher for southern states (8 Stroke Belt states plus Florida and Virginia) than 16 nonsouthern states,²⁴ showing the impact of the Stroke Belt is larger for the black than white population. Figure 2C shows the race-specific stroke mortality ratio contrasting the Stroke Belt to the non–Stroke Belt. In the late 1960s, the Stroke Belt was associated with an approximate 40% increased stroke mortality for blacks compared with only 25% for whites. Over the 1968 to 2016 period, the excess mortality in the Stroke Belt declined to 25% for blacks but remained higher than the 20% excess for whites (Figure 2C).⁸ Hence, not only does a higher proportion of blacks living in the Stroke Belt contribute to the overall Stroke Belt, but a larger impact of the Stroke Belt in the black (than white) population is also contributing. One could speculate this larger impact for the black population could be due to a larger racial disparity in the prevalence of risk factors in the Stroke Belt than other regions or greater stress through a closer tie to the heritage of slavery in the region.²⁵

Prevalence of Stroke Risk Factors

Analyses contrasting the Stroke Belt to the rest of the nation have shown an inconsistent pattern regarding disparities in risk factors. Some,²⁶ but not all,²⁷ reports have shown a higher prevalence of hypertension in the Stroke Belt. The prevalence of diabetes mellitus has been shown to be higher in the Stroke Belt.²⁸ However, the prevalence of cigarette smoking is highest in the Midwest (18.5%), with the South having the second highest prevalence (16.9%) and lower rates in the Northeast (13.3%) and West (12.3%).²⁹ Finally, the Framingham Risk Score, a risk factor–based estimate of the 10-year risk of a stroke, was only slightly higher in the Stroke Belt compared with other regions.²⁷

Data from the Behavioral Risk Factor Surveillance System has shown the prevalence of obesity (>30 kg/m² body mass index) to be higher in the South, a finding based on self-reported height and weight. This pattern of higher body mass index in the Southeastern United States is not supported by studies that directly measure weight using scales (REGARDS and the National Health and Nutrition Examination Survey).³⁰

Heat maps of the estimated prevalence of the 3 risk factors with the largest population attributable risk for stroke (hypertension, diabetes mellitus, and smoking) support the hypothesis of higher prevalence of stroke risk factors in the Stroke Belt (Figure 5).³¹ However, for the white population, the region with the highest prevalence of these risk factors appears to the west of the coastal plain of North Carolina, South Carolina, and Georgia, the region of the Stroke Belt with the highest stroke mortality (aka, the Buckle of the Stroke Belt).³² The alignment of regions with the highest prevalence of risk factors is less concordant for the black population. While smoking rates are high in all Stroke Belt states, the state-level estimates of average smoking prevalence for the years 2011 to 2013 are higher for the Western Stroke Belt states (Arkansas 26.0%, Mississippi 24.9%, Louisiana 23.5%, and Tennessee 24.3%) than for the Eastern Stroke Belt states (Alabama 23.2%, South Carolina 22.5%, North Carolina 21.0%, and Georgia 20.1%).³³ Most of the work on the role of risk factors as contributors to the Stroke Belt has focused on associations with the prevalence of the factors, with less work focused on impact of control of these same risk factors; however, what data are available suggests relatively small differences in achieved blood pressure levels between the regions.^27,34 Also, we are unaware of additional estimates of hypertension and diabetes mellitus at the state (or smaller) area that incorporated direct measures of blood pressure and glucose/hemoglobin-A1c.

• Download figure
• Download PowerPoint

As such, it seems likely that a higher prevalence of risk factors is a contributor to the higher stroke mortality in the Stroke Belt; however, the risk factor prevalence appears to be higher in the western Stroke Belt, an area that does not encompass the Buckle of the Stroke Belt.

Lifestyle Choices of Diet and Physical Activity

Another commonly hypothesized contributor to the Stroke Belt is presumed poorer dietary intake in the region. The Southern Diet Score is one of 5 eating patterns that was empirically defined by a factor analysis of food frequency data and includes high intake of fried foods, organ meats, processed meats, egg and egg dishes, added fats, high-fat dairy foods, sugar-sweetened beverages, and bread.³⁵ This dietary pattern has been strongly associated with stroke, with a 39% (95% CI, 5%–84%; P=0.009) higher stroke risk in the highest (versus lowest) quartile.³⁵ However, surprisingly there were not large geographic differences in intake, with the mean Southern Diet Score only 0.105 SD higher for residents of the Stroke Belt compared with those from other regions. Because of this lack of geographic difference in consumption, it did not mediate regional differences in stroke risk.³⁵

