Multiple social vulnerabilities to health disparities and hypertension and death in the REasons for Geographic And Racial Differences in Stroke (REGARDS) study

Abstract

Social vulnerabilities increase the risk of developing hypertension and lower life expectancy, but the effect of an individual’s overall vulnerability burden is unknown. Our objective was to determine the association of social vulnerability count and the risk of developing hypertension or dying over ten years, and whether these associations vary by race. We used the Reasons for Geographic and Racial Differences in Stroke (REGARDS) study and included participants without baseline hypertension. The primary exposure was the count of social vulnerabilities defined across economic, education, health and health care, neighborhood and built environment, and social and community context domains. Among 5,425 participants of mean age 64 ±10 SD years of which 24% were Black participants, 1,468 (31%) had 1 vulnerability and 717 (15%) had ≥2 vulnerabilities. Compared to participants without vulnerabilities, the adjusted relative risk ratio (aRRR) for developing hypertension was 1.16 (95% CI 0.99–1.36) and 1.49 (95% CI 1.20–1.85) for individuals with 1 and ≥2 vulnerabilities, respectively. The aRRR for death was 1.55 (95% CI 1.24–1.93) and 2.30 (95% CI 1.75–3.04) for individuals with 1 and ≥2 vulnerabilities, respectively. A greater proportion of Black participants developed hypertension and died than did White participants (hypertension, 38% vs 31%; death, 25% vs 20%). The vulnerability count association was strongest in White participants (p-value for vulnerability count*race interaction: hypertension = 0.046, death =0.015). Overall, a greater number of socially determined vulnerabilities was associated with progressively higher risk of developing hypertension, and an even higher risk of dying over 10 years.

INTRODUCTION

Hypertension is a leading modifiable risk factor for fatal and nonfatal cardiovascular disease (CVD).¹ Preventing its development is a major public health challenge.² Nearly half of US adults have hypertension and among those taking antihypertensive medications, only half have their blood pressure controlled according to the most recent hypertension guideline.¹ This problem is particularly accentuated in Black Americans among whom hypertension prevalence is among the highest in the world.¹ As a result, Black-White disparities in CVD event and mortality rates persist despite safe, effective, and low-cost preventive therapies.¹

In addition to race and ethnicity, social factors such as low educational attainment or low income, among others, are increasingly recognized as leading influences on health outcomes,³ including the development of hypertension^4–6 and premature death.⁷ Individuals in lower socioeconomic strata are particularly likely to have deleterious dietary, smoking, and alcohol consumption habits,⁸ all of which are risk factors for developing hypertension and premature death. Yet, health-related behaviors may not entirely explain the influence of socioeconomic status on poor health outcomes.¹ The Healthy People 2020 public health campaign proposed five domains of social determinants of health, including education, income, neighborhood factors, social context, and health care access.⁹ Compared to White adults, Black adults share a disproportionate amount of these social vulnerabilities which may explain, in part, their higher overall hypertension prevalence and lower overall life expectancy.^{10, 11} Although it is clear that social factors are related to health outcomes independent of clinical risk factors,¹² no studies to date have described the association between experiencing multiple social vulnerabilities simultaneously and the risk of developing hypertension or dying among normotensive adults over the long term.^{13, 14}

Further, the 2017 American College of Cardiology/American Heart Association Guideline for the Prevention, Detection, Evaluation and Management of High Blood Pressure in Adults (2017 ACC/AHA Guideline) recently recommended lower blood pressure thresholds to define hypertension.¹⁵ This lower threshold increases the number of individuals with hypertension and indications for treatment.¹⁶ However, neither the relationship between socially determined vulnerabilities and the risk of developing hypertension or dying, nor the impact of the burden of socially determined vulnerabilities with this lower threshold are well described. We therefore evaluated the association of individual social vulnerabilities and the overall burden of social vulnerabilities and the risk of developing hypertension or dying over ten years. We also determined if these associations varied by race.

METHODS

Data and Resource Availability

The data underlying the findings include potentially identifying participant information and cannot be made publicly available because of ethical/legal restrictions. However, data including statistical code from this article are available to researchers who meet the criteria for access to confidential data. Data can be obtained upon request through the University of Alabama at Birmingham at regardsadmin@uab.edu. Additional information on the REGARDS study is available at www.regardsstudy.org.

Study Design and Population

Details of the Reasons for Geographic and Racial Differences in Stroke (REGARDS) study methods have been previously reported.¹⁷ Briefly, the REGARDS study is a national prospective cohort study of Black and White adults aged >45 years who were recruited between January 2003 and October 2007. Due to the greater prevalence of stroke mortality in Blacks and in the Southeast US, the study oversampled Black adults and residents of the “Stroke Buckle” (coastal North Carolina, South Carolina, and Georgia) and the “Stroke Belt” (the remainder of North Carolina, South Carolina, and Georgia, as well as Alabama, Mississippi, Tennessee, Arkansas, and Louisiana). At baseline, medical history and self-reported health status was collected via computer-assisted telephone interview (CATI). An in-home exam collected fasting blood and urine samples and recorded blood pressure, height, weight, and current medication use. Participants repeated a similar CATI and in-home assessment approximately 10 years later between 2013 and 2016. We restricted the sample to participants free of hypertension at baseline who had either died prior to or completed the 10-year follow-up assessment. Participants who elected not to participate in the 10-year follow-up assessment and were not known to be deceased by the time of the expected assessment, continue to receive follow-up calls and vital status ascertainment; they were included in sensitivity analyses.¹⁸ All REGARDS participants provided written informed consent and the study was approved by all participating institutional review boards.

