Social Networks and Cardiovascular Disease Events in the Jackson Heart Study

Abstract

Background

Cardiovascular disease (CVD) disproportionately affects Black adults. Greater social networks (SNs), or social connectedness, may lower the risk of CVD events. This study determined the association of SNs and incident CVD and tested mediation by depressive symptoms, hypertension control, and diabetes control.

Methods and Results

We used the Social Network Index at exam 1 (2000–2004) to develop a continuous standardized SN score and binary categories (high versus low) among participants in the Jackson Heart Study (n=4686; mean age, 54.8 years). Surveillance of coronary heart disease, stroke, and heart failure events occurred after exam 1 (2005 for HF) until 2016. Using Cox proportional hazards regression, we estimated the association of SNs and CVD events by sex and tested the mediation of depressive symptoms, hypertension control, and diabetes control. Models adjusted for age, education, health behaviors, CVD comorbidities, and depressive symptoms. Among women, the SN score was associated with a lower hazard of stroke, coronary heart disease, and heart failure after full adjustment (hazard ratio [HR], 0.78 [95% CI, 0.64–0.95]; HR, 0.79 [95% CI, 0.71–0.88]; and HR, 0.78 [95% CI, 0.66–0.92], respectively). SN scores were also associated with a lower hazard of coronary heart disease in men (HR, 0.84 [95% CI, 0.75–0.94]) after full adjustment. High versus low SNs were associated with a lower hazard of coronary heart disease and heart failure among women after full adjustment. There was no evidence of mediation by depressive symptoms, diabetes control, and hypertension control.

Conclusions

Higher SNs may lower the risk of CVD events, especially in women.

Clinical Perspective.

What Is New?

• Social connectedness is associated with lower risk of coronary heart disease among Black adults, and health behaviors such as smoking may partially explain the association.

• Control of risk factors such as hypertension and diabetes did not mediate the association between social connectedness and heart disease.

What Are the Clinical Implications?

• Increasing social connections may benefit women's risk of coronary heart disease, stroke, and heart failure.

Nonstandard Abbreviations and Acronyms

ARIC

Atherosclerosis Risk in Communities

JHS

Jackson Heart Study

SN

social network

SNI

Social Network Index

WISE

Women's Ischemia Syndrome Evaluation

Every year in the United States, >600 000 deaths are attributed to cardiovascular disease (CVD), which equates to 1 in 4 deaths. ¹ Age‐adjusted death rates for heart disease differ by race and ethnicity, where non‐Hispanic Black adults are more likely to have a higher rate than non‐Hispanic White, Hispanic, and Asian or Pacific Islander adults. ¹ Additionally, the burden of CVD among Black adults is estimated at 9.5%, and the burden of risk factors for CVD, such as hypertension and diabetes, are more common among Black adults than other racial and ethnic groups. ²

Social isolation, or having few or limited social connections with others, is an upstream social factor associated with inflammation, high cholesterol, and a greater risk of a CVD event. ³ , ⁴ , ⁵ Although social isolation is related to a greater risk of CVD, social integration, or having social connections, is associated with a lower risk of CVD, even without examining the quality of these interactions or examining social support. ⁶ Social networks (SNs), or the social connectedness related to the number, density, and characteristics of family and community connections, prevent social isolation and may moderate and improve disease risk through a variety of mechanisms. According to Havranek et al, ⁷ an SN may influence health in 2 ways: (1) through social influence on behaviors and (2) through resources embedded in an SN that are available to its members. Having a larger and positive SN could also moderate disease risk by reducing certain negative health behaviors such as smoking, by directly or indirectly improving immune system responses, by offering socioeconomic benefits, and by providing greater social support to reduce emotional distress. ⁸ Although previous reports have found that cigarette smoking and physical activity are in the pathway of SNs and coronary heart disease (CHD), ⁶ it is uncertain whether other factors strongly associated with CVD are in the pathway, such as having better control of risk factors such as hypertension ⁹ and diabetes, ¹⁰ and low depressive symptoms. ¹¹ These key risk factors for CVD may be better managed among those who are better connected and may be the reason for connectedness lowering the risk of CVD.