Several studies have also documented an association between the Mediterranean Diet Score and stroke risk, with ≈15% to 25% lower incident stroke risk for those with the highest Mediterranean Diet Score.^36–39 While we are not aware of a formal mediation analysis, there does not appear to be large regional differences in the Mediterranean Diet Score, with ≈24% of residents of the Stroke Belt having Mediterranean diet scores between 6 and 9 (high intake), 42% with scores of 4 to 5 (moderate intake), and 34% with scores between 0 and 3 (low intake), while the respective scores for those from other regions were 28%, 40%, and 31%.³⁷

While much work remains to better understand the potential contribution of diet to the Stroke Belt, it appears that while the association of diet with stroke risk has been established, there is little evidence that residents of the Stroke Belt consume the diet associated with higher stroke risk.

Physical activity is associated with ≈25% lower risk of incident stroke and a 17% lower stroke mortality, with some differences by sex and by level of physical activity.⁴⁰ In REGARDS, physical activity was assessed by the question “How many times per week do you engage in intense physical activity, enough to work up a sweat?,” categorized as 4 or more times, 1 to 3 times, or no times. This measure has been well-validated⁴¹ and includes both aerobic and resistance training. Those inactive had a 20% (95% CI, 2%–42%) increased stroke risk (relative to those with 4+ times). However, there was virtually no differences in activity levels between residents of the Stroke Belt versus other regions with ≈29.4% versus 30.3% reporting 4+ times per week, 36.7% versus 36.4% reporting 1 to 3 times per week, and 33.8% versus 33.2% reporting no such activity.⁴² Again, while more work remains to assess regional differences in physical inactivity, the lack of geographic differences in the proportion of the population that is inactive implies it is an unlikely contributor to the Stroke Belt.

Prevalence of Inflammation and Infection

Inflammation and infection are established risk factors for stroke, with multiple pathways of action.⁴³ Our previous review identified a higher prevalence of inflammation and infection as one of the most understudied potential contributors to the Stroke Belt (Figure 4). Since that time, a geographic clustering of sepsis mortality has been documented that largely conforms with the Stroke Belt,⁴⁴ and residents of the Stroke Belt have been shown to have a higher adjusted risk of incident sepsis (OR=1.14 [95% CI, 1.02–1.24]).⁴⁵

A higher prevalence of inflammation/infection in the Stroke Belt remains a promising area for investigation, as it has been noted that residents of the Stroke Belt have higher levels of CRP (C-reactive protein) compared with non–Stroke Belt residents (41% versus 38% above 3 mg/L)⁴⁶ and also higher levels of interleukin-6 (mean±SD: 4.2±2.8 ng/mL in the Stroke Buckle, 4.3±2.8 for the rest of the Stroke Belt, and 3.9±2.8 for other regions of the United States).⁴⁷ The contributors to higher levels of CRP are broad but include many factors more prevalent in the Stroke Belt including black race, lower socioeconomic status (SES), lower household cleanliness and smoking.³⁴

Environmental Exposures

Our previous report noted that geographic differences in environmental exposures were also substantially understudied. Since that report, several studies provide insight to these exposures as potential contributors to the Stroke Belt. Low levels of magnesium and selenium are hypothesized to be associated with higher stroke risk, and residents of the Stroke Belt are more likely to have low levels of each of these (OR=5.48 [95% CI, 5.05–5.95] and OR=2.37 [95% CI, 2.22–2.54], respectively).⁴⁸ Conversely, high levels of arsenic and mercury are hypothesized to be associated with higher stroke risk, but Stroke Belt residents are less likely to have high exposure to either (OR=0.33 [95% CI, 0.31–0.35] and OR=0.65 [95% CI, 0.62–0.70], respectively).⁴⁸

The role of selenium as a stroke risk factor is complex and inconsistent.⁴⁹ Higher level of environmental selenium was associated with higher stroke risk in REGARDS, with a monotonic increase in stroke risk across quartiles of selenium with a 34% (95% CI, 10%–64%) higher stroke risk in the highest quartile of selenium relative to the lowest (ie, an association in the unanticipated direction).⁵⁰ If this association is confirmed in other studies, it is unlikely to be a contributor to the Stroke Belt since residents of the Stroke Belt are more likely to be exposed to lower levels of selenium.

