Background The main strategy for alleviating heart disease has been to target individuals and encourage them to change their health behaviours. Although important, emphasis on individuals has diverted focus and responsibility away from neighbourhood characteristics, which also strongly influence people's behaviours. Although a growing body of research has repeatedly demonstrated strong associations between neighbourhood characteristics and cardiovascular health, it has typically focused on negative neighbourhood characteristics. Only a few studies have examined the potential health enhancing effects of positive neighbourhood characteristics, such as perceived neighbourhood social cohesion.
Methods Using multiple logistic regression models, we tested whether higher perceived neighbourhood social cohesion was associated with lower incidence of myocardial infarction. Prospective data from the Health and Retirement Study—a nationally representative panel study of American adults over the age of 50—were used to analyse 5276 participants with no history of heart disease. Respondents were tracked for 4 years and analyses adjusted for relevant sociodemographic, behavioural, biological and psychosocial factors.
Results In a model that adjusted for age, gender, race, marital status, education and total wealth, each SD increase in perceived neighbourhood social cohesion was associated with a 22% reduced odds of myocardial infarction (OR=0.78, 95% CI 0.63 to 0.94. The association between perceived neighbourhood social cohesion and myocardial infarction remained even after adjusting for behavioural, biological and psychosocial covariates.
Conclusions Higher perceived neighbourhood social cohesion may have a protective effect against myocardial infarction.
- Epidemiology of Ageing
- Epidemiology of Cardiovascular Disease
- Psychosocial Factors
- Public Health
Statistics from Altmetric.com
Heart disease has been the leading cause of death in the USA for the past 80 years.1 The main strategy for alleviating heart disease has been trying to convince individuals to change their health behaviours. Although important, emphasis on individuals has diverted focus and responsibility away from higher order factors, such as neighbourhood-level factors. A growing body of literature suggests that neighbourhood characteristics impact cardiovascular health.2 Research examining neighbourhood factors and health, however, has historically emphasised how negative neighbourhood characteristics (eg, density of fast food restaurants, violence, noise, traffic, poor air quality, vandalism, drug use and physical decay) harm health.2 ,3 Only a few studies have examined the possible role positive neighbourhood characteristics, such as neighbourhood social cohesion, have in enhancing health.4
Neighbourhood social cohesion is the perceived degree of connectedness between and among neighbours and their willingness to intervene for the common good.5 It is also characterised by the degree to which a resident: feels secure, feels connected to the area and trusts its inhabitants. The construct is distinct from individual-level social networks and support because it characterises the entire community and impacts the whole neighbourhood, regardless of an individual's characteristics.2
Two pioneering studies explored the association between positive neighbourhood social climate and cardiovascular events. In one study, approximately 7800 Swedish adults over the age of 45 were tracked for 4 years. Higher neighbourhood social climate was linked with lower rates of myocardial infarction.6 The study, however, used a one-item measure of neighbourhood social interaction that may not have captured the multifaceted nature of the construct of interest in this study. Another study followed 2.8 million Swedish adults (aged 45–74) for 2 years. The researchers found that lower neighbourhood social capital was associated with higher incidence of coronary heart disease.7 However, neighbourhood social capital, was operationalised as the percentage of people in a neighbourhood that voted, and also may not have accurately captured the construct of interest in this study.
Although the exact mechanisms responsible for the associations between neighbourhood social cohesion and enhanced cardiovascular health are unknown, studies have linked neighbourhood social cohesion with intermediate outcomes that predict cardiovascular events. For example, higher neighbourhood social cohesion is associated with more physical activity,8 ,9 such as walking,10 and less coronary artery calcification.11 Neighbourhood social cohesion has also been linked with reduced risk of related outcomes such as stroke.12 ,13
We built on the important research of the two studies examining the link between positive neighbourhood social climate and cardiovascular events by using a four-item measure of neighbourhood social cohesion that was carefully constructed based on empirical, conceptual and theoretical evidence. We also controlled for a more comprehensive array of covariates including those that were sociodemographic, behavioural, biological and psychosocial in nature. Several individual-level psychological factors that may distort a person's perception of neighbourhood social cohesion were also controlled for (eg, a depressed person may artificially decrease their neighbourhood social cohesion rating because of their condition). Also, psychological factors are important to control for because several have been linked with an altered risk of cardiovascular events.14–17 We further controlled for two measures of individual-level social engagement because these factors may confound neighbourhood social cohesion ratings.
