Association between neighborhood social cohesion, awareness of chronic diseases, and participation in healthy behaviors in a community cohort

We conducted a cross-sectional study examining the relationship between individual perception of neighborhood social cohesion, awareness of health conditions, and participation in healthy behaviors among individuals who participated in the prospective longitudinal Healthy Aging in Neighborhoods of Diversity Across the Life Span (HANDLS) study, taking place between 2013 and 2017. To identify the interaction of race and socioeconomic status on the development of cardiovascular and cerebrovascular health disparities, the HANDLS study recruited community-dwelling African Americans and Whites aged 30–64 years from an area probability sample of 13 neighborhoods representing contiguous U.S. census tracts in Baltimore City. Based on 2000 census data, these neighborhoods were thought to yield sufficient individuals to recruit a minimum of 30 participants per cell defined by race (African American, White) socioeconomic status (above and below 125% of the 2004 Health and Human Services Poverty guidelines), age (seven 5-year age groups) and sex (male, female), necessary to detect a difference in cardiovascular outcomes with 80% power after 20 years. HANDLS participants were recruited in two phases: household recruitment and interview followed by examination in medical research vehicles belonging to the study. Detailed methods for recruitment in HANDLS have been reported elsewhere [30].

Overall, 3720 individuals participated in a Wave 1 (baseline) study visit between August 2004 and March 2009. Wave 4 visits occurred between September 2013 and September 2017 and consisted of in-person health examination, a telephonic dietary recall, renal function assessment, and optional participation in a telephone-based survey. Nearly 58% of Wave 1 participants (n = 2171/3720) had a Wave 4 visit, during which most data were collected for this ancillary study, including the primary predictor. This study population excluded individuals missing serum creatinine and urine albuminuria values (n = 71), and those with an estimated glomerular filtration rate (eGFR) < 15 ml/min/1.73m² (n = 18) at Wave 4, for a final study population was 2082. National Institutes of Health Institutional Review Board (#09AGN248), approved the study protocol, as did the University of California, San Francisco Institutional Review Board (#10–02885). All participants provided written, informed consent to participate.

The primary explanatory variable was individual perception of neighborhood social cohesion, based on a validated questionnaire of 5 questions asking about unique neighborhood characteristics [17]. Individuals were asked to agree or disagree with the following statements using a 1–5 Likert scale in which 1 represented strong disagreement and 5 represented strong agreement: (1) “People in my neighborhood are willing to help their neighbors;” (2) “I live in a close-knit neighborhood;” (3) “People in my neighborhood can be trusted;” (4) “People in my neighborhood generally do not get along with each other;” and (5) People in my neighborhood do not share the same values.” Consistent with prior research, a composite neighborhood social cohesion score was calculated as the mean of the 5 items with items (4) and (5) reverse coded, with higher values representing more agreement that there was social cohesion [31, 32].

Awareness outcomes included awareness of kidney disease, diabetes, and hypertension, defined by accurate self-report of each of these chronic conditions during the Wave 4 health questionnaire compared to study measures of eGFR, fasting glucose/use of antidiabetic medications, and BP/use of anti-hypertensive medications [33]. Behavioral outcomes were defined by self-reported participation in healthy behaviors: no cigarette use, higher physical activity defined by the Baecke physical activity questionnaire [34], and heathy eating defined by the 2010 Healthy Eating Index [35].

Covariates included self-reported demographic information (age, sex, race, educational attainment, health insurance, having a regular healthcare provider, annual household income, and co-morbid conditions) obtained during Wave 1. Health literacy was measured with the hort Test of Functional Health Literacy in Adults (S-TOFHLA) using a cutoff score of 60 to differentiate between individuals with adequate vs. inadequate literacy levels [36]. Physical exam measures included blood pressure, height, and weight. Each participant underwent sitting and standing blood pressure measurements on each arm using the brachial artery auscultation method with an inflatable cuff of appropriate size. Laboratory measures included serum creatinine, urine microalbumin, serum glycated hemoglobin, and fasting glucose. Diabetes was defined by a fasting glucose > 126 ml/dl or use of an antidiabetic medication. Hypertension was defined by an uncontrolled measured BP defined by an average seated study systolic blood pressure > 140 mmHg or an average seated study diastolic BP > 90 mmHg or use of an anti-hypertensive medication. CKD was defined by single values of eGFR < 60 ml/min/1.73m² calculated by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation or the presence a urine microalbumin-to-creatinine ratio (UACR) > 30 mg/g [37, 38].

Characteristics of participants were compared by tertile of perceived social cohesion using χ² and ANOVA tests as appropriate. Multivariable logistic regression models were used to determine the presence, direction, strength and independence of an association between mean social cohesion score (modeled as a continuous variable) and awareness of each chronic condition (CKD, diabetes, hypertension) and cigarette use (all binary variables). Multivariable linear regression models were used to determine the association of mean social cohesion score with physical activity and healthy eating scores modeled as continuous variables. All models represented complete case analyses and included variables that were either determined a priori based on prior literature or statistically associated with tertile of social cohesion. Model 1 included age, sex, and race. Model 2 further adjusted for additional demographic variables (educational attainment, health insurance status, and having a regular healthcare provider). Model 3 further adjusted for clinical variables (diabetes status, hypertension status, eGFR, and UACR) when appropriate (i.e., models examining diabetes awareness did not include diabetes status; models examining hypertension awareness did not include hypertension status). Interaction between race and neighborhood social cohesion was assessed in each fully adjusted model. All statistical analyses were performed using STATA software, Version 14.2 (StataCorp, College Station, Texas, USA).