Social connectedness may differ by race and ethnicity, and sex. For instance, 1 study found that Black adults are more likely to be involved in religious organizations when compared with other ethnic groups. ¹² Additionally, Black, compared with White, adults tend to have smaller SNs, more family members in their SNs, and more frequent contact with their family members. ¹³ Differences in the effect of SNs by sex have also been reported. One study by Pantrell et al ¹⁴ reported slight differences in mortality by SNs among men and women, with infrequent social contact being more influential for survival among women than among men. Due to these differences, there is a need to interrogate how SNs relate to CVD end points among Black men and women to better understand whether improving connectedness is a valid intervention technique to lower CVD burden and determine who would benefit most from such interventions.

In this study, SNs consisted of 5 dimensions of social connectedness: marital status, close friends, closeness to relatives, visiting close friends and relatives, and group participation. We tested the association of SNs with CVD events (ie, stroke, CHD, and heart failure [HF]) by sex and examined the extent to which depressive symptoms and hypertension and diabetes control mediated the association of SNs with CVD events. Our guiding hypothesis was that there will be an inverse association between SNs and incident CVD, with differences among men and women. We also hypothesized that management or control of hypertension and diabetes, and having low depressive symptoms, will mediate the association among Black men and women in the JHS (Jackson Heart Study).

Methods

The data used for this study can be requested for purposes of reproducing results. Request to access this data set (or other data in the JHS) may be directed to the qualified researchers trained in human subject confidentiality within the JHS Coordinating Center at jhsccrc@umc.edu.

Participants

The JHS is a prospective and longitudinal cohort study of 5306 (3371 women, 1935 men; 20–95 years of age) Black adults from the tricounty area (Hinds, Madison, and Rankin) of Jackson, Mississippi. The purpose of the study was to determine risk factors for CVD. Eligible participants were enrolled at baseline (2000–2004) and followed annually and at each exam (exam 2 [2005–2008] and exam 3 [2009–2013]). Participants were sampled from the ARIC (Atherosclerosis Risk in Communities) study (30%), the Mississippi Department of Transportation Driver's License and Identification List (17%), volunteers (22%), or family members of those who had already agreed to be a part of the study. ¹⁵ , ¹⁶ Participants provided demographic, socioeconomic, psychosocial, health history, and clinical data from in‐home interviews, self‐administered questionnaires, and in‐clinic examinations at each exam visit. Annual follow‐up data and surveillance of CVD events were conducted by trained research staff. ¹⁷ All JHS participants provided written informed consent, and all study protocols conformed to the 1975 Declaration of Helsinki guidelines. The JHS was also approved by the institutional review boards of the following participating institutions: the University of Mississippi Medical Center, Jackson State University, and Tougaloo College.

Social Networks

We defined SNs using 5 measures based on the definition of social connection and ties from the Berkman and Syme Social Networks Index (SNI): marital status, number of close friends, number of close relatives, number of friends or relatives visited monthly, and belonging to social groups. ¹⁸ For instance, participants were asked: How many relatives do you feel you are close to? How many of these friends or relatives do you see at least once per month? We generated a continuous SN score based on the 5 items (range, 0–5) and standardized the score using standard deviation units. To observe threshold effects, we also dichotomized the score as high and low (above and below the median, respectively). The SNI has not been validated in the JHS; however, it has previously been used successfully in other large cohort studies. ¹⁹ , ²⁰

Mediators

We considered greater hypertension control, diabetes control, and low (versus high) depressive symptoms as possible mediators in the association between SNs and CVD events. Among those with hypertension at exam 1, hypertension control was defined as having a measured blood pressure <140/90 mm Hg among those treated for hypertension at exam 1. Among those with type 2 diabetes at exam 1, diabetes control was defined as having measured hemoglobin A1c (HbA1c) levels <7.0%. We used the Center for Epidemiologic Studies‐Depression Scale to define depressive symptoms, which counts the number of clinically significant depressive symptoms experienced during a given week (range, 0–60). As recommended, participants who scored ≥16 were classified as having high depressive symptoms, whereas those who scored <16 had low depressive symptoms. ²¹ , ²²

CVD Events

Surveillance of CVD events began during baseline examination (except heart failure, which began in 2005). Surveillance data were available through December 31, 2016. The JHS medical record abstraction team contacted participants by phone to identify CVD events (diagnostic tests, hospitalizations, or death), and also used related records (discharge lists and death certificates) from hospitals and state offices for verification. ²³ , ²⁴

CHD was defined as definite or probable hospitalized myocardial infarction, a related death, an unrecognized myocardial infarction by ECG, or coronary revascularization based on combinations of chest pain symptoms, ECG changes, and cardiac enzyme levels. A fatal CHD event was classified as an underlying cause of death from the death certificate and other associated hospital information or medical history.

