Abstract P368: Neighborhoods and Subclinical Cardiovascular Disease

Abstract

Background: Research on the health impact of the built environment has focused on health behaviors such as diet and exercise, and conditions such as obesity and diabetes. Few studies have examined its influence on downstream outcomes such as cardiovascular disease. We investigated the proportional variance in the 10-year and 30-year Framingham risk score (FRS) attributable to neighborhoods in the Framingham Heart Study. Methods: Offspring- and Generation 3 cohort members’ homes at the time of exam 7 (Offspring, 1998-2001) or exam 1 (Generation 3, 2002-2005) were geocoded to 2000 Census block groups. We evaluated Framingham Offspring and Generation 3 cohort participants inhabiting private residences in block groups within Massachusetts containing the residences of 5 or more participants. Analyses of the 10-year FRS were further restricted to participants aged 30-74 at the time of the relevant exam and those of the 30-year FRS to participants 20-59 years old. Cardiovascular risk was determined on the bases of sex, age, systolic blood pressure, anti-hypertensive medication, smoking, diabetes, total cholesterol, and HDL. The outcomes were the standardized residuals of log-transformed FRS regressed on age and sex. We analyzed the percentage of variance of FRS explained at the block-group level and 95% confidence intervals using multilevel linear regression. An empty model was first used to estimate the total variance and the following factors were then added singly to evaluate their influence on the group-level FRS variance explained by education, body mass index, waist circumference, physical activity score, and depression (CES-D score ≥16). Analyses were repeated stratified by sex. Results: The analysis of 10-year FRS included a total of 2,882 participants in 188 census block groups. The block-group-level variance explained for this outcome was 1.77% (95% CI=0.69, 4.44). Upon the addition of BMI to the model, the variance explained dropped to 1.11% (95% CI=0.28%, 4.30%). None of the other covariates had a substantial impact. Among 1,363 women in 117 block groups, the block level group explained a total of 2.03% of the FRS variance at the block level group (95% CI=0.61, 6.56), which dropped to 0.64% (95% CI=0.03, 13.82) when BMI was added to the model. Results were somewhat stronger in analysis of the 30-year FRS. The group-level FRS variance explained by census blocks was 3.56% (95% CI=1.75, 7.12) among 2096 participants in 156 neighborhoods. Similar to the 10-year CVD risk score, the variance explained among women (959 in 97 block groups) was higher (6.06%, 95% CI=2.75, 12.83), but null among males. Conclusions: In this relatively homogenous suburban white population, census block groups explained a small percentage of the variance in CVD risk. The explained variance was higher among women (19% non-working vs. 5% of males), and largely explained by the clustering of obesity.