Main Exposure: Socially Determined Vulnerabilities

The Healthy People 2020 framework guided our categorization of nine social vulnerabilities to health disparities into five domains: 1) economic (annual household income <$35,000), 2) education (<High School), 3) health and health care (residence in a Health Professional Shortage Area [HPSA], lacking health insurance), 4) neighborhood and built environment (living in a zip code with >25% of residents living below the federal poverty line; living in a state with poor public health infrastructure; rural vs. urban residence), and 5) social and community context (social isolation).¹⁹ Counties with HPSA status were classified as “complete HPSA” and counties with only a portion designated as HPSA were classified as “partial HPSA”.²⁰ Public health infrastructure was assessed using data from the America’s Health Ranking (AHR).¹⁷ AHR identified and modeled influences of determinants on health grouped into categories that included Lifestyle, Access to Care, and Disability. Based on these categories, states were annually ranked from 1993 to 2002. States ranking in the bottom quintile for ≥8 years were considered to have poor public health infrastructure. Rural residence was defined using rural-urban commuting area (RUCA) codes as defined by the US Department of Agriculture, with rural defined as RUCA codes 9 (small towns with low commuting to urban areas) and 10 (rural areas) and urban defined as RUCA codes 1–8 (at least a being a small town with high commuting to urban areas). Social isolation from family/friends was defined as having 0 or 1 visits from family or friends in the past month. Social isolation for care was present if the participant reported having no one to care for them if they became ill. All vulnerabilities were assessed at baseline and dichotomized. We did not include race as a social vulnerability because race has a complex relationship with hypertension, including the possibility of a biological contribution.²¹ We stratified by race in our primary analysis and adjusted for race in a pooled, prespecified secondary analysis.

Outcomes

Our primary outcome variable had three levels: 1) alive at follow-up without hypertension (reference group), 2) alive at follow-up with hypertension, or 3) deceased prior to follow-up, as detected through the National Death Index. Blood pressure and prescription medication use were assessed at baseline and follow-up in the home. Blood pressure was defined as the mean of two blood pressure measures taken after a seated five-minute rest with both feet flat on the floor. Blood pressure measurements were performed using a standard aneroid sphygmomanometer.¹⁷ Prescription medication use was based on all medications taken in the previous two weeks and verified via pill bottle review. Prescription antihypertensive medication use was determined based on participant self-report.

The thresholds at which blood pressure values were considered indicative of hypertension were defined by the 2017 ACC/AHA guideline recommendation (systolic blood pressure ≥130 or diastolic blood pressure ≥80 mm Hg).¹⁵ The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation and Treatment of High Blood Pressure (JNC-7) was published in 2003, the same year that the REGARDS study began enrolling patients, and defined hypertension as systolic blood pressure ≥140 or diastolic blood pressure ≥90 mm Hg.¹⁹ As a sensitivity analysis, analyses were repeated using the JNC-7 thresholds. Development of hypertension over the 10 years of follow-up was defined at the follow-up visit (i.e., systolic or diastolic blood pressure values at or above the hypertension thresholds or antihypertensive medication use).

Covariates

Variables collected at baseline were included in analyses as potential confounders when there was potential for the covariate to be associated with both socially determined vulnerability domains and incident hypertension at the 10-year follow-up visit. Model covariates included 1) demographics, 2) medical conditions, 3) functional status, 4) health behaviors, and 5) physiologic variables. Demographics included age at baseline, race, region of residence (Stroke Belt vs non-Stroke Belt), and gender.¹⁷ Medical conditions included history of CVD, history of stroke, high cholesterol (use of lipid lowering medications, low density lipoprotein [LDL] cholesterol >130 mg/dL, or self-report of a physician or nurse diagnosis) and diabetes (use of glucose-lowering medications or fasting blood glucose ≥126 mg/dL or non-fasting glucose ≥200 mg/dL). Use of insulin (self-reported or from medication data) was entered as a separate covariate. Functional status was measured by the Physical Component Summary (PCS) and Mental Component Summary (MCS) scores from the Short Form 12-item survey.²² Health behaviors included cigarette smoking (currently smoking vs. not), alcohol use based on sex-specific National Institute on Alcohol Abuse and Alcoholism cut points (7+ drinks per week for women, 14+ drinks per week for men), physical activity (enough activity to work up a sweat on 0, 1–3, or ≥4 days of the week), and the Mediterranean diet adherence score (low adherence 0–4; high adherence 5–9).^23–26 Physiologic variables included obesity (≥30.0 kg/m²), systolic blood pressure, diastolic blood pressure, log transformed C-reactive protein (CRP), urinary albumin-to-creatinine ratio (ACR, ≥30 vs <30 mg/g), and estimated glomerular filtration rate (eGFR) using the CKD-EPI equation (<60 vs >60 ml/minute/1.73m²).

Statistical Analysis

Following our previous approach, we first examined age-adjusted associations between each of the candidate socially determined vulnerabilities and being alive without hypertension (referent), being alive with hypertension, or having died before the second in-home visit using multinomial logistic regression to estimate relative risk ratios (RRR; i.e., exponentiated coefficients) and 95% confidence intervals (CI).^27–30 Vulnerabilities associated with both incident hypertension and death with p<0.10 in the age-adjusted models were retained for inclusion in the count of vulnerabilities. Included vulnerabilities were examined in a correlation matrix and correlation coefficients were tabulated. Participants were classified according to their number of vulnerabilities. We also examined the age-adjusted association between each vulnerability and death prior to the follow-up visit.

We then described the distribution of characteristics of participants by vulnerability count. Differences across vulnerability counts were assessed using logistic regression, modeling vulnerability count as the independent variable.

Multinomial logistic regression was used to determine associations between each individual vulnerability, as well as, separately, the count of vulnerabilities, with hypertension and survival status at the follow-up visit expressed as adjusted RRR (aRRR). Sequential models were constructed for each exposure: model 1 included only for age; model 2 included age and demographics, medical conditions, functional status, and health behaviors; and model 3 included model 2 covariates and physiologic variables.