We also shifted the focus from aggregate levels of neighbourhood social cohesion to individual perceptions of neighbourhood social cohesion. Neighbourhood social cohesion is often measured as an aggregated group indicator or analysed using multilevel modelling. There are several good reasons to do so, but there are also benefits to examining this construct at the individual level. For example, the boundaries of a neighbourhood are difficult to identify. They are often identified by researchers (using Zip codes, census tracts or census block groups) but often differ from a resident's perception of neighbourhood boundaries.18 ,19 Therefore, misidentifying neighbourhoods in a multilevel model, and clustering together respondents who do not consider themselves neighbours, may skew results. Researchers discuss how neighbourhood-level measures of social cohesion often have poor agreement among inhabitants in the same neighbourhood.18 ,20 Further, aggregated neighbourhood-level data would have required a larger number of respondents in each neighbourhood than were available in our sample.
On the basis of our question of interest, conceptual grounds, practical reasons and the merits of supplementing one type of modelling (typically multilevel models in this line of research) with other types of justifiable modelling, we named our predictor variable perceived neighbourhood social cohesion. We use this term to indicate that neighbourhood social cohesion was investigated at the individual level, rather than at the aggregated neighbourhood level.
In this study, we hypothesised that higher perceived neighbourhood social cohesion would be prospectively associated with a lower risk of myocardial infarction. Because the risk of cardiovascular events increases with age, the examination of factors associated with cardiovascular health is particularly important for the expanding segment of older adults facing the threat of declining health and rising healthcare costs. Therefore, we used the Health and Retirement Study (HRS) to test our hypothesis.
The HRS began in 1992 and has surveyed more than 22 000 people biannually since then. It is a nationally representative panel study of American adults over the age of 50 and the age of respondents in this study ranged from 51 to 105. In 2006, the HRS added a detailed section that assessed several psychosocial factors. Therefore, we used 2006 as our baseline and that is also when all of the covariate data were assessed. Incident myocardial infarction was assessed in follow-up waves 2008, 2010 and exit surveys. For respondents who died during the follow-up period, exit interviews were completed by knowledgeable informants. The University of Michigan's Institute for Social Research is responsible for the study and provides extensive documentation about the protocol, instrumentation, sampling strategy and statistical weighting procedures.21 HRS is sponsored by the National Institute on Aging (grant number NIA U01AG009740) and is conducted by the University of Michigan.21 ,22 The present study used de-identified and publicly available data, therefore the Institutional Review Board at the University of Michigan exempted it from review.
In 2006, approximately 50% of the HRS respondents were selected for an enhanced face-to-face interview. At the end of the interview, respondents were asked to complete a self-report leave-behind psychological questionnaire, which they then returned by mail. Among people who were interviewed, the response rate for the leave-behind questionnaire was 90%, resulting in 7168 respondents. We excluded 1816 participants who self-reported a history of heart disease at baseline, and 76 participants who self-identified themselves in a race/ethnicity category other than Caucasian, Black or Hispanic. These 76 participants were dropped because there were not enough cases of myocardial infarction to power the analyses for this group. The final sample consisted of 5276 respondents.
Myocardial infarction outcome measurement
Incidence of myocardial infarction was defined as a first non-fatal or fatal myocardial infarction based on self-report or proxy-report of a physician's diagnosis using the 2008, 2010 and exit surveys. Myocardial infarctions that are assessed through self-report correspond imperfectly with medical records. Although imperfect, the high agreement between self-reported myocardial infarctions and hospital records has been well documented.23–26 For example, in a recent longitudinal study of 41 438 Spanish adults, self-reported myocardial infarctions were compared against hospital records. The sensitivity of self-reported acute myocardial infarction was 97.7%, with a positive predictive value of 60.7%, and a specificity of 99.7%.23 Also, self-report data are particularly precise for acute events, like myocardial infarction.24
Perceived neighbourhood social cohesion measurement
Perceived neighbourhood social cohesion was measured using a four-item scale developed and tested for use in two nationally representative studies of older adults (the HRS and the English Longitudinal Study of Aging).27 The items were derived from widely used and widely cited neighbourhood cohesion scales that have been validated.10 ,28–30 The scale assesses the perceived social cohesion and perceived social trust of the respondent's neighbourhood. Using a seven-point Likert scale, respondents specified the degree to which they endorsed the following four items: ‘I really feel part of this area’, ‘If you were in trouble, there are lots of people in this area who would help you’, ‘Most people in this area can be trusted’ and ‘Most people in this area are friendly’. The scores were averaged and higher scores reflected higher perceived neighbourhood social cohesion (Cronbach's α=0.83). The perceived neighbourhood social cohesion scores were then standardised (M=0, SD=1) so that the outcome OR could be interpreted as the result of one SD increase in perceived neighbourhood social cohesion.
