Factors Explaining Excess Stroke Prevalence in the US Stroke Belt
Background and Purpose— Higher risk and burden of stroke have been observed within the southeastern states (the Stroke Belt) compared with elsewhere in the United States. We examined reasons for these disparities using a large data set from a nationwide cross-sectional study.
Methods— Self-reported data from the 2005 and 2007 Behavioral Risk Factor Surveillance System were used (n=765 368). The potential contributors for self-reported stroke prevalence (n=27 962) were demographics (age, sex, geography, and race/ethnicity), socioeconomic status (education and income), common risk factors (smoking and obesity), and chronic diseases (hypertension, diabetes, and coronary heart disease). Multivariate logistic regression was used in the analysis.
Results— The age- and sex-adjusted OR comparing self-reported stroke prevalence in the 11-state Stroke Belt versus non-Stroke Belt region was 1.25 (95% CI, 1.19 to 1.31). Unequal black/white distribution by region accounted for 20% of the excess prevalence in the Stroke Belt (OR reduced to 1.20; 1.15 to 1.26). Approximately one third (32%) of the excess prevalence was accounted either by socioeconomic status alone or by risk factors and chronic disease alone (OR, 1.12). The OR was further reduced to 1.07 (1.02 to 1.13) in the fully adjusted logistic model, a 72% reduction.
Conclusions– Differences in socioeconomic status, risk factors, and prevalence of common chronic diseases account for most of the regional differences in stroke prevalence.
The risk and burden of stroke do not equally affect populations in different regions. Higher stroke mortality has long been found common among residents of the southeastern states (Stroke Belt region).1,2 Although much is known about the excess burden of stroke in some regions, the reasons for these disparities are less well understood. Suggested potential contributors for the regional difference in stroke mortality were geographic differences in the distribution of major cerebrovascular disease risk factors such as high blood pressure (including lifestyle characteristics leading to hypertension), diabetes, heart disease, cigarette smoking, and obesity, and differences in socioeconomic factors and environmental factors such as soil and water characteristics.3,4 However, many of the hypotheses were based on descriptive or ecological observations.3–6 Available person-level data were not linked to person-level data on stroke,7 limited to a few geographic regions, or did not have large numbers of strokes among both white and black samples from the Stroke Belt states.8–10
Identifying the factors contributing to stroke disparities is a necessary first step in developing appropriate interventions to eliminate the disparities. To accomplish this goal, we used large person-level self-reported data sets from the 2005 and 2007 Behavioral Risk Factor Surveillance System (BRFSS) to examine and quantitatively estimate the contribution of selected indicators for demographics (race/ethnicity), socioeconomic status (education and income), risk factors (smoking, obesity), and chronic diseases (hypertension, diabetes, coronary heart disease) to the excess prevalence of stroke in the Stroke Belt region of the country.
Materials and Methods
The BRFSS is a nationwide, state-based, annual cross-sectional health survey.11 More than 300 000 interviews are completed each year, making BRFSS the largest telephone health survey in the world. The survey uses a multistage, random-digit-dialing method to gather a representative sample from each state’s noninstitutionalized, civilian residents aged ≥18 years. The BRFSS questionnaire consists of core component questions asked in all states and optional questions (modules) asked at the choice of the states. In 2005 and 2007, questions on stroke were asked in all 50 states and the District of Columbia. The median cooperation (participation) rate among all eligible households across 50 states and the District of Columbia was 75.1% and 72.1%, respectively, in 2005 and 2007, and ranged from 49.6% in New Jersey (2007) to 85.3% in Minnesota (2005). A total of 767 495 adults responded in these 2 aggregated survey years. After excluding 2127 respondents whose information on stroke was missing, this analysis included data from 765 368 respondents.
