Rural/Urban Differences in the Prevalence of Stroke Risk Factors: A Cross-Sectional Analysis from the REGARDS Study

Abstract

Purpose:

We previously described the magnitude of rural-urban differences in the prevalence of stroke risk factors and stroke mortality. In this report, we sought to extend understanding of rural-urban differences in the prevalence of stroke risk factors by using an enhanced definition of rural-urban status and assessing the impact of neighborhood socioeconomic status (nSES) on risk factor differences.

Methods:

This analysis included 28,242 participants without a history of stroke from the REasons for Geographic And Racial Differences in Stroke (REGARDS) cohort. Participants were categorized into the 6-level ordinal National Center for Health Statistics Urban-Rural Classification Scheme. The prevalence of stroke risk factors (hypertension, diabetes, smoking, atrial fibrillation, left ventricular hypertrophy, and heart disease) was assessed across the rural-urban scale with adjustment for demographic characteristics and further adjustment for nSES score.

Findings:

Hypertension, diabetes, and heart disease were more prevalent in rural than urban regions. Higher odds were observed for these risk factors in the most rural compared to the most urban areas [Odds ratios (95% CI): 1.25 (1.11 – 1.42) for hypertension, 1.15 (0.99 – 1.33) for diabetes, and 1.19 (1.02 – 1.39) for heart disease]. Adjustment for nSES score partially attenuated the odds of hypertension and heart disease with rurality, completely attenuated the odds of diabetes, and unmasked an association of current smoking.

Conclusions:

Some of the higher stroke mortality in rural areas may be due to the higher burden of stroke risk factors in rural areas. Lower nSES contributed most notably to rural-urban differences for diabetes and smoking.

One of the United States Department of Health and Human Services (DHHS)’s primary missions is to “achieve health equity, eliminate disparities, and improve the health of all groups.”¹ Strategic plans of the National Institutes of Health (NIH) within the DHHS to eliminate disparities have specifically defined “health disparity populations” to include “racial and ethnic minorities …, low socioeconomic status, and rural persons [italics added].”²

A previous analysis used the Rural/Urban Commuting Area (RUCA) classification (urban, large city/town, and rural town/isolated region).³ We reported that higher stroke mortality in rural areas reflected higher stroke incidence, not higher case fatality. Differences in the prevalence of risk factors were generally minor across the rural-urban spectrum. Rural areas had a more detrimental profile for hypertension, diabetes, and heart disease. There was no difference in the odds of smoking, left ventricular hypertrophy (LVH), and atrial fibrillation (AF). Further, stroke risk factors modestly attenuated the disparity in stroke incidence across the rural-urban spectrum.³ Consequently, this report seeks to supplement previous findings by describing risk factor differences across an expanded definition of the rural-urban spectrum and examining the impact of neighborhood socioeconomic status (nSES) on associations.

Methods

Data

Data for this cross-sectional analysis came from the baseline assessment of the REGARDS Study cohort, a national, population-based, longitudinal study designed to investigate factors associated with the excess stroke mortality among Black people and residents of the stroke belt region, defined as the eight southern states of North Carolina, South Carolina, Georgia, Tennessee, Mississippi, Alabama, Louisiana, and Arkansas.⁴ Details of REGARDS have been published previously.⁵ Community-dwelling individuals, aged ≥45 years, self-identified as non-Hispanic Black or White, were enrolled between January 2003 and October 2007, resulting in a cohort of 30,239 participants. Twenty percent of the sample was randomly selected from the “buckle” of the Stroke Belt (coastal plain region of North Carolina, South Carolina, and Georgia),⁶ 30% from the Stroke Belt states, and the remaining 50% from the other 40 contiguous states. Individuals were identified from commercially available lists of residents and recruited using an initial mailing followed by telephone contact. Using a computer-assisted telephone interview, trained interviewers obtained demographic information and medical history. A brief physical exam including blood pressure measurements, blood samples, and an electrocardiogram (ECG) was conducted in person 3–4 weeks after the telephone interview. Consent was obtained verbally and later in writing during the in-person assessment. All involved Institutional Review Boards approved the study methods.