HF events were defined as the occurrence of either inpatient or outpatient diagnoses of unspecified failure of the heart from the International Classification of Diseases, Ninth Revision (ICD‐9) diagnosis code of 428.X in any position or 428 on a death certificate. The definition of HF also includes, but is not limited to, radiographic findings that were similar with congestive HF, increased venous pressure >16 mm Hg, dilated ventricle/left ventricular function <40% from an ECG or multiple gated acquisition, or autopsy finding of pulmonary edema.

Stroke events were defined as a definite or probable hospitalized stroke from neuroimaging and autopsy based on the classification from the National Survey of Stroke. ²⁵ The minimum criterion for probable or definite stroke was rapid onset of neurological symptoms lasting >24 hours or symptoms that led to death. Neurological symptoms lasting <24 hours or symptoms seen before or during admission to the hospital were not considered a definite or probable stroke. Only stroke events with a medical diagnosis were considered definite or probable. A computer algorithm was used to determine all CVD events, as well as physician reviews. Disagreements were adjudicated by another reviewer.

Covariates

To determine the inclusion of covariates relevant to the association between SNs and CVD, we used a directed acyclic graph (data not shown) based on previous studies of social isolation and heart disease. ²⁶ , ²⁷ Thus, we chose covariates that would control for confounding in the association between SNs and incident CVD events. These covariates included demographic, socioeconomic, behavioral, psychosocial, and traditional risk factors.

We included demographic variables such as age (years), self‐identified sex (male=men/female=women), and a marker of socioeconomic status, education (less than a high school diploma, some college, college degree or more) as covariates. Additionally, we included 2 American Heart Association Life's Simple 7 characteristics (classified as ideal versus nonideal): physical activity and smoking, which are traditional risk factors for CVD. ²⁸ Alcohol intake was defined by alcohol drinking in the past 12 months at baseline, and dietary fat was defined as daily percent fat from the Food Frequency Questionnaire. CVD comorbidities were also considered: hypertension status (derived variable that includes antihypertensive medication use, self‐reported physician diagnosis, and measured systolic and diastolic blood pressure), obesity (based on body mass index >30 kg/m²), and diabetes prevalence (derived diabetes status based on American Diabetes Association 2010, which included fasting glucose ≥126 mg/dL or confirmed medication use from the medication inventory, or self‐reported use of antidiabetic medications within the past 2 weeks of the examination, or self‐reported diabetes diagnosis). Depressive symptoms, also a risk factor for incident CVD ²⁹ and a factor that influences social connections, ²⁶ was measured using the Center for Epidemiologic Studies‐Depression Scale described previously.

Statistical Analysis

There were 5252 participants who had complete responses from the SNI at exam 1. Of the 5252, 566 participants were removed from the sample because they reported prevalent CVD at baseline, leaving 4686. Covariates with missing observations were the following: education (n=11), smoking (n=79), obesity (n=14), dietary fat (n=177), diabetes prevalence (n=63), and hypertension prevalence (n=2). Missing values for depressive symptoms were imputed to the mean depressive symptoms score (n=1667; mean, 10.8).