Respondents were considered to have had a stroke if they answered “yes” to the question “Has a doctor, nurse, or other health professional ever told you that you had a stroke?” The presence of other medical conditions was assessed by similar questions for hypertension, diabetes, and coronary heart disease (including questions on heart attack or myocardial infarction and angina or coronary heart disease). Smoking status was categorized as follows: never smoked (<100 cigarettes in lifetime), currently smoke (smoked at least 100 cigarettes in lifetime and currently smoke everyday or some days), or formerly smoked (at least 100 cigarettes in lifetime but not currently smoking). Respondents were asked to report their height and weight. Persons with body mass index ≥25 kg/m2 were defined as overweight, and persons with body mass index ≥30 kg/m2 were defined as obese.
Respondents were asked to report the racial group that best represented their race (eg, white, black, Asian and Pacific Islander, American Indian). In a separate question, they were asked if they were Hispanic or Latino. Respondents were asked to select one of the following 8 categories of the annual household income (in dollars) from all sources: <10 000, 10 000 to <15 000, 15 000 to <20 000, 20 000 to <25 000, 25 000 to <35 000, 35 000 to <50 000, 50 000 to <75 000, and ≥75,000. They were also asked the highest grade or year of school completed for the following 6 categories: never attended school or only kindergarten, Grades 1 to 8 (elementary), Grades 9 to 11 (some high school), Grade 12 or GED (high school graduate or equivalent), college 1 to 3 years (some college or technical school), and college ≥4 years (college graduate).
We used the definition from the National Heart Lung and Blood Institute of “Stroke Belt”3 as the 11 states that had the highest age-adjusted stroke mortality rate in 1980: Alabama, Arkansas, Georgia, Indiana, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia. The remaining 39 states and the District of Columbia combined were defined as the “non-Stroke Belt.” For the current study, self-reported race was classified as black, white, or other racial groups regardless of Hispanic origin.
The χ2 test was used to compare persons in Stroke Belt states with those in non-Stroke Belt states by demographics (age, sex, race, and Hispanic origin), socioeconomic status (education and income), risk factors (overweight, obesity, and smoking status), and prevalence of selected chronic diseases (hypertension, diabetes, and coronary heart disease). The age-standardized prevalence of self-reported stroke was calculated by the direct method using the 2000 US population as the standard. Logistic regression analysis was used to examine the independent impact of geographic region on stroke prevalence adjusting for other covariates (ie, demographics, socioeconomic status, risk factors, and other chronic diseases) and to obtain adjusted ORs and 95% CIs. By entering various covariates in a step-by-step manner, we assessed their contribution to the regional difference in stroke prevalence. First, we compared Stroke Belt versus non-Stroke Belt in a basic model, adjusting for age and sex (with additional adjustment for Hispanic origin in the later models). Then we included race/ethnicity and region simultaneously in the model in addition to age and sex to examine whether race was the contributing factor for the regional difference in stroke prevalence. Based on this model, we further entered either socioeconomic variables or risk factors and chronic diseases variables into the model. The final model included all covariates. To estimate the contribution of a covariate or a group of covariates in the excess stroke prevalence in the Stroke Belt region, we calculated the percentage change of the OR with and without the covariate(s) using the following formula: equation
where OR1 represents OR derived from the basic model; OR2, OR after adjusting for additional covariate(s); and 1.0, OR when there was no excess risk.
The analysis included 153 106 respondents from the Stroke Belt region with 6615 cases of stroke and 612 262 respondents with 21 347 cases of stroke from the non-Stroke Belt region. The software, SUDAAN, was used to analyze these data using a complex sample design.
Table 1 summarizes the demographic and health characteristics of adults aged ≥18 in Stroke Belt versus non-Stroke Belt states. The 2 regions had similar age and sex distribution, although the Stroke Belt had a somewhat lower proportion of older adults aged ≥65. The Stroke Belt had a greater proportion of blacks than did the non-Stroke Belt, whereas a higher proportion of Hispanics lived in non-Stroke Belt states. Residents of the Stroke Belt were less likely to be college graduates and had lower income levels and a higher prevalence of obesity, current smoking, hypertension, diabetes, and coronary heart disease than those in non-Stroke Belt states (P<0.0001).