The primary predictor variable was the rural-urban status at the county level, which was classified into a six-level ordinal scale of rural-urban status according to the 2006 National Center for Health Statistics (NCHS) Urban-Rural Classification Scheme for Counties.⁷ The NCHS Urban-Rural Classification Scheme is defined by the following: large central metro counties, which are counties in metropolitan statistical areas (MSA) of 1 million or more population that 1) contain the entire population of the largest principal city of the MSA, or 2) are entirely contained in the largest principal city of the MSA, or 3) contain at least 250,000 residents of any principal city of the MSA; large fringe metro counties which are counties in MSAs of 1 million or more population that do not qualify as large central; medium metro counties, which are counties in MSAs of 250,000 to 999,999 population; small metro counties, which are counties in MSAs of less than 250,000 population; micropolitan counties, which are counties in micropolitan statistical areas; and non-core counties, which are nonmetropolitan counties that are not in a micropolitan statistical area. A micropolitan statistical area centers on at least one urban cluster of 10,000–49,999 inhabitants. A micropolitan statistical area may also contain outlying counties that meet commuting criteria to or from the central counties.⁷

We evaluated traditional stroke risk factors.^8,9 Hypertension was classified as having systolic blood pressure greater than or equal to 140 mmHg, diastolic blood pressure greater than or equal to 90 mmHg, or self-reported use of antihypertensive medications. Diabetes was classified as having a fasting blood glucose greater than or equal to 126 mg/dL (or glucose greater than or equal to 200 mg/dL for those failing to fast) or using medications for glucose control. Current cigarette smoking was determined by self-report; participants were asked, “Are you currently smoking cigarettes now, even occasionally?” Left ventricular hypertrophy (LVH) was determined by a centrally read electrocardiogram (ECG) using the Sokolow criteria.¹⁰ Atrial fibrillation was determined by ECG evidence or self-report of a physician diagnosis of atrial fibrillation.¹¹ History of heart disease was determined by self-reported myocardial infarction, ECG evidence of myocardial infarction, or self-reported coronary artery bypass, angioplasty, or coronary stent.

Potential confounding factors for the association of rural-urban status and prevalence of risk factors included age, race, and sex. In addition, since the Southern U.S. has a higher prevalence of risk factors,¹² and all eight of the stroke belt states are more rural on average than other U.S. states,¹³ we adjusted for state of residence to control for the potential confounding effect of region of the country with rurality.

To address the causal pathway of socioeconomic status (SES) on the association of rurality and stroke risk factors, we adjusted for a measure of the nSES at the census tract level. Using the participant’s address at the time of enrollment, we used the approach developed by Diez-Roux and colleagues to create a summary score characterizing the nSES, which incorporates information on wealth/income, education, and occupation that are available from U.S. Census data.^14,15 Use of this metric instead of individual measures of SES allows evaluation of the “SES of the place.”

The primary analysis was limited to 28,242 (93%) participants stroke-free at baseline and having data that could be linked to NCHS Urban-Rural Classification Scheme. The nSES score was available on 25,297 (87%) of the participants.

Statistical Analyses

Differences in the demographic risk factors and nSES score were described by level of rurality, with a test of trend (Pearson correlation for continuous factors and Cochran-Mantel-Haenszel test for categorical). Logistic regression was used to estimate odds ratios for the risk factors across the urban-rural spectrum after adjusting for age, race, sex, and state of residence and test for trend in the prevalence of risk factors across the urban-rural spectrum. This analysis was then repeated with further adjustment for nSES score.

Results

Residents of more rural counties tended to be approximately one year younger and were much less likely to be Black; there were no significant differences in the sex distribution. The nSES score in this cohort ranged from −11.8 to 29.0, with an increasing score indicating an increase in nSES. Atrial fibrillation and heart disease were more prevalent in rural counties, and LVH was less prevalent in rural counties. Hypertension was less prevalent in medium/large fringe metropolitan areas. Smoking and diabetes did not vary by rurality (Table 1).