Potential confounders or pathways linking perceived neighbourhood social cohesion with risk of myocardial infarction were all assessed at baseline. Potential covariates included sociodemographic, behavioural, biological and psychosocial factors that past research suggests are relevant to myocardial infarction risk.15 ,31 Potential confounders included the following sociodemographic variables: age, gender, race/ethnicity (Caucasian, African-American, Hispanic and Other), marital status (married/not married), educational attainment (no degree, GED or high school diploma, college degree or higher) and total wealth (<25 000; 25 000–124 999; 125 000–299 999; 300 000–649 999; >650 000—based on quintiles of the score distribution in this sample).
Psychosocial factors that might confound the primary association of interest included depression, anxiety, cynical hostility, positive affect, social participation and social integration. Further information about how these constructs were measured can be found in the online supplementary methods and the HRS Psychosocial Manual.27
Potential behavioural and biological pathway covariates that might link neighbourhood social cohesion to myocardial infarction were also considered. Behavioural covariates included smoking status (never, former and current), frequency of moderate (eg, gardening, dancing, walking at a moderate pace) and vigorous exercise (eg, running, swimming and aerobics) reported as (never, 1–4 times per month, more than once a week), and frequency of alcohol consumption (abstinent, less than 1 or 2 days/month, 1–2 days/week and more than 3 days/week).
Biological covariates included self-reported weight in pounds, converted into kilograms and height in inches, converted into metres (used to calculate body mass index (BMI) according to kg/m2); hypertension and diabetes (each yes/no based on self-report of a doctor's diagnosis). BMI was categorised as 18.5–24.9 (normal), 25–29.9 (overweight), ≥30 (obese). There is no ‘underweight’ category because it contained only 1.51% of the sample and was unstable in statistical analyses, therefore it was collapsed with the ‘normal’ category.
We conducted multiple logistic regression analyses to test if perceived neighbourhood social cohesion was associated with a reduced risk of myocardial infarction. Logistic regression was used because we did not have detailed information about the date each myocardial infarction occurred. ORs, however, offer a good approximation of HRs in this study for several reasons: the follow-up time was short, the outcome incidence ratio was low (probability of myocardial infarction was 2.81% in our sample),32 the risk ratio was moderate in size and the sample size was large. The potential impact of covariates on the relationship between perceived neighbourhood social cohesion and myocardial infarction was estimated by adjusting for blocks of covariates.
We first examined a minimally adjusted model and then considered the impact that adding potentially confounding demographic factors had on the association between perceived neighbourhood social cohesion and myocardial infarction. We then considered the impact of biological or behavioural pathways in a third and a fourth model. In models 3 and 4, an observed reduction in the association between perceived neighbourhood social cohesion and myocardial infarction, after adding either biological or behavioural covariates, may be consistent with the possibility that each block of variables represents a potential pathway linking perceived neighbourhood social cohesion to risk of myocardial infarction. Model 1 adjusted for only age and gender. Model 2, the core model, included: age, gender, race/ethnicity, marital status, educational degree and total wealth. Three additional models were created; model 3—core model+health behaviours (smoking status, exercise and alcohol frequency) and model 4—core model+biological factors (hypertension, diabetes and BMI).
Some additional analyses were performed. First, we examined if associations found between perceived neighbourhood social cohesion and myocardial infarction were maintained even when controlling for depression, anxiety, cynical hostility, optimism, positive affect and two factors that tapped into individual-level social engagement (social participation and social integration). Using the core model, we added each psychosocial factor one at a time. Then three additional models were created: model 5—core model+negative psychological factors (depression, anxiety and cynical hostility); model 6—core model+positive psychological factors (optimism and positive affect); model 7—core model+individual-level social engagement factors (social participation and social integration) and finally a model 8, which included all covariates.
In addition, we created quartiles of perceived neighbourhood social cohesion based on the score distribution in this sample, in order to consider the possibility of threshold or discontinuous effects. Finally, we tested a potential interaction between perceived neighbourhood social cohesion and gender to assess possible gender differences in the association of interest.