In REGARDS, no association was observed between environmental arsenic, mercury or magnesium and stroke risk.⁵⁰ An investigation of serum mercury with stroke risk in REGARDS also failed to find an association,⁵¹ and mercury exposure (assessed in toenail clippings) was not associated with stroke risk in the Health Professional Follow-up Study and the Nurses’ Health Study.⁵²

Finally, in a nested case-cohort study of REGARDS, higher levels of urinary cadmium were associated with higher risk of ischemic stroke (quintile 5 versus quintile 1: HR=1.50 [95% CI, 1.01–2.22], P_trend=0.02); however, no regional differences were observed (P=0.09).⁵³

While more work is clearly needed investigating the potential contribution of environmental exposures to the Stroke Belt, these recent reports seem to suggest that the more commonly hypothesized exposures are not playing a role.

Socioeconomic Status

Lower regional and individual-level SES are well-accepted stroke risk factors.⁵⁴Figure 6 shows the county level SES calculated by standardizing (subtracting mean and dividing by the SD) and then averaging the 2017 county level data for (1) unemployment rate, (2) percent of the population with a high school or lower education, and (3) proportion of the population in poverty.⁵⁵ There is a striking similarity between this county level SES and the county level stroke mortality (Figure 1), with a Spearman correlation of 0.50 (95% CI, 0.47–0.53). Insights for the potential contribution of SES on the Stroke Belt are provided by a REGARDS analysis assessing the association of neighborhood SES (using 6 neighborhood variables assessed for the census block of the participant) with incident stroke risk.⁵⁶ There was a strong association of neighborhood SES with stroke risk (HR=1.60 [95% CI, 1.33–1.93]) comparing those in the lowest quartile (most deprived) to those in the highest quartile. This association was only slightly mediated with adjustment for demographic factors (HR=1.56 [95% CI, 1.36–1.92]) and was substantially mediated but remained marginally significant after adjustment for individual SES as indexed by household income and education (HR=1.25 [95% CI, 0.99–1.56]). However, about half of this remaining association was mediated by further adjustment for the Framingham stroke risk factors, suggesting the pathway for the impact of SES could be through a higher prevalence of these stroke risk factors among individuals with lower SES.⁵⁶

• Download figure
• Download PowerPoint

Despite the strong relationship between SES and incident stroke risk, the only (admittedly dated) attempt to quantify the role of SES as a contributor to the Stroke Belt suggests it plays a relatively minor role. The proportion of the excess risk for death from stroke for residents of the Stroke Belt and Stroke Buckle attributable to individual SES (indexed by individual income and education) was estimated from the National Longitudinal Follow-up Survey for the years 1979 to 1989. This analysis suggested that individual SES is responsible for only ≈5% of the higher stroke risk in the Stroke Belt/Buckle.³² Not only are these data dated, but they rely on mortality data; hence, additional work using data on adjudicated incident stroke events is warranted to assess the role of SES as a contributor to the Stroke Belt.

Other Potential Contributors

Our previous report also considered other potential contributors to the Stroke Belt including regional differences in the coding of death certificates and regional differences in genetic factors. The previous report judged the coding of death certificates as unlikely to be a major contributor, and (to our knowledge) no subsequent reports change this belief. There could be geographic differences in the genetic pool introduced during the settlement of the United States, for example, the higher proportion of residents in the Southeastern United States with English heritage (a region of Europe with high stroke mortality).⁵⁷ As such, genetics remains an uninvestigated potential contributor. Depression, stress, discrimination, and other novel risk factors could also be contributors to the Stroke Belt; however, these factors remain uninvestigated.

Conclusions

Over the past 20 years, there have been substantial advancements in understanding some of the potential contributors to the higher stroke mortality in the Stroke Belt. It appears these contributors are multifaceted and complex. Since the mid-1960s, there has been a minor decrease in the magnitude of the disparity; however, the disparity persists. A current shortcoming of the literature is that many of the reports have focused on an any stroke outcome, while risk factors and intervention differ between hemorrhagic and infarction stroke subtypes. It seems likely that while both higher stroke incidence and higher case-fatality could be contributing, higher stroke incidence is potentially playing a larger role. Although it remains speculative, the larger contributors to the Stroke Belt include a higher risk factor burden, higher levels of inflammation and infection, and lower SES, with potentially smaller contributions from environmental exposures and lifestyle choices. The reasons for an apparent larger impact of the Stroke Belt for African Americans than for whites remains a mystery. Additional investigations should consider access to care and quality of care that could contribute to the Stroke Belt through a higher case-fatality in the region. While solid progress is being made to solving the mystery of the Stroke Belt, much work remains to address and reduce this important disparity.

Footnotes