Table 1.

Demographic and Stroke Risk Factors by Urban-Rural Category (N= 28,242)

		Urban-Rural Census Code (URCC)
		Non-Core (non-metro)	Micropolitan (non-metro)	Small Metro	Medium Metro	Large Fringe Metro	Large Central Metro	P value for trenda
N		1841	3593	3366	7550	3813	8079
Demographic factors	Age, mean±SD, years	64.0±9.1	64.3±9.4	64.4±9.5	64.5±9.4	64.4±9.1	65.2±9.6	< .0001
	Black (%)	24.4	26.9	31.3	35.6	32.0	63.5	< .0001
	Male (%)	43.7	45.2	44.2	44.2	51.2	41.9	.2
Stroke Risk factors	Hypertension (%)	57.8	57.7	58.2	56.5	55.4	60.3	.03
	Diabetes (%)	20.3	21.5	21.0	20.4	18.9	22.3	.2
	Cigarette smoking (%)	13.7	14.1	14.6	14.3	12.5	15.4	.1
	LVH (%)	9.4	8.9	8.2	8.7	10.4	11.1	< .0001
	Atrial Fibrillation (%)	9.5	9.4	8.2	8.7	8.0	7.6	.0003
	Heart Disease (%)	18.5	18.4	19.0	16.1	16.5	15.6	< .0001
Census Tract-level Summary Score for nSES, mean±SD		−2.7 ±2.8	−1.4±4.5	−1.0±4.6	0.3±5.1	2.8±5.7	0.2±5.5	< .0001

^a:P values from analysis of variance for continuous variables and Cochran-Mantel-Haenszel tests of nonzero correlation for categorical variables.

LVH: left ventricular hypertrophy; nSES: neighborhood socioeconomic status

Table 2 shows that after adjustment for age, race, sex, and state of residence, there were increases in the odds of hypertension, diabetes, and heart disease across the urban-rural spectrum (all P < .002). Comparing those in the most rural counties to those in the most urban, the odds of hypertension were 1.25 times (95% CI: 1.11 – 1.42) greater; there was a 1.15 times (95% CI: 0.99 – 1.33) higher odds of diabetes, and a 1.19 times (95% CI: 1.02 – 1.39) higher odds of heart disease. There was no trend for increased smoking, atrial fibrillation, or LVH across the urban-rural spectrum (P > .05).

Table 2.

Odds Ratio (and 95% Confidence Interval) for Stroke Risk Factors by Urban-Rural Category, Adjusting for Age, Race, Sex, State, and Neighborhood Socioeconomic status (nSES) Score