Logits were converted into ORs for ease of interpretation. Given that the probability of myocardial infarction was rare in our sample (2.81%), our reported ORs can be regarded as relative risks.33 All results in this study were weighted, using HRS sampling weights to account for the complex multistage probability survey design, which includes individual non-response, stratification, sample clustering and additional post-stratification using Stata.34
Missing data analysis
The item non-response rate was less than 1% for all study variables. These missing data, however, were distributed across variables, resulting in a 5.29% loss of respondents when complete-case analyses were attempted. Therefore, to examine the impact of missing data on our results and to obtain less biased estimates, multiple imputation procedures were used to impute missing data.3,5 Results were largely the same between the original and imputed data sets. Therefore, we used the data set with multiple imputations for all analyses reported in this study.3,5
The average age of respondents at baseline was 70 years (SD=10.05)—ages ranged from 51 to 105 years (47% of the sample was 70 or older). The majority of respondents were women (62%) and married (62%). Most had a high school degree (55%) or attended some college (20%). Respondents identified as being Caucasian (70%), African-American (17%) or Hispanic (12%). Among the 5276 participants, 148 respondents (66 women and 82 men) had a myocardial infarction over the 4-year follow-up. Table 1 contains further descriptive statistics about the covariates. Online supplementary table S1 shows the correlations among the continuous and binary factors in our study, and online supplementary table S2 shows the distribution of perceived neighbourhood social cohesion scores among the categorical factors.
Perceived neighbourhood social cohesion and myocardial infarction incidence
Associations between perceived neighbourhood social cohesion and myocardial infarction were highly consistent across all five models. In the core model (model 2, table 2), each SD increase in perceived neighbourhood social cohesion was associated with a multivariate-adjusted OR of 0.78 for myocardial infarction (95% CI 0.63 to 0.94). When considering each block of potential pathway covariates, the association between perceived neighbourhood social cohesion and myocardial infarction were somewhat attenuated, but the association remained significant or marginally significant in all the models (models 2–8, table 2). See online supplementary table S3 for more detailed information about these results.
When examining quartiles of perceived neighbourhood social cohesion, the findings suggested a threshold relationship (table 3). For example, in the core model (model 2, table 3) relative to people with the lowest neighbourhood cohesion, people with moderately low neighbourhood cohesion had a somewhat reduced risk of myocardial infarction (OR=0.66, 95% CI 0.35 to 1.26), while people with moderately high neighbourhood cohesion had an even lower risk of myocardial infarction (OR=0.56, 95% CI 0.30 to 1.05). However, people with the highest neighbourhood cohesion did not have a substantially lower risk of myocardial infarction when compared against people with moderately high neighbourhood cohesion (OR=0.55, 95% CI 0.32 to 0.96). Adjusting for additional covariates made the associations between perceived neighbourhood social cohesion and myocardial infarction marginally significant in some models (table 3).
Considering additional psychosocial factors
Adding each psychosocial factor sequentially to the base model contributed to a modest decrease in the association between perceived neighbourhood social cohesion and myocardial infarction but the association between perceived neighbourhood social cohesion and myocardial infarction remained significant in each of these analyses. When all three negative psychological factors were simultaneously added to the core model, the association between perceived neighbourhood social cohesion and myocardial infarction was marginally significant (OR=0.79, 95% CI 0.63 to 1.00; model 5, table 2). The association between perceived neighbourhood social cohesion and myocardial infarction also remained significant when positive psychological factors (OR=0.79, 95% CI 0.65 to 0.96; model 6, table 2) or individual-level social engagement factors (OR=0.78, 95% CI 0.64 to 0.96; model 7, table 2) were added to the core model. A fully adjusted model that controlled for every covariate showed a marginally significant association (OR=0.82, 95% CI 0.66 to 1.02; model 8, table 2). Finally, a potential interaction between perceived neighbourhood social cohesion and myocardial infarction was formally tested and the result was not significant (p=0.469).
In a prospective and nationally representative sample of 5276 US adults over the age of 50, who had no history of heart disease at baseline, perceived neighbourhood social cohesion was associated with a reduced likelihood of incident myocardial infarction over the 4-year follow-up period. After adjusting for sociodemographic factors, each SD increase in perceived neighbourhood social cohesion was associated with a 22% reduced risk in incident myocardial infarction. Even after further adjustments for behavioural, biological and psychosocial factors the association between perceived neighbourhood social cohesion and myocardial infarction persisted. Our results are consistent with previous studies that found associations between positive neighbourhood social climate and cardiovascular events.6 ,7 Although adjusting for potential behavioural and biological pathway variables attenuated the parameter estimates, the magnitude of attenuation was modest. This suggests that other mechanisms may be at work. Further, we observed a threshold relationship. After a moderately high amount of perceived neighbourhood social cohesion, additional neighbourhood cohesion did not appear to reduce the risk of myocardial infarction.