Next, we assessed if the association between vulnerability count and likelihood of developing hypertension during follow-up differed between Black and White participants. To do so, we included an interaction term (vulnerability count*race) in the age-adjusted models, using a Wald test to determine statistical significance of the interaction. Statistical significance was defined as a two-tailed p <0.05. The sample was then stratified by race and the crude RRR and aRRR for vulnerability count were calculated separately within each race group.

To account for potential selection bias introduced by participants electing not to participate in the follow-up visit, we conducted two sensitivity analyses. First, we used weights for the inverse probability of not participating in the follow-up visit (Supplemental Tables S9 & S10).³¹ We modeled the relationship between vulnerability count and nonparticipation in the follow-up visit using logistic regression. Predicted probabilities from these models were inverted to create stabilized weights. The weights for the inverse probability of nonparticipation were then applied to the multinomial logistic regression models described above with hypertension and death as the outcomes. Second, we conducted an analysis including a fourth outcome category in the multinomial logistic regression for participants who were alive but did not participate in the follow-up visit (Supplemental Tables S11 & S12).

All analyses were conducted for the cohort defined by the 2017 ACC/AHA guideline, and, separately, for the cohort defined by the JNC-7 guideline.

Missing data for covariates and exposure information at baseline were multiply imputed using chained equations with 30 imputations. Non-imputed data were used to describe participants’ baseline characteristics; all other results are after multiple imputation was performed. The imputed variables included those described in the Covariates section above. We also included the outcome variable as a predictor in the imputation models, stratified models by race, and checked for convergence. Variables with the highest percentage of missing values included diet (13%), annual household income (6%) and social isolation (3%). A complete case analysis was conducted as a sensitivity analysis. All analyses were performed using SAS software version 9.4 (SAS Institute, Cary, NC) and Stata 14 (StataCorp, College Station, TX).

RESULTS

Study Populations

Between 2003 and 2007, 30,239 participants enrolled in the REGARDS study. Of these, 7,702 (25%) had blood pressure <130/80 mm Hg (ACC/AHA threshold) and were not using antihypertensive medications. Among participants without hypertension according to the ACC/AHA threshold, 5,425 (70%) had survival or hypertension status available at the follow-up visit of which 4,736 (87%) had complete baseline data on socially determined vulnerabilities and were included in the 2017 ACC/AHA cohort analysis (Supplemental Figure S1). Of the 22,028 participants who were excluded due to hypertension status at baseline, 21,954 had hypertension or were using antihypertensive medications and 74 had unknown hypertension status. Supplemental Tables S1-S2 display the baseline characteristics for participants who did and did not complete the 10-year follow-up but were alive.

Selection of Vulnerabilities

Over median 9.5 years [IQR 8.7–9.9] of follow-up, 1,785 (33%) participants developed hypertension and 1,135 (21%) participants died. In age-adjusted models, four individual vulnerabilities were associated with developing hypertension with p<0.10 and were selected for inclusion in the count of vulnerabilities: 1) low educational attainment, 2) low annual household income, 3) living in a zip code with high poverty, and 4) lacking health insurance (Table 1). The included vulnerabilities were not highly correlated (correlation coefficients <0.30, Supplemental Tables S21 and S22).

Table 1.

Age-Adjusted Association of Individual Vulnerabilities and Hypertension or Death at follow-up in the 2017 ACC/AHA Cohort (n = 5,425)

Vulnerability	Alive with Hypertension vs Alive without Hypertension		Deceased vs Alive without Hypertension
	RRR (95% CI)	P-value	RRR (95% CI)	P-value
Income <$35,000	1.45 (1.26,1.66)	<0.001	2.91 (2.45,3.45)	<0.001
Education < High School	1.79 (1.35,2.38)	<0.001	3.93 (2.94,5.24)	<0.001
No health insurance	1.47 (1.12,1.92)	0.005	3.07 (2.13,4.42)	<0.001
>25% residents of zip code live below poverty line	1.35 (1.13,1.61)	0.001	1.54 (1.23,1.92)	<0.001
Residence in Complete or Partial HPSA	0.96 (0.84,1.08)	0.468	1.03 (0.87,1.20)	0.750
Lives in a state with poor public health infrastructure	1.07 (0.94,1.21)	0.332	1.16 (0.99,1.37)	0.075
Rural/ Urban Residence	0.95 (0.53,1.70)	0.868	1.11 (0.53,2.32)	0.790
Social Isolation from friends/family	1.06 (0.88,1.29)	0.530	1.36 (1.07,1.73)	0.014
Social Isolation – no one to provide care if fall ill	0.95 (0.79,1.14)	0.600	1.06 (0.84,1.33)	0.620

CI = Confidence interval, HPSA = Healthcare Professional Shortage Area, REGARDS = Reasons for Geographic and Racial Differences in Strokes, RRR = Relative Risk Ratio.

Baseline Characteristics by Count of Vulnerabilities

Among participants who did not have missing vulnerability count information, 2,551 (54%) participants had no vulnerabilities, 1,468 (31%) had one, 582 (12%) had two, 126 (3%) had three, and 9 (<1%) had four. Older age, Black race, and female sex were associated with higher numbers of vulnerabilities (Table 2). Participants with more vulnerabilities were more likely to reside in the Stroke Belt or HPSAs, and to be socially isolated. As the number of vulnerabilities increased, participants were more likely to have diabetes, use insulin, have worse functional status, engage in poor health behaviors (e.g., cigarette smoking, no physical activity, low adherence to a Mediterranean diet), be obese, have higher systolic blood pressure, have elevated C-reactive protein, and have worse renal function.

Table 2.