Table 2 shows the age-standardized prevalence (and 95% CI) of self-reported stroke by region (overall, whites, and blacks). The overall prevalence of stroke was 2.50% (95% CI, 2.43 to 2.56) for noninstitutionalized adults in the 50 states and the District of Columbia. Comparing the 2 geographic regions, the prevalence of stroke was significantly higher in the Stroke Belt than in the non-Stroke Belt region for all racial groups combined and for whites (P<0.001). Although blacks in the Stroke Belt region had a higher prevalence of stroke than blacks in the non-Stroke Belt region, the difference was not statistically significant (P=0.413).
Logistic regression analysis was performed to examine the possible reasons for the excess stroke prevalence in the Stroke Belt (Table 3). The age- and sex-adjusted OR comparing stroke prevalence in the Stroke Belt versus non-Stroke Belt was 1.25 (the first model, basic model, in Table 3). After we additionally adjusted for race/ethnicity (including blacks, whites, and other racial groups as well as Hispanic origin in the model), the OR was reduced to 1.20 indicating that the difference in race/ethnicity distributions between the 2 comparison regions accounted for approximately 20% of the excess stroke risk in the Stroke Belt. The OR was further reduced to 1.12 after we additionally adjusted for socioeconomic status (education and income), a reduction of over one half (52%) from the age- and sex-adjusted OR. Hence, socioeconomic status (differences in educational attainment and income) alone accounted for 32% (ie, 52% minus 20%) of the excess risk. Differences in race/ethnicity and the prevalence of risk factors (overweight, obesity, and smoking) and chronic diseases (hypertension, diabetes, and coronary heart disease; without adjustment for socioeconomic status) also accounted for half (52%) of the excess stroke risk in the Stroke Belt; 32% was the result of the risk factors and chronic diseases alone. When all the covariates were included in the model, the adjusted OR was 1.07 (95% CI, 1.02 to 1.13). Race/ethnicity, socioeconomic status, risk factor, and chronic disease variables were all independently associated with stroke risk (data not shown) and explained approximately three fourths (72%) of the excess stroke risk between the Stroke Belt and non-Stroke Belt regions.
We performed the following additional analyses: (1) using the “Expanded Stroke Belt”3 with 2 additional states (Oklahoma and Texas) adding to the original 11 Stroke Belt states; (2) using the 12 states with the highest stroke mortality in 2004 to 2005 as the high stroke risk region (Alabama, Tennessee, South Carolina, Arkansas, North Carolina, Oklahoma, Oregon, Louisiana, Mississippi, Georgia, Virginia, and Kentucky); and (3) limiting the analyses to persons ≥35 years of age. The results were similar to those presented in Table 3. The ORs were 1.08 to 1.09 comparing high with low stroke risk regions in the fully adjusted logistic models. Race/ethnicity, socioeconomic status, risk factors, and chronic diseases explained 64% to 68% of excess stroke risk.
In an earlier brief report of 23 states and the District of Columbia in 2003, we found that the differences in education level, risk factors, and chronic diseases accounted for 46% of the difference in stroke prevalence between southeastern states and nonsoutheastern states.12 The current study reported here included multiple years’ data in all 50 states and the District of Columbia with a much bigger sample size. These national cross-sectional data demonstrate that the prevalence of stroke was significantly higher in the Stroke Belt region than in the non-Stroke Belt region. Although a greater proportion of blacks lived in the Stroke Belt region, this factor combined with the lower prevalence of Hispanics accounted for only approximately one fifth of excess risk in the Stroke Belt. In addition to race/ethnicity, socioeconomic indicators such as education and income, risk factors (overweight, obesity, and smoking), and prevalent chronic diseases (hypertension, diabetes, and coronary heart disease) were the primary contributors to geographic difference in stroke prevalence. These factors combined accounted for approximately three fourths of the excess stroke prevalence in the Stroke Belt.