Participants excluded due to prevalent CVD and missing SN scores (n=620) compared with the analytic sample (data not shown) were older (62.8±10.7 years versus 53.9±12.8 years), were more likely to not have a high school diploma (38.9% versus 16.2%), were more likely to have nonideal physical activity (86.8% versus 80.1%), and were more likely to have nonideal smoking habits (19.9% versus 13.8%). However, alcohol intake was lower (35.8% versus 47.2%), and the dietary fat percentage was lower (33.7±6.7% versus 35.1±6.9%). CVD morbidities were higher among participants with CVD and missing SN scores compared with the analytic sample: hypertension (59.6% versus 53.2%), obesity (56.4% versus 53.3%), and diabetes (37.8% versus 21.4%). Of those with hypertension and diabetes, 61.8% (versus 23.5%) had hypertension control, and 36.1% (versus 39.3%) had diabetes control. Depressive symptoms scores were higher (before imputing) in the excluded sample (12.7±8.7 versus 10.8±8.0).

Descriptive statistics, specifically differences in select characteristics by sex, were calculated using χ² and ANOVA tests. Time‐to‐CVD event was approximated based on years from exam 1 to the adjudicated incident CVD event (CHD, stroke, HF). By December 2016, there were missing observations for each CVD outcome: CHD (n=733), HF (n=965), and stroke (n=550). Participants who had deaths unrelated to CVD (n=940) were censored. Cox proportional hazards regression was used to estimate the association of high versus low SNs (and the standardized SN score) with incident CVD, where hazard ratios (HRs) and 95% CI estimated the risk of CVD. Models were sequential in the adjustment of covariates, where Model 1 adjusted for age and education. Model 2 added physical activity, smoking, alcohol, and dietary fat. Model 3 added hypertension, obesity, diabetes, and depressive symptoms. The proportional hazards assumption was tested by examining Kaplan‐Meier curves and by examining the statistical significance of time‐dependent interaction terms.

Modification by sex was determined by evaluating the interaction term (sex×high SNs) in Model 3. A P value <0.05 was considered statistically significant, as well as confidence intervals without a value of 1.00.

We used causal mediation techniques as described by Pearl ³⁰ and Bind et al ³¹ to relax linearity and interaction assumptions found in the Baron and Kenny method in SAS 9.4 (PROC CAUSALMED) to evaluate mediation. Separate models were constructed to test mediation of hypertension control, diabetes control, and low (versus high) depressive symptoms, respectively in the association between high versus low SNs and each type of CVD event. Models 1 and 2 were used to adjust for confounders, whereas covariates in Model 3 were not included if the mediator was a covariate. For each potential mediator, we reported the total effect, direct effect, natural direct effect, and the natural indirect effect. Percent mediation was applicable when the natural indirect effect was statistically significant. A P value <0.05 was considered statistically significant, as well as confidence intervals excluding 1.00.

Results

Table 1 shows the descriptive characteristics of participants without CVD at baseline, stratified by low and high SN scores. SN scores were associated with age, education, smoking, alcohol, hypertension, hypertension control, and depressive symptoms. Those who had high (versus low) SN scores were more likely to be older (P<0.001), have at least some college or more (P<0.001), have ideal smoking habits (P<0.001), and were more likely to drink alcohol (P<0.001). Hypertension prevalence (P=0.004) and hypertension control (P=0.003) were also higher for those with high SNs. Depressive symptoms were lower for those who had high SNs (P<0.001).

Table 1.

Select Characteristics of Participants by Low and High Social Network Score: Jackson Heart Study (n=4686)

Characteristics	Low social networks (N=780)	High social networks (N=3906)	χ2/ANOVA P value
Age, y, mean±SD	50.30±14.30	54.57±12.36	<0.001
Men	272 (16.30)	1397 (83.70)	0.634
Women	508 (16.84)	2509 (83.16)
Education
Less than high school	204 (23.78)	654 (76.22)	<0.001
High school diploma	271 (19.54)	1528 (80.46)
Some college or more	203 (10.59)	1713 (89.41)
Smoking
Ideal	591 (14.83)	3395 (85.17)	<0.001
Nonideal	174 (27.23)	465 (72.77)
Physical activity
Ideal	144 (15.45)	788 (84.55)	0.274
Nonideal	636 (16.94)	3118 (83.06)
Alcohol
Yes	426 (19.38)	1772 (80.62)	<0.001
No	347 (14.10)	2114 (85.90)
Percent dietary fat, mean±SD	35.03±7.28	35.16±6.85	0.638
Hypertension	378 (15.17)	2114 (84.83)	0.004
Diabetes	169 (17.09)	820 (82.91)	0.689
Obesity	426 (17.09)	2066 (82.91)	0.352
Hypertension control	77 (13.49)	494 (86.51)	0.003
Diabetes control	41 (17.52)	193 (82.48)	0.696
Depressive symptoms, mean±SD	13.38±9.99	10.30±7.47	<0.001