Baseline Characteristics, Overall and by Vulnerability Count in the 2017 ACC/AHA Cohort (n = 5,425).

	All included participants		Participants who did not have a missing vulnerability count*
	Overall	N with missing data	Overall	N with missing data	Number of Vulnerabilities			p-value
					0	1	≥ 2
No. of participants	5425		4736		2551	1468	717
Vulnerabilities Included in the count
Income <$35,000	1782 (37.2)	63 (11.6)	1758 (37.3)	28 (0.6)	0	1093 (74.5)	665 (96.5)	<0.001
Education < High School	395 (7.3)	2 (<0.1)	346 (7.3)	0	0	46 (3.1)	300 (41.8)	<0.001
No health insurance	285 (5.3)	3 (<0.1)	259 (5.5)	0	0	49 (3.3)	210 (29.3)	<0.001
>25% residents of zip code live below poverty line	745 (14.0)	95 (1.8)	683 (14.4)	7 (0.1)	0	280 (19.1)	403 (56.8)	<0.001
Vulnerabilities Considered but not included in the count
Lives in a state with poor public health infrastructure	1809 (33.4)	1 (<0.1)	1577 (33.3)	0	726 (28.5)	538 (36.6)	313 (43.7)	<0.001
Lives in complete or partial HPSA	2140 (39.5)	1 (<0.1)	1850 (39.1)	0	943 (37.0)	562 (38.3)	345 (48.1)	<0.001
Rural/Urban Residence	61 (1.3)	562 (10.4)	51 (1.2)	485 (10.2)	21 (0.9)	14 (1.1)	16 (2.5)	0.004
Social Isolation from friends/family	615 (11.5)	79 (1.5)	535 (11.4)	56 (1.2)	266 (10.5)	167 (11.5)	102 (14.7)	0.009
Social Isolation – no one to provide care if fall ill	714 (14.0)	322 (5.9)	614 (13.7)	265 (5.6)	254 (10.5)	230 (16.5)	130 (19.8)	<0.001
Demographics
Black race	1277 (23.5)	0	1154 (24.4)	0	410 (16.1)	376 (25.6)	368 (51.3)	<0.001
Age in years	63.6 (9.6)	0	63.4 (9.5)	0	61.5 ± 8.8	65.7 ± 9.6	65.2 ± 10.4	<0.001
Female	2987 (55.1)	0	2552 (53.9)	0	1289 (50.5)	840 (57.2)	423 (59.0)	<0.001
Stroke Belt residence	3006 (55.4)	0	2615 (55.2)	0	1295 (50.8)	854 (58.2)	466 (65.0)	<0.001
Medical Conditions or Medications
History of CVD	695 (13.0)	96 (1.8)	603 (12.9)	78 (1.6)	285 (10.4)	347 (13.8)	330 (15.4)	<0.001
History of stroke	179 (3.3)	23 (0.4)	156 (3.3)	17 (0.4)	49 (1.8)	99 (3.9)	115 (5.3)	<0.001
High cholesterol	3343 (61.6)	0	2907 (61.4)	0	1518 (59.5)	942 (64.2)	447 (62.3)	0.012
Diabetes	563 (10.7)	178 (3.3)	484 (10.6)	156 (3.3)	181 (7.3)	186 (13.1)	117 (17.2)	<0.001
Statin Use	1273 (23.6)	21 (0.4)	1099 (23.3)	17 (0.4)	587 (23.1)	345 (23.6)	167 (23.4)	0.928
Insulin use	145 (2.7)	2 (<0.1)	125 (2.6)	1 (<0.1)	47 (1.8)	45 (3.1)	33 (4.6)	<0.001
Functional Status
PCS-12: SF-12 Physical	52.5 [45.3, 55.9]	187 (3.4)	52.7 [45.6, 55.9]	139 (2.9)	53.8 [48.9, 56.2]	50.7 [42.3, 55.5]	48.7 [38.3, 54.4]	<0.001
MCS-12: SF-12 Mental	56.6 [52.4, 59.0]	187 (3.4)	56.4 [52.4, 59.0]	139 (2.9)	56.8 [53.3, 58.8]	56.4 [51.7, 59.3]	55.5 [48.4, 59.4]	<0.001
Health Behaviors
Current cigarette smoking	754 (13.9)	15 (0.3)	671 (14.2)	15 (0.3)	230 (9.1)	261 (17.8)	180 (25.1)	<0.001
Heavy Alcohol Use	210 (3.9)	79 (1.5)	182 (3.9)	62 (1.3)	111 (4.4)	48 (3.3)	23 (3.3)	0.150
No physical activity	1590 (29.7)	74 (1.4)	1371 (29.3)	59 (1.2)	617 (24.4)	491 (33.8)	263 (37.6)	<0.001
Low adherence to a Mediterranean diet	2167 (51.0)	1177 (21.7)	1883 (50.9)	1036 (21.9)	975 (46.6)	599 (53.5)	309 (63.3)	<0.001
Physiologic Variables
BMI ≥30.0 kg/m2	1130 (20.9)	16 (0.3)	1013 (21.4)	13 (0.3)	502 (19.7)	320 (21.9)	191 (26.9)	<0.001
SBP, mm Hg	113.9 (9.4)	0	113.9 (9.4)	0	113.4 ± 9.3	114.4 ± 9.2	114.7 ± 9.7	<0.001
DBP, mm Hg	69.5 (6.5)	0	69.5 (6.5)	0	69.8 ± 6.3	69.3 ± 6.6	69.3 ± 6.9	0.028
CRP > 3 mg/L	1570 (30.7)	307 (5.7)	1379 (30.8)	258 (5.4)	638 (26.1)	477 (34.5)	264 (40.6)	<0.001
UACR ≥ 30 mg/g	383 (7.4)	220 (4.1)	333 (7.3)	198 (4.2)	145 (5.9)	99 (7.1)	89 (13.3)	<0.001
eGFR < 60 ml/minute/1.73m2	308 (5.9)	192 (3.5)	260 (5.7)	160 (3.4)	89 (3.6)	109 (7.7)	62 (9.2)	<0.001

Values are count (%), mean ± standard deviation, or median [interquartile range].