Higher stroke death rates have been observed in the southeastern region of the United States than in other regions for approximately 50 years.1,2 Most of the proposed hypotheses for the region difference were based on descriptive or ecological observations. For example, regional variations in the prevalence of hypertension5,6 followed a pattern similar to the reported stroke mortality. Hence, the excess hypertension in the southeastern United States has been thought to contribute to a higher risk of stroke in the Stroke Belt. Socioeconomic status is associated with both stroke risk factors and increased stroke risk13 and aggregate levels of socioeconomic status are lower in the Southeast than the rest of the nation.14 This led to the hypothesis that the stroke belt may be attributable, at least in part, to regional differences in socioeconomic status.3,4 The observation that the coastal southeast had very soft water and has had very high stroke mortality suggested the possibility that water hardness may help to explain the high stroke risk in the Stroke Belt.3 There is a practical advantage of ecological observation that makes it appealing for hypothesis generation. However, the ecological method has inherent weakness and it poses major problems in making biological inferences due to ecological bias (ecological fallacy).15 The strength of the current study is that we were able to study person-level data on the stroke outcome (prevalence) as well as demographic characteristics, a geographic indicator (state), socioeconomic indicators, risk factors, and current chronic disease status.
Limited studies have examined the impact of factors on the elevated stroke risk in the Stroke Belt using person-level data.8,9 In the National Health and Nutrition Examination Survey I Epidemiological Followup Study, Gillum et al found that the major cardiovascular risk factors played a minor role in accounting for the excess incidence of stroke in the southeast region.8 In the evaluation of social status as a contributing factor to stroke mortality in the Stroke Belt region, the National Longitudinal Mortality Study showed that socioeconomic status does not appear to be a major contributor to the excess mortality.9 Less than 16% of this excess stroke mortality was attributable to socioeconomic status. Data on other stroke risk factors were not available in this study, however. In our study, socioeconomic status variables alone accounted for approximately 32% of excess stroke prevalence. Collectively, all the factors considered in this report explain approximately three fourths of the regional difference in stroke prevalence. The remaining effect accounting for the difference could be attributable to residual confounding resulting from incomplete adjustment for hypertension, diabetes, or socioeconomic status. For example, blacks are more likely to have hypertension and more likely to be pharmacologically treated, but they are less likely to achieve blood pressure control.16 Adjusting for prevalence of hypertension did not account for this “residual” confounder. The residual confounding of socioeconomic status measured by education and income has been well described.17
The etiology for stroke risk is multifaceted and related to various factors. However, the study reported here indicates that socioeconomic status, hypertension, diabetes, coronary heart disease, and smoking are still the basic crucial contributors to the disparities. Most of these factors are either modifiable or potentially amenable to interventions. Given these findings, public health interventions are key to reducing the stroke burden in the Stroke Belt region. Stroke is a process that begins with developing risk factors and continues through caring for stroke patients. Policy-level interventions include improving social and physical environments by promoting emergency and health systems policies that eliminate unequal access to preventive, treatment, and rehabilitation services; public education ranges from educating the public on risk factor prevention and control to recognizing stroke signs and symptoms and taking appropriate responses. The interventions also include primary prevention of stroke risk factors beginning in early life and secondary prevention of stroke. As the public health community successfully translates and implements evidence-based interventions to practice in the population level, we should be able to eliminate most of the geographic disparity in stroke.
There are several limitations in this study. Data were collected through a cross-sectional telephone survey. Institutionalized persons, persons without a landline telephone, and persons not able to answer questions because of stroke-related disability were not included. Self-reported data are subject to misclassification. Validation studies showed that a single question on stroke history had sensitivity between 80% and 95% and specificity between 96% and 99%.18,19 Approximately 6% to 10% of the self-reported strokes may be accurately transient ischemic attacks.19 Nonetheless, studies have concluded that simple questionnaire can be confidently used in assessing the prevalence of stroke at the population level.18,19 Other measures such as self-reported age, sex, race/ethnicity, education, income, current smoker, awareness of having hypertension or diabetes, and obesity have been shown to have high or moderate reliability and validity.20 We assume that any bias resulting from the methodological limitations occurred in all regions.