Urban-Rural Census Code (URCC)
Risk Factors	Non-Core (non-metro)	Micropolitan (non-metro)	Small Metro	Medium Metro	Large Fringe Metro	Large Central Metro	P value for trenda
Hypertension
Base model	1.25 (1.11–1.42)	1.19 (1.08 −1.32)	1.18 (1.07 −1.30)	1.05 (0.97–1.14)	1.11 (1.01–1.22)	1.00	< .0001
+ nSES	1.21 (1.06–1.39)	1.14 (1.02–1.27)	1.12 (1.01–1.24)	1.04 (0.95–1.13)	1.15 (1.05–1.27)	1.00	.003
Diabetes
Base model	1.15 (0.99–1.33)	1.19 (1.05–1.34)	1.15 (1.02–1.29)	1.03 (0.93–1.14)	1.06 (0.95–1.19)	1.00	.002
+ nSES	1.00 (0.85–1.19)	1.12 (0.98–1.27)	1.09 (0.96–1.23)	1.02 (0.92–1.13)	1.18 (1.05–1.33)	1.00	.4
Current smoking
Base model	0.89 (0.75–1.05)	0.93 (0.81–1.06)	0.96 (0.84–1.09)	0.92 (0.83–1.03)	0.81 (0.71–0.92)	1.00	.4
+ nSES	0.70 (0.57–0.84)	0.80 (0.69–0.93)	0.89 (0.78–1.03)	0.92 (0.82–1.04)	0.91 (0.79–1.04)	1.00	< .0001
Atrial fibrillation
Base model	1.15 (0.94–1.41)	1.10 (0.93–1.31)	0.97 (0.81–1.15)	1.02 (0.89–1.18)	0.97 (0.82–1.14)	1.00	.1
+ nSES	1.00 (0.79–1.26)	1.10 (0.92–1.32)	0.93 (0.78–1.12)	1.02 (0.86–1.18)	0.96 (0.81–1.14)	1.00	.7
LVH
Base model	1.31 (1.13–1.38)	1.02 (0.86–1.21)	0.93 (0.78–1.09)	0.94 (0.82–1.07)	1.20 (1.04–1.39)	1.00	.7
+ nSES	1.07 (0.86–1.34)	0.97 (0.81–1.16)	0.88 (0.74–1.05)	0.93 (0.80–1.07)	1.22 (1.04–1.42)	1.00	.7
Heart disease
Base model	1.19 (1.02–1.39)	1.14 (1.00–1.30)	1.23 (1.09–1.40)	0.98 (0.88–1.10)	1.02 (0.90–1.15)	1.00	.0006
+ nSES	1.10 (0.92–1.31)	1.12 (0.98–1.29)	1.17 (1.03–1.34)	0.97 (0.87–1.09)	1.07 (0.94–1.21)	1.00	.03

^a:P values for categorical variables are calculated from logistic regression.

LVH: left ventricular hypertrophy

Base model: age, race, sex, and state

Further adjustment for the nSES had a minor attenuating effect on the magnitude of the association of hypertension with the level of rurality, attenuating the most rural to most urban odds ratio from 1.25 to 1.21, and the P value for trend from P < .0001 to P = .0003 (Table 2). This adjustment attenuated the rural-urban odds ratio for heart disease from 1.19 to 1.10, and the test for trend remained significant (P = .03). Adjustment for nSES score completely attenuated the association of diabetes with rurality; for example, from 1.15 for the non-core (non-metro) with adjustment for age-race-sex to 1.00 with further adjustment for nSES score (test for trend changing from P = .002 to P = .4). In contrast, adjustment for nSES score unmasked an association between level of rurality and current smoking, where the odds of smoking in rural counties was 30% lower than in urban counties.

Discussion

Our results indicated hypertension, diabetes, and heart disease increased across the rurality spectrum in this large States cohort. The higher prevalence of these risk factors could be contributing to a higher incidence of stroke in rural areas and thereby be contributing to the greater burden of stroke mortality in rural areas. Using a narrower definition of residence (RUCA), we previously reported that rural regions had a more detrimental stroke and stroke risk factor profile.³ The current results confirm these prior findings using an expanded rural-urban classification. This study also adjusted for nSES score, unlike the previous analysis, when assessing associations between rurality and stroke risk factors. The nSES partly (for hypertension and heart disease) or completely (for diabetes) explained rural-urban differences. Adjustment for nSES score also unmasked an association of current smoking with rurality.