^*Includes only those participants who either had complete data on the vulnerabilities included in the count or sufficient data to categorize the participants vulnerability count. That is, although 146 and 21 of these participants had missing income and zip code level poverty, respectively, all of these had a vulnerability count score of ≥ 2 when considering the other vulnerabilities, so the missing values could not affect their categorization. Overall, 959 participants in the JNC-7 Cohort had a missing vulnerability count which was imputed.

ACC/AHA = American College of Cardiology/American Heart Association; BMI = Body mass index; CRP = C reactive protein; DBP = Diastolic blood pressure; eGFR = Estimated glomerular filtration rate; HPSA = Healthcare Professional Shortage Area; MCS = Mental Component Summary; PCS = Physical component summary; REGARDS = Reasons for Geographic and Racial Differences in Strokes; SBP = Systolic Blood Pressure; SF-12 = 12-Item Short Form Survey; UACR = Urinary Albumin/Creatinine Ratio.

Association of vulnerabilities and 10-year outcomes

After multivariable adjustment, participants with low annual household income had higher risk of developing hypertension (aRRR 1.27, 95% CI 1.09–1.48) and even higher risk of dying (aRRR 1.96, 95% CI 1.60, 2.40). This pattern was also observed for low education (aRRR for hypertension 1.42, 95% CI 1.05–1.91 and for dying 1.71, 95% CI 1.24–2.37). Those without health insurance had nonsignificantly higher risk of developing hypertension (aRRR 1.29, 95% CI 0.97–1.71) but the highest risk for dying of any of the four vulnerabilities (aRRR 2.29, 95% CI 1.54–3.41). Area poverty was not significantly associated with either developing hypertension (aRRR 1.06, 95% CI 0.88–1.29) or dying (aRRR 0.94, 95% CI 0.72–1.22) after full adjustment (Table 3). The proportion of participants who developed hypertension over follow-up was similar regardless of the number of vulnerabilities, but the proportion who died during follow-up was progressively higher as the number of vulnerabilities increased: 0 vs 1 vs ≥2 vulnerabilities for hypertension: 33% vs 31% vs 33%, respectively; for death: 15% vs 33% vs 40%, respectively (Table 4). Compared to participants without any vulnerabilities, the aRRR for developing hypertension was 1.16 (95% CI 0.99–1.36) for individuals with 1 vulnerability and 1.49 (95% CI 1.20–1.85) for individuals with ≥2 vulnerabilities. The aRRR for death was 1.55 (95% CI 1.24–1.93) and 2.30 (95% CI 1.75–3.04) for individuals with 1 and ≥2 vulnerabilities, respectively. For both hypertension and death, the p-values for trend across vulnerability count categories were statistically significant (p<0.001) in all models.

Table 3.

Individual Vulnerabilities and 10-Year Risk of Hypertension and Death in the 2017 ACC/AHA Cohort (n = 5,425).

	Individual Vulnerability
	Income <$35,000	Education < High School	No health insurance	>25% residents of zip code live below poverty line
Alive without Hypertension
Percent of participants	30.0%	21.8%	38.4%	34.0%
RRR (95% CI)
All Models	1 [Reference]	1 [Reference]	1 [Reference]	1 [Reference]
Alive with Hypertension
Percent of participants	30.4%	28.3%	39.4%	36.3%
RRR (95% CI)
Crude	1.50 (1.31,1.71)	1.84 (1.39,2.44)	1.34 (1.03,1.75)	1.34 (1.12,1.60)
Model 1*	1.45 (1.26,1.66)	1.79 (1.35,2.38)	1.47 (1.12,1.92)	1.35 (1.13,1.61)
Model 2†	1.28 (1.10,1.48)	1.43 (1.07,1.91)	1.28 (0.97,1.68)	1.10 (0.91,1.33)
Model 3‡	1.27 (1.09,1.48)	1.42 (1.05,1.91)	1.29 (0.97,1.71)	1.06 (0.88,1.29)
Deceased
Percent of participants	39.6%	49.9%	22.2%	29.7%
RRR (95% CI)
Crude	4.47 (3.82,5.23)	5.27 (4.06,6.84)	1.01 (0.73,1.40)	1.51 (1.24,1.84)
Model 1*	2.91 (2.45,3.45)	3.93 (2.94,5.24)	3.07 (2.13,4.42)	1.54 (1.23,1.92)
Model 2†	1.97 (1.62,2.41)	1.88 (1.37,2.58)	2.28 (1.53,3.38)	0.96 (0.74,1.24)
Model 3‡	1.96 (1.60,2.40)	1.71 (1.24,2.37)	2.29 (1.54,3.41)	0.94 (0.72,1.22)

Values are relative risk ratios (95% confidence interval) obtained from multinomial logistic regression models, unless otherwise indicated. In all models, “alive without hypertension” is the base outcome.

^*Model 1 is adjusted for age.

^†Model 2 is adjusted for the covariates in Mode1 and socio-demographics (gender, region, social isolation) + medical conditions (history of CVD, history of stroke, high cholesterol, diabetes, statin use, insulin use) + functional status (physical and mental component summary from the Short Form-12 instrument) + health behaviors (alcohol use, smoking status, physical activity, Mediterranean diet).

^‡Model 3 is adjusted for the covariates in Model 2 and physiologic and laboratory characteristics (body mass index, systolic blood pressure, diastolic blood pressure, log transformed C-reactive protein, urinary albumin-to-creatinine ratio, and estimated glomerular filtration rate).