Several mechanisms may explain the link between neighbourhood social cohesion and cardiovascular health. Studies consistently report links between higher individual-level social support and better health outcomes, and perceived neighbourhood social cohesion may work through similar mechanisms. For instance, greater perceived social support—one's perception of access to social support—has been linked with better cardiovascular health.3,6–,40 Perceived neighbourhood social cohesion could be a type of social support that is available in the neighbourhood social environment outside the realm of family and friends. Further, this additional type of neighbourhood-level social support may create and reinforce neighbourhood norms. These norms may then impact the behaviour of neighbourhood residents by creating a system of incentives for adopting and maintaining certain behaviours.2 ,10 ,4,1 Further research examining the potential mechanisms between neighbourhood-level factors and cardiovascular health are needed.
Our study has limitations and strengths. Some risk factors, such as family history of cardiovascular disease and genetic vulnerability were not available for analysis. We were also unable to examine many ethnic minority groups (other than Blacks or Hispanics) due to sample size issues. Studies examining the relationship between neighbourhood cohesion and health across different ethnic groups report mixed results.4,2 ,4,3 Therefore, the associations examined in this study should be further researched in more diverse populations, particularly because such a small percentage of the psychosocial literature addresses ethnic minority groups sufficiently.4,4 Further, we do not know how long a respondent has been living in a particular neighbourhood. A respondent could have moved from a low to highly cohesive neighbourhood (or vice versa) and such a move would impact the associations examined in this study. Future research should examine this issue further. Additionally, myocardial infarctions that are assessed through self-report correspond imperfectly with medical records. Although imperfect, the high correlation between self-reported myocardial infarctions and hospital records has been well documented and self-reported data are particularly precise for acute events, such as myocardial infarction.23–26 Finally, only 4 years of HRS follow-up data were available at the time of analysis. The risk of myocardial infarction increases rapidly with age and 47% our study sample was 70 years old and over. Therefore, we thought 4 years was a long enough follow-up period for this study. However, the mechanisms that link perceived neighbourhood social cohesion with myocardial infarction likely develop over the course of many years, and we recognise that a longer follow-up is ideal. Further, the strength of the association between perceived neighbourhood social cohesion and myocardial infarction may change over a longer follow-up period. Therefore this study should be replicated with longer follow-up periods. With time, future waves of data collection and future releases of HRS data will allow researchers to examine longer follow-up periods.
Despite these limitations, our study has several strengths. First, we controlled for a wide range of important covariates that were sociodemographic, behavioural, biological and psychosocial in nature. Second, we used a large nationally representative sample. Third, we controlled for several psychosocial factors that may skew a person's neighbourhood social cohesion ratings. Finally, we tried minimising the potential impact of missing data by using a multiple imputation technique, which has been shown to provide more accurate estimates of associations than other methods of handling missing data.3,5 If future work replicates our findings, this line of research may justify future research which examines the potential health benefits of policy and public-health interventions that bolster the social infrastructure of neighbourhoods.
What is already known on this subject
Past research examining the associations between neighbourhood-level factors and health, has largely focused on how negative neighbourhood factors are associated with poorer health.
However, a growing body of research shows that positive neighbourhood-level factors, such as neighbourhood social cohesion, is associated with an array of positive outcomes including better: mental health, health behaviours, and physical health.
What this study adds
To the best of our knowledge, this is the first study to prospectively examine the association between perceived neighbourhood social cohesion and myocardial infarction.
Higher perceived neighbourhood social cohesion was associated with lower myocardial infarction risk, even after adjusting for a wide range of covariates.
If future research replicates these findings, more neighbourhood-level public health approaches that target neighbourhood social cohesion may be warranted.
The authors would like to thank the editors and the anonymous reviewers for their valuable comments and suggestions.
Contributors ESK had full access to all data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. All authors contributed to the study design and interpretation of the findings, and have read, commented on and approved the manuscript.
Competing interests None.
Patient consent Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The authors agree to allow the journal and any researcher to review the data if requested. The data set we used for this study (The Health and Retirement study) is also free and publicly available.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.