Rural-urban differences in self-reported diabetes and heart disease rates have also been reported from the Behavioral Risk Factor Surveillance System (BRFSS)¹⁶ and the National Health Interview Survey (NHIS).¹⁷ In these data (crude models for the BRFSS and only age-adjusted for the NHIS), for those 18 years of age or older, there was a higher prevalence in rural compared to urban areas for both diabetes and heart disease. For the BRFSS data, adjustment for demographic factors and individual measures of SES attenuated the difference, so the prevalence of diabetes in rural areas became lower than in urban areas (OR = 0.94; 95% CI: 0.84 – 0.99), while the prevalence of heart disease remained significantly higher (OR = 1.09; 95% CI: 1.02 – 1.17). However, both these reports were limited by the self-report of risk factors (so prevalent risk factors would be both underascertained and falsely reported as prevalent), use of a narrow gradation on the rural-urban scale, and lack of adjustment for nSES score. In contrast, in the current analysis, adjustment for nSES score substantially attenuated the magnitude of the rural-urban disparity in diabetes, though only partly attenuated the rural-urban disparity in hypertension and heart disease, which persisted even with this adjustment. This attenuation suggests that some of the rural-urban differences in stroke risk factors are attributable to lower nSES of the rural counties, suggesting that education, occupation, and wealth/income of the community may mediate some of the observed urban-rural differences in disease. A more detailed investigation of the components of nSES would be useful. Reasons for differential attenuation among risk factors, when adjusting for neighborhood-level SES, also require further study.

Higher levels of tobacco use in rural areas have been reported from the BRFSS,¹⁸ NHIS,¹⁷ and our prior REGARDS report.³ Interestingly, our current observation indicated a lower smoking rate in rural areas. Arguably, our expanded definition of the urban-rural spectrum provides a more refined analysis of current smoking by rurality status. Additionally, in the BRFSS data, after adjustment for demographic and individual SES measures, compared to those in rural areas, smoking was significantly lower among those in suburban areas (OR = 0.87; 95% CI: 0.83 – 0.91) and urban areas (OR = 0.90; 95% CI: 0.86 – 0.95) compared to rural areas. This contrasts with the current REGARDS data, where smoking rates were lower in rural areas after adjustment for the nSES score. These differences in findings could exist because rural counties with lower nSES are poorer (with individuals living in neighborhoods with lower SES and poor people being more likely to smoke). When this was considered here, we observed lower smoking rates in rural areas. However, these results must be interpreted with caution because the BFRSS was a younger age sample, and smoking is more common at a young age.¹⁹

This report extends our previous study, which to our knowledge was the first to describe the prevalence of stroke risk factors across the rural-urban spectrum at the national level using direct measurement of risk factors. Given that rural populations have been identified by NIH and the Agency for Health Care Research and Quality as a health disparity population,^{2, 20} the dearth of information on differences in the prevalence of stroke risk factors is somewhat surprising, as stroke is the 5th leading cause of death. Further, many of these same factors are also risk factors for heart disease, the leading cause of death in the U.S. The lack of available data has implications for the generalizability of policies and programs from non-rural and rural areas, further generalizing policies for rural areas from region to region.²¹ One explanation for this lack of data may be that most epidemiological approaches use a clinic-based model where study participants come to an academic medical center clinic for assessments, and these clinics tend to be in urban areas.

A major strength of this report is the design of the REGARDS study that de-links the collection of data from clinical centers but still uses approaches that allow for the direct measurement of risk factors such as blood pressures and glucose levels. This provides the geographic heterogeneity needed to assess rural-urban disparities in risk factors. Limitations of the study warrant discussion. This study was cross-sectional, so we cannot infer the temporal association between rural/urban status and risk factors. People who chose to participate in REGARDS might be different from those who did not, affecting generalizability; whether participation was differential by rural-urban status is difficult to predict, so any impact on findings is uncertain. Additionally, some risk factors relied on self-reported data. For example, smoking status self-report may be susceptible to social desirability bias, which arises when individuals are asked about sociocultural taboos, illegal behavior, and extreme opinions.²² This bias may lead to underreporting of smoking prevalence, though we don’t know if this would differ by rural-urban status. Although REGARDS participants were recruited nationally, only Black and White people were included, so we cannot generalize findings to other racial/ethnic groups.

In conclusion, these findings build on our previous analysis that at least part of the rural-urban disparity in stroke mortality could be attributable to a higher prevalence of hypertension, diabetes, and heart disease in rural areas. Some, but not all, of the higher risk factor prevalence in rural areas is attributable to lower SES in rural areas. More research is needed to assess differences in smoking and in the incidence of risk factors in rural regions compared to urban regions.