2017 ACC/AHA = 2017 American College of Cardiology/American Heart Association Guideline for the Prevention, Detection, Evaluation and Management of High Blood Pressure in Adults; CI = confidence interval; RRR = relative risk ratio.

Table 4.

Cumulative Vulnerabilities and 10-Year Risk of Hypertension and Death in the 2017 ACC/AHA Cohort (n = 5,425).

	Number of Vulnerabilities			P for trend
	0	1	≥2
Alive without Hypertension
Percent of participants	52.3%	36.5%	26.2%
RRR (95% CI)
All Models	1 [Reference]	1 [Reference]	1 [Reference]
Alive with Hypertension
Percent of participants	32.7%	30.7%	33.4%
RRR (95% CI)
Crude	1 [Reference]	1.33 (1.15, 1.54)	1.90 (1.56, 2.31)	<0.001
Model 1*	1 [Reference]	1.29 (1.11, 1.49)	1.88 (1.55, 2.29)	<0.001
Model 2†	1 [Reference]	1.17 (1.00, 1.36)	1.51 (1.22, 1.86)	<0.001
Model 3‡	1 [Reference]	1.16 (0.99, 1.36)	1.49 (1.20, 1.85)	<0.001
Deceased
Percent of participants	15.0%	32.8%	40.4%
RRR (95% CI)
Crude	1 [Reference]	3.14 (2.64, 3.73)	5.59 (4.53, 6.89)	<0.001
Model 1*	1 [Reference]	2.16 (1.78, 2.61)	4.58 (3.64, 5.78)	<0.001
Model 2†	1 [Reference]	1.55 (1.25, 1.93)	2.43 (1.85, 3.18)	<0.001
Model 3‡	1 [Reference]	1.55 (1.24, 1.93)	2.30 (1.75, 3.04)	<0.001

Values are relative risk ratio (95% confidence interval) obtained from multinomial logistic regression models, unless otherwise indicated. In all models, “alive without hypertension” is the base outcome.

^*Model 1 is adjusted for age.

^†Model 2 is adjusted for the covariates in Mode1 and socio-demographics (gender, region, social isolation) + medical conditions (history of CVD, history of stroke, high cholesterol, diabetes, statin use, insulin use) + functional status (physical and mental component summary from the Short Form-12 instrument) + health behaviors (alcohol use, smoking status, physical activity, Mediterranean diet).

^‡Model 3 is adjusted for the covariates in Model 2 and physiologic and laboratory characteristics (body mass index, systolic blood pressure, diastolic blood pressure, log transformed C-reactive protein, urinary albumin-to-creatinine ratio, and estimated glomerular filtration rate).

2017 ACC/AHA = 2017 American College of Cardiology/American Heart Association Guideline for the Prevention, Detection, Evaluation and Management of High Blood Pressure in Adults; CI = confidence interval; RRR = relative risk ratio.

Race stratified results

Compared to White participants, Black participants were more likely to have a higher vulnerability count at baseline and were as a group more likely to develop hypertension or die during follow-up (Black vs White participants: hypertension, 38% vs 31%; death, 25% vs 20%) (Table 5). However, vulnerability count was associated with increased risk of developing hypertension in White but not Black participants (p for vulnerability count*race interaction = 0.046). Vulnerability count was associated with increased risk of death in both race groups, but the association was stronger in White than in Black participants (p for vulnerability count*race interaction = 0.015).

Table 5.

Cumulative Vulnerabilities and 10-Year Risk of Hypertension and Death in the 2017 ACC/AHA Cohort, by Race (n = 5,425).

	White Participants				Black Participants				P for interaction
	Number of Vulnerabilities			P for trend	Number of Vulnerabilities			P for trend
	0	1	≥2		0	1	≥2
Alive without Hypertension
Percent of participants	54.7%	36.8%	27.2%		41.9%	35.9%	25.3%
RRR (95% CI)
All Models	1 [Reference]	1 [Reference]	1 [Reference]		1 [Reference]	1 [Reference]	1 [Reference]
Alive with Hypertension
Percent of participants	31.0%	28.2%	35.6%		39.7%	37.1%	31.6%
RRR (95% CI)
Crude	1 [Reference]	1.32 (1.12,1.57)	2.01 (1.55,2.61)	<0.001	1 [Reference]	1.10 (0.81,1.49)	1.29 (0.93,1.78)	0.135	0.031
Model 1*	1 [Reference]	1.28 (1.08,1.52)	2.01 (1.54,2.62)	<0.001	1 [Reference]	1.05 (0.77,1.43)	1.22 (0.88,1.70)	0.251	0.037
Model 2†	1 [Reference]	1.23 (1.03,1.46)	1.81 (1.37,2.39)	<0.001	1 [Reference]	0.99 (0.73,1.36)	1.11 (0.78,1.58)	0.587	0.046
Model 3‡	1 [Reference)	1.21 (1.01,1.45)	1.82 (1.37,2.43)	<0.001	1 [Reference]	0.99 (0.72,1.37)	1.06 (0.74,1.52)	0.791	0.046
Deceased
Percent of participants	14.3%	35.0%	37.2%		18.4%	27.0%	43.1%
RRR (95% CI)
Crude	1 [Reference]	3.50 (2.88,4.25)	5.41 (4.08,7.16)	<0.001	1 [Reference]	1.87 (1.27,2.76)	4.28 (2.95,6.21)	<0.001	0.075
Model 1*	1 [Reference]	2.28 (1.84,2.83)	4.82 (3.52,6.60)	<0.001	1 [Reference]	1.30 (0.85,1.97)	2.59 (1.73,3.88)	<0.001	0.009
Model 2†	1 [Reference]	1.70 (1.33,2.19)	2.82 (1.98,4.00)	<0.001	1 [Reference]	1.00 (0.63,1.58)	1.78 (1.13,2.80)	0.008	0.017
Model 3‡	1 [Reference]	1.70 (1.32,2.19)	2.69 (1.88,3.85)	<0.001	1 [Reference]	1.00 (0.62,1.60)	1.70 (1.06,2.73)	0.019	0.015

Values are relative risk ratio (95% confidence interval) obtained from multinomial logistic regression models, unless otherwise indicated. In all models, “alive without hypertension” is the base outcome.