Data reported as N (%) unless otherwise indicated. Participants without a social networks score (n=54) and those with cardiovascular disease history (n=566) at baseline were excluded. The following variables had missing observations: education (n=11), smoking (n=79), obesity (n=14), dietary fat (n=177), diabetes prevalence (n=63), hypertension (n=2), hypertension control (n=339), and diabetes control (n=487).

There were 239 CHD events, 184 stroke events, and 261 HF events over a median follow‐up of ≈8 years (data not shown). Table 2 presents the unadjusted cumulative incidence of CVD events by sex. Among women, the cumulative incidence was 5.0% for stroke, 5.6% for CHD, and 7.1% for HF. Among men, the cumulative incidence was 5.6% for stroke, 7.5% for CHD, and 7.4% for HF. The cumulative incidence of HF differed by high and low SNs among women. The 10‐year unadjusted cumulative incidence of HF was 6.6% (95% CI, 5.6%–7.8%) for women with a high SN score, whereas the cumulative incidence of HF was 10% (95% CI, 7.1%–13.5%) for women with a low SN score (Gray test, P=0.0258). The cumulative incidence of HF did not differ by SN score among men (Gray test, P=0.2933) (Figure). Differences in the cumulative incidence of CHD and stroke by SN score and sex were not statistically significant (Figures S1 and S2).

Table 2.

Unadjusted Cumulative Incidence (95% CI) of Cardiovascular Disease Events by Sex

	Women (n=2976)	Men (n=1653)	P value
Stroke
Events	113	71	0.2698
Cumulative incidence	5.0% (4.0%–6.2%)	5.6% (4.2%–7.2%)
CHD
Events	138	101	0.0156
Cumulative incidence	5.6% (4.7%–6.7%)	7.5% (6.0%–9.1%)
HF
Events	168	93	0.7998
Cumulative incidence	7.1% (6.1%–8.2%)	7.4% (6.0%–9.0%)

Median follow‐up time was 10.99 years for HF, 12.86 years for CHD, and 12.86 years for stroke.

CHD indicates coronary heart disease; and HF, heart failure.

Figure 1. Kaplan‐Meier curve for heart failure among women (top) and men (bottom), with high and low social networks.

The 10‐year unadjusted cumulative incidence among women with high social networks was 6.6% (95% CI, 5.6%–7.8%). The 10‐year unadjusted cumulative incidence among women with low social networks was 10% (95% CI, 7.1%–13.5%). The 10‐year unadjusted cumulative incidence among men with high social networks was 7.1% (95% CI, 5.6%–8.7%). The 10‐year unadjusted cumulative incidence among men with low social networks was 9.3% (95% CI, 5.6%–14.1%).

Table 3 shows the association between SNs and CVD events stratified by sex. Among women, a 1‐SD‐unit increase in the SN score was associated with a 0.78 (95% CI, 0.64–0.95) hazard of stroke after full adjustment. High versus low SNs were not associated with stroke in women or men. Among women and men, a high versus low SN score was associated with a lower hazard of CHD after full adjustment (HR, 0.65 [95% CI, 0.48–0.87] and HR, 0.71 [95% CI, 0.51–0.98], respectively) (Table 3). Among both men and women, a 1‐SD‐unit increase in SN score was associated with a lower hazard of CHD after full adjustment (women: HR, 0.79 [95% CI, 0.71–0.88]; men: HR, 0.84 [95% CI, 0.75–0.94]). Among women and men, a high versus low SN score was associated with a lower hazard for HF in Model 2 (women: HR, 0.62 [95% CI, 0.42–0.91]; men: HR, 0.56 [95% CI, 0.33–0.98]). The association between high (versus low) SNs remained significant in Model 3 for women (HR, 0.60 [95% CI, 0.41–0.89]), but not for men (HR, 0.68 [95% CI, 0.38–1.21]). Additionally, a 1‐SD‐unit increase in the SN score was associated with a lower hazard of HF among women (HR, 0.78 [95% CI, 0.66–0.92]) after full adjustment.