Among the study population of 2082 individuals, mean (SD) age was 56.5 (9.1) years, 40.9% were men, 38.7% were White, and 61.4% were African American. Mean NSC for the entire cohort was 3.3 (SD = 0.80) with a mean score in the first NSC tertile of 2.55 (standard deviation [SD] = 0.51), a mean score of 3.40 (SD = 0.16) in the second NSC tertile, and a mean score of 4.23 (SD = 0.38) in the third NSC tertile. Among the entire cohort, 32.3% did not complete their high school education and 12.7% had low health literacy, percentages that did not differ by social cohesion tertile. Approximately 39% of the cohort lived below the 125% federal poverty level with fewer individuals living below the poverty level among those who self-reported higher social cohesion (p < 0.001). Nearly 65% of participants reported having health insurance and 68.8% cited a regular source of health care, with higher percentages of individuals reporting either characteristic with increasing social cohesion score (p = 0.002 and p = 0.04, respectively). Of the overall cohort, 26.3% were diabetic and 70% had hypertension. Approximately 58.9% of individuals had CKD stage 3 or 4 and 13.9% had urine albuminuria levels > 30 mg/g of creatinine (Table 1).

Wave 4 participants were generally similar to those who completed the baseline study visit. However, Wave 4 participants were more likely to be female compared to those participants who were lost to follow-up (58.6% vs. 49.2%, p < 0.01) and more likely to be African American (61.1% vs. 56.3%, p = 0.003). Education, poverty status and insurance were not different among the two groups (data not shown, p > 0.05). Of the Wave 4 study population, 1918 individuals answered questions about social cohesion. While individuals who did not complete the NSC questionnaire were generally similar to those who did, individuals missing NSC scores were less likely to be a college graduate, more likely to have low health literacy, and more likely to have diagnosed hypertension and diabetes. (Supplemental Table).

Overall, 18.7% (86/460) of individuals with CKD were aware of their disease. Among those aware of their CKD, mean NSC score was 3.32 (SD = 0.73) with similar prevalence of awareness by tertile of neighborhood social cohesion (Fig. 1). Nearly 90% of individuals with diabetes (89.7%, 330/368) were aware of their condition. Of those aware, mean NSC score was 3.20 (SD = 0.78). A similar, high percentage of individuals (90.7%, n = 936/1032) reported awareness of hypertension with a similar distribution of awareness by neighborhood social cohesion scores; mean NSC among those aware was 3.28 (SD = 0.79).

Using multivariable logistic regression, increasing mean social cohesion score was not significantly associated with awareness of CKD (adjusted odds ratio [aOR] = 1.13; 95% Confidence Interval = 0.81, 1.56), awareness of diabetes (aOR = 0.94; 0.57–1.55) or awareness of hypertension (aOR = 1.15; 0.87–1.51). None of the associations between NSC and awareness differed by race (P_{interaction NSC x race} > 0.10).

Neighborhood social cohesion scores were independently associated with many healthy lifestyle habits and often differed by race (Table 2). For example, there were 13.1% decreased odds of cigarette use with every one-point gain in self-reported mean social cohesion score (aOR = 0.87; 0.76, 0.99), an association largely driven among White participants (P_{interaction NSC x race} = 0.08). Among Whites, odds of cigarette use were 24% lower with each higher point in NSC (aOR = 0.76; 0.61–0.94) while there was no association between cigarette use and NSC score among African Americans (aOR = 0.95; 0.79–1.13). Similarly, every one-point increase in mean NSC score was associated with an increase in healthy eating engagement according to the Healthy Eating Index, (beta coefficient [βc]: 1.08; 95% Confidence Interval = 0.34–1.82), largely driven by White participants (P_{interaction NSC x race} = 0.06). Among Whites, every one-point increase in mean NSC was associated with an increase of healthy eating engagement score (βc: 1.75; 0.55–2.97) while the association was not statistically significant among African Americans (βc: 0.46; − 0.48–1.39). Higher neighborhood social cohesion scores were associated with increased physical activity (βc: 0.12; 0.08–0.16) without any difference by race (P_{interaction NSC x race} = 0.40).

Other sociodemographic characteristics were also independently associated with participation in healthy behaviors. As depicted in Table 2, younger age was associated with higher odds of cigarette use (aOR = 0.97; 0.96–0.98) and less physical activity (βc: − 0.005; − 0.01–0.001) whereas older age was associated with higher report of healthy eating habits (βc: 0.23; 0.12–0.31). Having a greater than high school education was uniformly helpful, associated with lesser odds of cigarette use (aOR = 0.58; 0.47–0.73), greater healthy eating (βc: 3.23; 1.95–4.51), and more physical activity (βc: 0.11; 0.04–0.19). Poverty, on the other hand, was associated with greater odds of cigarette use (aOR = 1.51; 1.21–1.87) and non-statistically associated with less healthy eating (βc: − 2.45; − 3.68 – − 1.23) and less physical activity (βc: − 0.08; − 0.15–0.01). These values are stratified by race in Table 2.