^*Model 1 is adjusted for age.

^†Model 2 is adjusted for the covariates in Mode1 and socio-demographics (gender, region, social isolation) + medical conditions (history of CVD, history of stroke, high cholesterol, diabetes, statin use, insulin use) + functional status (physical and mental component summary from the Short Form-12 instrument) + health behaviors (alcohol use, smoking status, physical activity, Mediterranean diet).

^‡Model 3 is adjusted for the covariates in Model 2 and physiologic and laboratory characteristics (body mass index, systolic blood pressure, diastolic blood pressure, log transformed C-reactive protein, urinary albumin-to-creatinine ratio, and estimated glomerular filtration rate).

2017 ACC/AHA = 2017 American College of Cardiology/American Heart Association Guideline for the Prevention, Detection, Evaluation and Management of High Blood Pressure in Adults; CI = confidence interval; RRR = relative risk ratio.

Sensitivity Analyses

Results of the analyses using the thresholds at which the JNC-7 guidelines considered blood pressure values to be indicative of hypertension are shown in Supplemental Figure S2 and Supplemental Tables S3–S8. In addition to the four individual vulnerabilities associated with hypertension in age-adjusted models when using the 2017 ACC/AHA guideline definitions, living in a state with poor public health infrastructure was also associated with risk of developing hypertension (p=0.075) and included in the vulnerability count. Results were similar but associations less pronounced when using the JNC-7 definitions. Unlike the analysis using the 2017 ACC/AHA definitions, associations did not statistically vary by race when using the JNC-7 definitions (Supplemental Table S8). To explore potential reasons why results between the JNC-7 and 2017 ACC/AHA cohorts differed, we (1) defined the multiple vulnerabilities exposure equivalently across cohorts (i.e., removed living in a state with poor public health infrastructure from the exposure definition of the JNC-7 cohort), (2) compared participant characteristics between those included in both the 2017 ACC/AHA and JNC-7 cohorts vs only the JNC-7 cohort, and (3) compared the percentage of patients categorized as having hypertension who reported being treated with antihypertensive medication in the 2017 ACC/AHA vs JNC-7 cohort (Supplemental Tables S17 – S20). None of the results were convincingly revealing.

Other sensitivity analyses (inverse probability weighting, 4-outcome multinomial, complete case, and stratifying by residence in the Stroke Belt region) did not substantially vary from the primary analysis for either cohort (Supplemental Tables S9–S16).

DISCUSSION

In this analysis of the REGARDS cohort, we observed a graded association between socially determined vulnerabilities to health disparities and the development of incident hypertension and dying over 9.4 years of follow-up. Graded associations persisted after multivariable adjustment, and when considering 2017 ACC/AHA Guideline and JNC-7 Guideline blood pressure thresholds for defining hypertension. The relative associations of multiple vulnerabilities varied between White and Black adults with associations of multiple vulnerabilities being stronger among White than Black participants. These findings suggest that the burden of socially determined vulnerabilities has a graded association with the development of hypertension but an even more pronounced association with the risk of death. To our knowledge, this is the first study to demonstrate the risks of developing hypertension and dying over nearly 10 years conferred by an individual’s burden of socially determined vulnerabilities to health disparities.

Concordant with prior research, we observed that participants living in poor socioeconomic conditions experienced the highest likelihood of developing hypertension.⁵ This was particularly true for participants with low income and low educational attainment. In a meta-analysis of 51 studies, the pooled odds ratio for risk of hypertension comparing the lowest vs. highest socioeconomic status categories was 1.19 (95% CI, 0.96 – 1.48) for income, 1.31 (95% CI, 1.04 – 1.64) for occupation, and 2.02 (95% CI, 1.55 – 2.63) for education.⁵ In both the current analysis and the prior meta-analysis, education level was the strongest indicator of hypertension risk.⁵ There are several potential reasons that education may influence hypertension risk. Those with higher education are more likely to know about hypertension, its causes, and consequences, and subsequently live a healthier lifestyle.³² For example, those with higher education are more likely to exercise and receive preventative health care, and less likely to abuse alcohol or smoke.³³ Receiving more education is associated with more employment options which potentially leads to less job and income-related stress, improved working conditions, income, and access to healthcare.³⁴ Finally, education may be a more important predictor of outcomes for older ages (>65 years), which are also the ages at which people develop hypertension.³⁵

A theoretical concern exists in that each of the social vulnerabilities included in our analysis may not be “unique.” That is, overlap in their causes and consequences may exist. For example, having low income and living around people who are living in poverty are distinct but potentially correlated variables. Despite this theoretical concern, included vulnerabilities demonstrated weak correlations. All variables had a correlation coefficient <0.3 and only educational attainment and income level had a correlation coefficient >0.2.