Table 3.

Hazard Ratios (HR, 95% CI) of Cardiovascular Disease Events by Social Networks, stratified by sex: Jackson Heart Study (2000–2016)

Stroke
	Women			Men
	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
Social networks
Low (referent)	1.0	1.0	1.0	1.0	1.0	1.0
High	0.66 (0.41, 1.04)	0.64 (0.40, 1.03)	0.61 (0.38, 1.00)	1.37 (0.62, 3.02)	1.73 (0.74, 4.05)	1.71 (0.72, 4.03)
SD units	0.80 (0.66, 0.97)	0.80 (0.65, 0.97)	0.78 (0.64, 0.95)	0.88 (0.70, 1.10)	0.96 (0.75, 1.22)	0.95 (0.75, 1.20)
Coronary heart disease
Social networks
Low (referent)	1.0	1.0	1.0	1.0	1.0	1.0
High	0.62 (0.50, 0.79)	0.66 (0.52, 0.84)	0.69 (0.54, 0.89)	0.65 (0.48, 0.87)	0.69 (0.50, 0.94)	0.71 (0.51, 0.98)
SD units	0.76 (0.69, 0.84)	0.78 (0.71, 0.87)	0.79 (0.71, 0.88)	0.80 (0.72, 0.88)	0.83 (0.74, 0.92)	0.84 (0.75, 0.94)
Heart failure
Social networks
Low (referent)	1.0	1.0	1.0	1.0	1.0	1.0*
High	0.59 (0.40, 0.86)	0.62 (0.42, 0.91)	0.60 (0.41, 0.89)	0.58 (0.34, 1.00)	0.56 (0.33, 0.98)	0.68 (0.38, 1.21)
SD units	0.78 (0.67, 0.92)	0.80 (0.68, 0.94)	0.78 (0.66, 0.92)	0.83 (0.68, 1.00)	0.82 (0.67, 0.99)	0.87 (0.71, 1.06)

Model 1‐age, education.

Model 2‐ Model 1+ physical activity, alcohol, smoking, dietary fat.

Model 3‐ Model 2+ hypertension, obesity, diabetes, depressive symptoms.

Abbreviations HR, hazard ratio; CI, confidence interval; SD, standard deviation.

Boldface = significant p<0.05.

Tables 4 and 5 show the results by sex of mediation tests in the association of SNs and CHD and SNs and HF (stroke did not have a significant main effect). Natural indirect effects were not statistically significant for any of the mediators in the association of high (versus low) SNs and CHD or SNs and HF for men or women. We additionally explored mediation of ideal smoking, ideal physical activity, and ideal nutrition, and only found marginally significant (P<0.10) indirect effects for ideal smoking in the association of SNs and CHD events for men and women (Tables S1 and S2).

Table 4.

Effects from Mediation Tests (HR 95% CI) of Hypertension control, Diabetes Control and Depressive Symptoms in the Association of High (vs. low) Social Network Score and CHD and HF Among Women, Jackson Heart Study

	CHD events=138
	Hypertension control	Diabetes control	Low depressive symptoms
Total effect	0.79 (0.21, 1.37)	0.88 (0.29, 1.47)	0.93 (0.58, 1.28)
Natural direct effect	0.78 (0.21, 1.36)	0.84 (0.29, 1.44)	0.94 (0.59, 1.29)
Natural indirect effect	1.01 (0.98, 1.04)	1.01 (0.97, 1.05)	0.99 (0.97, 1.01)
Percent mediation	N/A	N/A	N/A

	HF events=168
	Hypertension control	Diabetes control	Low depressive symptoms
Total effect	0.56 (0.21, 0.92)	0.46 (0.17, 0.74)	0.57 (0.39, 0.75)
Natural direct effect	0.56 (0.21, 0.91)	0.44 (0.17, 0.71)	0.57 (0.39, 0.75)
Natural indirect effect	1.01 (0.97, 1.04)	1.04 (0.96, 1.12)	1.00 (0.98, 1.01)
Percent mediation	N/A	N/A	N/A

Model adjusted for: age, education, physical activity, alcohol, smoking, dietary fat, hypertension (except in hypertension control model), obesity, and diabetes (except in diabetes control model), and depressive symptoms (except in depressive symptoms model). Boldface=statistically significant indirect effect p<0.05.