This study examined nonfatal stroke prevalence, not stroke mortality or incidence. Data from the Greater Cincinnati/Northern Kentucky region suggest that excess stroke mortality rates seen in blacks are likely the result of excess stroke incidence and not case fatality.21 Although we examined nonfatal prevalent stroke only, incidence, prevalence, and mortality of stroke in the population are closely interrelated. We correlated 2005 adult stroke mortality from 50 states and the District of Columbia with stroke prevalence at the state level from 2005 BRFSS (Figure). States with higher age-standardized stroke prevalence also had a higher age-standardized stroke mortality rate (r=0.52, P<0.01). The main factors related to disparity in stroke mortality are likely to be similar to those related to disparity in stroke prevalence. The implications of this study on stroke prevalence in the United States can be applied to similar issues on stroke mortality.
In conclusion, our findings suggest that currently known risk factors and socioeconomic status account for most of the difference in self-reported stroke prevalence in the Stroke Belt and non-Stroke Belt regions. To eliminate these stroke disparities, public health measures targeted at improving social determinants of health (educational attainment and income) and risk factor prevention/control should be given a high priority to disproportionately affected regions.
The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
- Received June 30, 2009.
- Accepted July 15, 2009.
Lanska DJ, Kuller LH. The geography of stroke mortality in the United States and the concept of a stroke belt. Stroke. 1995; 26: 1145–1149.
Perry HM, Roccella EJ. Conference report on stroke mortality in the southeastern United States. Hypertension. 1998; 31: 1206–1215.
Obisesan TO, Vargas CM, Gillum RF. Geographic variation in stroke risk in the United States. Region, urbanization, and hypertension in the Third National Health and Nutrition Examination Survey. Stroke. 2000; 31: 19–25.
Voeks JH, McClure LA, Go RC, Prineas RJ, Cushman M, Kissela BM, Roseman JM. Regional differences in diabetes as a possible contributor to the geographic disparity in stroke mortality: the Reasons for Geographic And Racial Differences in Stroke Study. Stroke. 2008; 39: 1675–1680.
Gillum RF, Ingram DD. Relation between residence in the southeast region of the United States and stroke incidence. The NHANES I Epidemiologic Followup Study. Am J Epidemiol. 1996; 144: 665–673.
Howard G, Anderson R, Johnson NJ, Sorlie P, Russell G, Howard VJ. Evaluation of social status as a contributing factor to the Stroke Belt region of the United States. Stroke. 1997; 28: 936–940.
El-Saed A, Kuller LH, Newman AB, Lopez O, Costantino J, McTigue K, Cushman M, Kronmal R. Factors associated with geographic variations in stroke incidence among older populations in four US communities. Stroke. 2006; 37: 1980–1985.
Mokdad AH, Stroup DF, Giles WH. Public health surveillance for behavioral risk factors in a changing environment. Recommendations from the Behavioral Risk Factor Surveillance Team. MMWR Recomm Rep. 23 2003; 52: 1–12.
Morgenstern H. Ecologic studies. In: Kenneth J, Rothman SG, eds. Modern Epidemiology. II ed. Philadelphia: Lippincott Williams & Wilkins; 1998: 459–480.
Ong KL, Cheung BM, Man YB, Lau CP, Lam KS. Prevalence, awareness, treatment, and control of hypertension among United States adults 1999–2004. Hypertension. 2007; 49: 69–75.
O'Mahony PG, Dobson R, Rodgers H, James OF, Thomson RG. Validation of a population screening questionnaire to assess prevalence of stroke. Stroke. 1995; 26: 1334–1337.
Engstad T, Bonaa KH, Viitanen M. Validity of self-reported stroke: the Tromsø Study. Stroke. 2000; 31: 1602–1607.
Nelson DE, Holtzman D, Bolen J, Stanwyck CA, Mack KA. Reliability and validity of measures from the Behavioral Risk Factor Surveillance System (BRFSS). Social Prev Med. 2001; 46 (suppl 1): S3–42.
Kleindorfer D, Broderick J, Khoury J, Flaherty M, Woo D, Alwell K, Moomaw CJ, Schneider A, Miller R, Shukla R, Kissela B. The unchanging incidence and case-fatality of stroke in the 1990s: a population-based study. Stroke. 2006; 37: 2473–2478.