Factors that contribute to Black-White differences in hypertension are multifactorial, with potentially both social and biological underpinnings. Social determinants include the effects of daily experiences of discrimination and racism,^{36, 37} distrust in the health care system due to historical atrocities,^{38, 39} and structural segregation into neighborhoods with poor social economic conditions.⁴⁰ Communities of individuals with low socioeconomic status have fewer healthcare providers and specialists, often of lower quality; lower access to healthy food and higher dietary sodium to potassium ratio; lower prevalence of health insurance; and lower paying jobs and the need to have multiple jobs, all of which makes it difficult to have a healthy lifestyle, avoid poor health behaviors, and attend healthcare visits. There are also proposed biologic effects of race, such as a genetic predisposition to salt sensitivity.⁴¹ For these reasons, we chose to examine race, fundamentally a social construct but with wide reaching consequences, in a more nuanced fashion than solely as another social determinant.

We examined race as both a covariate in the full sample and in stratified analyses. The combined sample (i.e., White and Black participants together) allowed us to test for statistical differences in the association between multiple social vulnerabilities and the development of hypertension or dying irrespective of race. The stratified analysis allowed us to examine differences in these relationships by race. This revealed that the association between socially determined vulnerabilities was stronger in Whites than in Blacks. However, as in prior analyses, Black participants in the current analysis experienced a higher absolute risk of both hypertension and death than did White participants, regardless of the blood pressure threshold used.^{42, 43} And, it is possible that the effects of the social vulnerabilities were masked by stronger, unmeasured determinants of hypertension and life expectancy.^{44, 45} For example, we could not directly account for the effect of repeated or chronic stressors in daily life (allostatic load) that result from structural racism, microaggressions, or discrimination uniquely experienced by Black members of our society as a function of their skin color, hair texture, and facial features.⁴⁶ Allostatic load alone at least partially explains Black-White disparities in mortality, even after adjusting for education level, income level, and health insurance status.⁴⁵

Also, there were differences regarding within-race variability of the distribution of baseline social vulnerabilities used in this analysis. For example, there was little variability in the baseline distributions among White participants with 90% of the group in the 0 or 1 vulnerability and <10% experiencing 2 or more (Table 2). Conversely, Black participants were approximately evenly distributed across these categories with one third of the group in each category. The observations in the current analysis may be due to a less confounded pathway in White individuals of social vulnerabilities leading to poorer health outcomes, and therefore, easier to detect. Nevertheless, in the current analysis, Black individuals had nearly a seven-percentage point higher incidence of hypertension and five-percentage point higher mortality than White individuals. Finally, and possibly most importantly, a greater number of socially determined vulnerabilities conferred a strikingly greater risk of death among both Blacks and Whites. While it is possible that the high risk of death may distort our ability to evaluate social vulnerabilities as a mechanism of the development of hypertension, clearly the extraordinary risk of death associated with a greater number of socially determined vulnerabilities demands urgent attention. This is especially concerning because the greatest risk of death was observed among those who lack health insurance, a finding that should inform policy.

We observed in the current analysis that the associations of multiple vulnerabilities and incident hypertension and death varied statistically by Black and White race when using the 2017 ACC/AHA hypertension definitions, but not when using the less stringent JNC-7 definitions. One hypothesis is that this may be due in part to the categorization of the exposure variable. For example, we included the variable “Lives in a state with poor public health infrastructure” in the JNC-7 cohort, but not in the 2017 ACC/AHA cohort, because this variable was significantly associated with hypertension development and death with a p-value of <0.10 only in the JNC-7 cohort.

A major strength of this analysis is the REGARDS cohort, a large biracial geographically dispersed study population. Detailed and systematic assessment of blood pressure, medication use, and social and clinical risk factors for hypertension were obtained at baseline and after nearly ten years of follow-up. The multinomial analyses incorporating death revealed additional risks beyond hypertension, and the inverse probability weighted analysis accounted for lack of participation in the second in-home visit. Some limitations are also worth noting. Despite the availability of multiple covariates and a sophisticated approach to the analysis, the study is observational, limiting the ability to draw causal inferences. Some variables were self-reported with known biases,⁴⁷ and other critically important socially determined influences such as perceived racism and structural racism were not available. Consistent with prior analyses, some variables were included in our models as dummy variables while others were included as continuous variables. Blood pressure was assessed rigorously but at a single point in time. Longer term influences beyond ten years should also be studied.

PERSPECTIVE

The current study suggests that a greater number of socially determined vulnerabilities is associated with a progressively higher risk of developing hypertension, and an even higher risk of dying over 10 years. These findings were observed regardless of the blood pressure threshold used to define hypertension. The relative association of multiple social vulnerabilities on hypertension development may be stronger in White adults than Black adults, particularly when using hypertension thresholds defined by the 2017 ACC/AHA guidelines. Nonetheless, Black adults experience the highest absolute rates of hypertension, and also experienced a very high, and concerning, risk of death attributable to their burden of social determinants. In order to optimize the health of the overall population and reduce inequities, policy makers, payers, and population health managers must choose which policies and interventions to test and implement, and who should receive them. Taken together, our current study suggests that policies and interventions targeting social vulnerability, specifically in populations with multiple vulnerabilities, should be developed, tested, and, if effective, implemented to reduce both mortality and the incidence of hypertension.

NOVELTY AND SIGNIFICANCE.

What is new?

• The current analysis is one of the first to document a graded association of multiple social vulnerabilities within the same individual and a greater risk of developing hypertension or dying over 10 years.

• The positive association between the number of social vulnerabilities and developing hypertension or dying was stronger in White adults than in Black adults when using the 2017 ACC/AHA hypertension definitions.

What is relevant?

• A count of social vulnerabilities to health disparities is an important indicator of an individual who is at increased risk of developing hypertension and of dying prematurely.

• The modest association of multiple vulnerabilities observed among Black participants may be due to stronger unmeasured determinants of hypertension such as structural racism.

Summary

Among REGARDS participants without hypertension, simultaneously occurring social vulnerabilities to health disparities were associated with a higher risk of developing hypertension or dying over 10 years. Counts of social vulnerabilities could be used to identify high risk adults for efforts to prevent incident hypertension and death.