CHD indicates coronary heart disease; CI, confidence interval; HF, heart failure; HR, Hazard Ratio and N/A, not applicable.

Table 5.

Effects from Mediation Tests (HR 95% CI) of Hypertension Control, Diabetes Control and Depressive Symptoms in the Association of High (vs. low) Social Network Score and CHD and HF Among Men, Jackson Heart Study

	CHD events=101
	Hypertension control	Diabetes control	Low depressive symptoms
Total effect	0.83 (0.16, 1.50)	0.87 (0.28, 1.46)	0.93 (0.58, 1.28)
Natural direct effect	0.81 (0.17, 1.45)	0.86 (0.27, 1.43)	0.94 (0.59, 1.29)
Natural indirect effect	1.01 (0.98, 1.04)	1.01 (0.97, 1.05)	0.99 (0.98, 1.01)
Percent mediation	N/A	N/A	N/A

	HF events=93
	Hypertension control	Diabetes control	Low depressive symptoms
Total effect	0.56 (0.20, 0.93)	0.45 (0.17, 0.73)	0.57 (0.39, 0.75)
Natural direct effect	0.56 (0.21, 0.91)	0.43 (0.16, 0.70)	0.57 (0.39, 0.75)
Natural indirect effect	1.01 (0.97, 1.04)	1.04 (0.96, 1.12)	1.00 (0.99, 1.01)
Percent mediation	N/A	N/A	N/A

Model adjusted for: age, education, physical activity, alcohol, smoking, dietary fat, hypertension (except in hypertension control model), obesity, and diabetes (except in diabetes control model), and depressive symptoms (except in depressive symptoms model). Boldface=statistically significant indirect effect P<0.05.

CHD indicates coronary heart disease; CI, confidence interval; HF, heart failure; HR, Hazard Ratio and N/A, not applicable.

Discussion

This study examined the association between SNs and incident CHD, stroke, and HF among Black men and women enrolled in the JHS. In addition, we explored mediation of hypertension control, diabetes control, and depressive symptoms in the association between high (versus low) SNs and incident CHD, stroke, and HF. We found partial support for our hypotheses. There were stronger inverse associations in women for SNs and incidence of each CVD outcome when examining the SNs’ SD unit score. The association between high (versus low) SNs and stroke was not as robust as the association between high (versus low) SNs and CHD and HF among women. When examining potential explanations for significant associations, we found that better control of hypertension and diabetes and lower depressive symptomatology were not mediators in the association of social networks and CHD or HF. We did see marginally significant evidence of smoking being a mediator in the association of SNs and CHD for both men and women.

SNI and Incident CVD

The inverse association between social connections and incident CVD reported in this study is consistent with previous reports. ³² , ³³ Rutledge et al ⁸ found a significant inverse association between SNs and stroke among women from the WISE (Women's Ischemia Syndrome Evaluation) study. Similar to our study findings, the association attenuated after adjustment for CVD risk factors. Among women from the WISE study who had suspected coronary artery disease, a type of CHD, Rutledge et al ³⁴ also reported that women with higher SN scores showed a consistent pattern of reduced coronary artery disease. Chang et al ⁶ explored the association between SNs and CHD among women in the Nurses' Health Study and found a lower hazard of CHD among women who had higher SN scores. To our knowledge, studies have only evaluated the association of social isolation and HF and found that greater isolation increased the risk of HF. ²⁶ , ³³ The associations between SNs and stroke among women was explained by greater social support from greater SNs. ⁸ Lower coronary artery disease risk by greater SN score was partially explained by income, but authors noted that the mediating effect was unlikely due to monetary gifts from networks but greater overall socioeconomic status, which potentially increases skills, access to treatments, knowledge, and support. ³⁴ Smoking was a mediator in the association between social integration and CHD risk in the Nurses' Health Study. ⁶

Although most studies have reported associations between SNs and incident CVD among women only, our results highlight sex differences, with stronger associations noted for women than men. Our findings indicate the effect of SNs on CVD is different when comparing women with men, demonstrating that women may gain more cardiovascular health benefits from SNs. We saw no significant differences by sex when examining SN scores; however, based on the previous studies conducted among women, ⁶ , ³⁴ SNs may provide better social support for women than men, which offers protection against CVD. It is also possible that women who have greater SNs are better equipped to handle stressors than men. ⁶

Mediation Analyses

We explored whether having better control of cardiovascular risk factors (hypertension, diabetes, and depressive symptoms) mediated the association of SNs and CHD and HF but found no evidence of mediation by these factors. Studies have suggested that greater SNs improve health behaviors, such as smoking and physical activity, which reduces the risk of CVD. ⁶ Women from the JHS were less likely to be alcoholics and were more likely to have ideal smoking behaviors, but men were more likely to be physically active. Although we explored smoking, physical activity, and dietary habits as mediators, there were only marginally indirect effects observed for smoking. Perhaps SNs offer other positive benefits, such as lower psychosocial stress and greater tangible support, which leads to lower downstream CVD risk. Future work may need to evaluate psychosocial factors as potential mediators in the pathway between SNs and CVD.

Limitations and Strengths

The results of this study should be interpreted with consideration of limitations. One limitation is that the SNI does not evaluate the quality of social connections in all index measures. For instance, it is assumed that being married and belonging to social groups are positive social ties that provide benefits to the participants. Although there is literature to support that being married and having greater social ties is linked to lower risk of CVD and better health, ³⁵ , ³⁶ we could not guarantee that all participants have positive social connections. Another limitation was generalizability, which was limited due to the analytic sample only including Black participants from Jackson, Mississippi. Deaths unrelated to CVD were censored rather than dropped, but missing data are a limitation of this study, particularly for covariates such as depressive symptoms. Although missing values for depressive symptoms were imputed to the mean, true values were unknown and could have caused bias. We compared estimates of the association of SNs with incident CVD in the fully adjusted model using the complete cases for depressive symptoms and the mean imputed values (Table S3), and the results were similar. We also compared estimates when testing depressive symptoms as a mediator in the association of SNs and incident CVD (Table S4), and results were not statistically significant when using mean imputed values. Although this study has limitations, inclusion of a large sample of Black participants in a study of social connectedness and CVD is a strength, due to this racial group's underrepresentation in scientific studies. Additionally, this study used surveillance data to capture CVD events and conducted time‐to‐event analyses. The use of formal mediation tests for diabetes control, hypertension control, and low depressive symptoms is another strength of this study.

Conclusions

A positive association between SN scores and CVD events was stronger among women. The results align with other studies conducted in women, which has implications for future public health interventions seeking to improve CVD burden and incidence in populations. These results suggest that increasing social network size (eg, number of close friends, belonging to organizations/groups) can reduce the risk of stroke, CHD, and HF among women. Based on results from other studies, increasing social network size may also have implications for control of CVD. In clinical settings, patients could be asked about their SNs. Isolated patients should be encouraged to improve their social connectedness to improve their quality of life and prevent depression and illness. Depending on the patients' environment and circumstance, clinicians could encourage group fitness, community service organizations, or visiting or receiving visits from loved ones more often. Health care organizations could also create groups that focus on patient‐centered care, while also engaging participation from the patient, multiple physicians, coordinators, and other health care workers. ³⁷ Interventionalists who focus on improving diet and physical activity could also add SNs to their program to improve adherence.

Future work should continue to explore the extent to which psychosocial resilience measures, besides SNs, may be protective of CVD risk, especially for men. The role of SNs could also be better understood by investigating other factors in the pathway of SNs and a lower risk of CVD among Black adults.

Sources of Funding

Dr Glover was supported by the National Heart, Lung, and Blood Institute under award F31HL159910. The views expressed in this article are those of the authors and do not represent the views of the National Heart, Lung, and Blood Institute, the National Institutes of Health, or the US Department of Health and Human Services.

Disclosures

None.

Supporting information

Tables S1–S4

Figures S1–S2